{"schema":"5lc-index-ledger/1","key":"infrastructure-servers-systems","layer":"infrastructure","sublayer":"servers-systems","asOf":"2026-10-22","about":"5LC Importance Index ledger for one sublayer: the measure card, the current ten with any editorial overrides and their reasons, the computed result for every scored company (Importance = share of AI output × years to replace, Reach, replacement years, Rising, the Monte Carlo rank simulation and its stability class, validation, the revenue baseline and rater agreement), each scored company's sourced inputs and irreplaceability ratings, and the sublayer's validation cases with their sources. Monte Carlo frequencies describe the stated input ranges, not true probabilities. Omitted: the free-text notes of the internal check passes and of rating reconciliations, and the location of the internal card review; references to local evidence files and to internal method decisions are replaced by plain descriptions.","roster":{"members":[{"entityId":"hon-hai","slot":"importance","since":"2026-10-22"},{"entityId":"quanta-computer","slot":"importance","since":"2026-10-22"},{"entityId":"dell","slot":"importance","since":"2026-10-22"},{"entityId":"wistron","slot":"importance","since":"2026-10-22"},{"entityId":"super-micro","slot":"importance","since":"2026-10-22"},{"entityId":"inspur","slot":"importance","since":"2026-10-22"},{"entityId":"huawei","slot":"importance","since":"2026-10-22"},{"entityId":"lenovo","slot":"importance","since":"2026-10-22"},{"entityId":"gigabyte","slot":"importance","since":"2026-10-22"},{"entityId":"asus","slot":"importance","since":"2026-10-22"}],"pending":{},"lastRun":"2026-10-22","resultsDate":"2026-10-22"},"card":{"schema":"5lc-index-card/1","key":"infrastructure-servers-systems","layer":"infrastructure","sublayer":"servers-systems","version":1,"scope":"Servers and rack-scale systems built around GPUs or other AI accelerators, credited to the company that sells the finished system to the data-centre operator, whether a branded system maker or a manufacturer selling directly to cloud companies; accelerators, CPUs and the boards and modules their designers sell belong to Chips, switches and cabling to Networking, power shelves and cold plates to Power & cooling, and general-purpose and storage servers are outside the measure.","outputMeasure":"AI servers and rack-scale systems delivered to data centres, by sales value, weighted by AI server sales.","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","hardToReplace":"Rack-scale systems need qualified manufacturing, liquid-cooling integration and full-rack testing that few makers provide at volume. Standard GPU servers have many qualified makers.","functions":[{"id":"gpu-systems","name":"GPU servers and racks outside mainland China","why":"Most AI computing runs on GPU servers and racks; about three-quarters of AI server sales.","weight":{"low":0.68,"central":0.74,"high":0.79},"weightSource":{"url":"https://www.idc.com/resource-center/blog/ai-infrastructure-spending-holds-near-90-billion-in-q1-2026-as-arm-overtakes-x86-in-accelerated-servers-2026-forecast-raised-to-497-billion/","title":"AI Infrastructure Spending Holds Near $90 Billion in Q1 2026 as ARM Overtakes x86 in Accelerated Servers; 2026 Forecast Raised to $497 Billion","publisher":"IDC","publishedDate":"2026-07-21","accessedDate":"2026-10-22","tier":"T4","locator":"Accelerated (AI) servers by quarter: 3Q25 $81.7B (ARM $29.8B + x86 $51.9B) and 4Q25 $90.2B ($47.5B + $42.7B) from this release; 1Q26 GPU $68.9B and other $17.7B (IDC, 15 Jun 2026); 2Q26 GPU $87.4B and other $27.5B (IDC, 'Worldwide Server Market Revenue Hits Record $166.3 Billion in the Second Quarter of 2026', 10 Sep 2026). 3Q25 and 4Q25 split by interpolating the other-accelerated share between 2Q25 (10.7%) and 1Q26 (20.4%). Four quarters: GPU $301.3B, other $72.1B, total $373.4B; China's 9% removed at 70% GPU and 30% other: GPU outside China $277.8B, 74%. Range: other-accelerated share in 3Q25–4Q25 between 10.7% and 20.4%, China 7–12%, China GPU share 60–80%."},"shareData":"Company GPU server and rack revenue outside China over IDC's total (IDC, filings); NVIDIA only for complete systems it sells (DGX)"},{"id":"custom-accelerator-systems","name":"Servers for cloud companies' own AI chips","why":"Servers built to order around chips such as Google TPUs and AWS Trainium; about a sixth of sales.","weight":{"low":0.12,"central":0.17,"high":0.21},"weightSource":{"url":"https://www.idc.com/resource-center/press-releases/1q26-server-tracker/","title":"Worldwide Server Market Revenue Surpasses $122 Billion in the First Quarter of 2026, Driven by AI Infrastructure Demand, according to IDC","publisher":"IDC","publishedDate":"2026-06-15","accessedDate":"2026-10-22","tier":"T4","locator":"'Other Accelerated servers surged 122.1% YoY to $17.7 billion' (1Q26); 2Q26 $27.5B (+237.6%) from IDC's 10 Sep 2026 release; 3Q25 and 4Q25 interpolated ($11.4B and $15.5B). Four quarters $72.1B less China's $10.1B: $62.0B of $373.4B, 17%. Range from the same interpolation and China bounds."},"shareData":"Company revenue from servers for other accelerators outside China over IDC's total (filings)"},{"id":"china","name":"AI servers sold in mainland China","why":"A separate market under export controls, with domestic makers and chips; about a tenth of sales.","weight":{"low":0.07,"central":0.09,"high":0.12},"weightSource":{"url":"https://www.idc.com/resource-center/blog/ai-infrastructure-spending-holds-near-90-billion-in-q1-2026-as-arm-overtakes-x86-in-accelerated-servers-2026-forecast-raised-to-497-billion/","title":"AI Infrastructure Spending Holds Near $90 Billion in Q1 2026 as ARM Overtakes x86 in Accelerated Servers; 2026 Forecast Raised to $497 Billion","publisher":"IDC","publishedDate":"2026-07-21","accessedDate":"2026-10-22","tier":"T4","locator":"'China (PRC) reached $7.8 billion, an 8.7% share' of AI infrastructure in 1Q26. Matched first-half 2025 check: IDC China's accelerated servers 'US$16.0B' (9 Oct 2025) against worldwide accelerated servers of $139.6B (1Q25 $63.2B and 2Q25 $76.4B, from the growth rates in IDC's 1Q26 and 2Q26 releases), 11.5%. Central 9% for 3Q25–2Q26, between the two; range 7–12%."},"shareData":"Company share of China's accelerated server market (IDC China)"}],"risingFloorPct":1,"validationCaseIds":[],"knownGaps":["IDC does not split 3Q25 and 4Q25 AI servers by chip type; those two quarters are interpolated between 2Q25 and 1Q26.","China's share comes from IDC's AI infrastructure total for 1Q26 and a matched first-half 2025 comparison, not from IDC's server tracker by chip type.","Manufacturers that build systems sold under another company's brand are not credited for that work, because each system is credited once, to its seller.","IDC reports direct sales by manufacturers to cloud companies as one total; each manufacturer's AI server revenue comes from filings that mix AI servers with other products.","When a cloud company buys accelerators itself and supplies them to the manufacturer, the manufacturer's revenue is lower than IDC's system value.","Fiscal years differ (Dell, HPE, Supermicro, Lenovo), so inputs use the latest four reported quarters.","System prices include the accelerators inside, so sales shares follow the value of chips passed through as well as the system maker's own work.","IDC does not split GPU systems into rack-scale systems and standard servers."],"approval":{"approver":"Claude (delegated by Dachi)","date":"2026-10-22","sha256":"d0c7a04c2aaf8e41973e41351e9d8f7cbc7d69d233cc6ad9f379697e80d829e8","delegatedOn":"2026-10-22"}},"result":{"schema":"5lc-index-result/1","key":"infrastructure-servers-systems","asOf":"2026-10-22","methodVersion":"1.1","methodSha256":"3cabed6b38514c263d8b86a494e8cc19cdbd0b2a1013543dc024192c49f7a6a6","cardSha256":"d0c7a04c2aaf8e41973e41351e9d8f7cbc7d69d233cc6ad9f379697e80d829e8","validationSha256":"409931bb458f38b02342c9bf43e61eab299e47d8679912af2c96074ec556f17e","inputsSha256":"9d0492f0b91401fdf82c223889c8c60516436d9a7d8390fd9a826975e8d3c829","unscreened":[],"companies":[{"entityId":"hon-hai","name":"Hon Hai (Foxconn)","reachPct":19.03327,"importance":12.4375941131,"importanceRank":1,"functions":[{"id":"gpu-systems","weight":0.74,"sharePct":21.632,"shareLowPct":9.456,"shareHighPct":37.543,"stale":false,"offBasis":false,"level":2.33333333333,"years":0.6299605249459811,"disagreement":true},{"id":"custom-accelerator-systems","weight":0.17,"sharePct":17.247,"shareLowPct":4.752,"shareHighPct":46.611,"stale":false,"offBasis":false,"level":2.66666666667,"years":0.7937005259859335,"disagreement":false},{"id":"china","weight":0.09,"sharePct":1.04,"shareLowPct":0,"shareHighPct":6.51,"stale":false,"offBasis":false,"level":1.16666666667,"years":0.2806155120779916,"disagreement":false}],"rising":78.5714285714,"risingEligible":true,"risingComponents":{"growth":57.1428571429,"shareChange":100},"mc":{"medianRank":1,"rankP10":1,"rankP90":1,"pTop8":1,"pSelected":1,"stability":"firm"},"baselineRevenueRank":1},{"entityId":"quanta-computer","name":"Quanta Computer (incl. QCT)","reachPct":12.81563,"importance":5.86741657702,"importanceRank":2,"functions":[{"id":"gpu-systems","weight":0.74,"sharePct":14.62,"shareLowPct":7.259,"shareHighPct":21.634,"stale":false,"offBasis":false,"level":1.66666666667,"years":0.3968502629929668,"disagreement":false},{"id":"custom-accelerator-systems","weight":0.17,"sharePct":11.628,"shareLowPct":3.845,"shareHighPct":33.574,"stale":false,"offBasis":false,"level":2.66666666667,"years":0.7937005259859335,"disagreement":false},{"id":"china","weight":0.09,"sharePct":0.223,"shareLowPct":0,"shareHighPct":1.15,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true}],"rising":51.4285714286,"risingEligible":true,"risingComponents":{"growth":42.8571428571,"shareChange":60},"mc":{"medianRank":3,"rankP10":2,"rankP90":4,"pTop8":1,"pSelected":1,"stability":"firm"},"baselineRevenueRank":3},{"entityId":"dell","name":"Dell Technologies","reachPct":12.63405,"importance":5.61726543709,"importanceRank":3,"functions":[{"id":"gpu-systems","weight":0.74,"sharePct":17,"shareLowPct":15.2,"shareHighPct":19.2,"stale":false,"offBasis":false,"level":1.83333333333,"years":0.4454493590691404,"disagreement":false},{"id":"custom-accelerator-systems","weight":0.17,"sharePct":0.121,"shareLowPct":0,"shareHighPct":0.656,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true},{"id":"china","weight":0.09,"sharePct":0.372,"shareLowPct":0,"shareHighPct":1.92,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true}],"rising":93.3333333333,"risingEligible":true,"risingComponents":{"growth":100,"shareChange":80,"committed":100},"mc":{"medianRank":3,"rankP10":2,"rankP90":4,"pTop8":1,"pSelected":1,"stability":"firm"},"baselineRevenueRank":4},{"entityId":"wistron","name":"Wistron (group, incl. Wiwynn)","reachPct":9.32114,"importance":4.64382403514,"importanceRank":4,"functions":[{"id":"gpu-systems","weight":0.74,"sharePct":9.342,"shareLowPct":4.355,"shareHighPct":18.615,"stale":false,"offBasis":false,"level":1.66666666667,"years":0.3968502629929668,"disagreement":false},{"id":"custom-accelerator-systems","weight":0.17,"sharePct":14.047,"shareLowPct":4.001,"shareHighPct":42.824,"stale":false,"offBasis":false,"level":2.66666666667,"years":0.7937005259859335,"disagreement":false},{"id":"china","weight":0.09,"sharePct":0.223,"shareLowPct":0,"shareHighPct":1.15,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true}],"rising":55.7142857143,"risingEligible":true,"risingComponents":{"growth":71.4285714286,"shareChange":40},"mc":{"medianRank":3,"rankP10":2,"rankP90":4,"pTop8":1,"pSelected":1,"stability":"firm"},"baselineRevenueRank":2},{"entityId":"super-micro","name":"Supermicro","reachPct":8.03421,"importance":2.83561479106,"importanceRank":5,"functions":[{"id":"gpu-systems","weight":0.74,"sharePct":10.793,"shareLowPct":9.838,"shareHighPct":12.277,"stale":false,"offBasis":false,"level":1.5,"years":0.3535533905932738,"disagreement":false},{"id":"custom-accelerator-systems","weight":0.17,"sharePct":0.121,"shareLowPct":0,"shareHighPct":0.656,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true},{"id":"china","weight":0.09,"sharePct":0.298,"shareLowPct":0,"shareHighPct":1.53,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true}],"rising":32.8571428571,"risingEligible":true,"risingComponents":{"growth":28.5714285714,"shareChange":20,"committed":50},"mc":{"medianRank":5,"rankP10":5,"rankP90":5,"pTop8":1,"pSelected":1,"stability":"firm"},"baselineRevenueRank":5},{"entityId":"inspur","name":"Inspur Information (IEIT Systems)","reachPct":3.56094,"importance":1.32457468194,"importanceRank":6,"functions":[{"id":"gpu-systems","weight":0.74,"sharePct":2.09,"shareLowPct":0.802,"shareHighPct":3.577,"stale":false,"offBasis":false,"level":1.16666666667,"years":0.2806155120779916,"disagreement":false},{"id":"custom-accelerator-systems","weight":0.17,"sharePct":0.202,"shareLowPct":0,"shareHighPct":1.09,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true},{"id":"china","weight":0.09,"sharePct":22,"shareLowPct":9.6,"shareHighPct":36.4,"stale":true,"offBasis":false,"level":1.83333333333,"years":0.4454493590691404,"disagreement":false}],"rising":null,"risingEligible":true,"risingComponents":{},"mc":{"medianRank":6,"rankP10":6,"rankP90":7,"pTop8":0.9976,"pSelected":0.9999,"stability":"firm"},"baselineRevenueRank":6},{"entityId":"huawei","name":"Huawei (Atlas and Ascend systems)","reachPct":1.00388,"importance":0.89222,"importanceRank":7,"functions":[{"id":"gpu-systems","weight":0.74,"sharePct":0.09,"shareLowPct":0,"shareHighPct":0.389,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true},{"id":"custom-accelerator-systems","weight":0.17,"sharePct":0.484,"shareLowPct":0,"shareHighPct":3.28,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true},{"id":"china","weight":0.09,"sharePct":9.5,"shareLowPct":3.1,"shareHighPct":18.9,"stale":true,"offBasis":false,"level":3,"years":1,"disagreement":true}],"rising":null,"risingEligible":true,"risingComponents":{"growth":14.2857142857},"mc":{"medianRank":7,"rankP10":6,"rankP90":11,"pTop8":0.7209,"pSelected":0.8609,"stability":"likely"},"baselineRevenueRank":null},{"entityId":"lenovo","name":"Lenovo","reachPct":2.1935,"importance":0.611335188278,"importanceRank":8,"functions":[{"id":"gpu-systems","weight":0.74,"sharePct":1.502,"shareLowPct":1.02,"shareHighPct":2.412,"stale":false,"offBasis":false,"level":1.16666666667,"years":0.2806155120779916,"disagreement":false},{"id":"custom-accelerator-systems","weight":0.17,"sharePct":0.806,"shareLowPct":0,"shareHighPct":2.609,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true},{"id":"china","weight":0.09,"sharePct":10.5,"shareLowPct":4,"shareHighPct":16,"stale":false,"offBasis":false,"level":1.16666666667,"years":0.2806155120779916,"disagreement":false}],"rising":null,"risingEligible":true,"risingComponents":{},"mc":{"medianRank":8,"rankP10":7,"rankP90":10,"pTop8":0.5401,"pSelected":0.9139,"stability":"firm"},"baselineRevenueRank":7},{"entityId":"gigabyte","name":"GIGABYTE (Giga Computing)","reachPct":1.75949,"importance":0.490009074889,"importanceRank":9,"functions":[{"id":"gpu-systems","weight":0.74,"sharePct":2.213,"shareLowPct":1.453,"shareHighPct":3.145,"stale":false,"offBasis":false,"level":1.16666666667,"years":0.2806155120779916,"disagreement":false},{"id":"custom-accelerator-systems","weight":0.17,"sharePct":0.323,"shareLowPct":0,"shareHighPct":1.75,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true},{"id":"china","weight":0.09,"sharePct":0.744,"shareLowPct":0,"shareHighPct":3.83,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true}],"rising":null,"risingEligible":true,"risingComponents":{},"mc":{"medianRank":10,"rankP10":8,"rankP90":13,"pTop8":0.1636,"pSelected":0.5973,"stability":"borderline"},"baselineRevenueRank":8},{"entityId":"asus","name":"ASUS","reachPct":1.60998,"importance":0.449505118796,"importanceRank":10,"functions":[{"id":"gpu-systems","weight":0.74,"sharePct":2.075,"shareLowPct":1.468,"shareHighPct":2.756,"stale":false,"offBasis":false,"level":1.16666666667,"years":0.2806155120779916,"disagreement":false},{"id":"custom-accelerator-systems","weight":0.17,"sharePct":0.202,"shareLowPct":0,"shareHighPct":1.09,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true},{"id":"china","weight":0.09,"sharePct":0.446,"shareLowPct":0,"shareHighPct":2.3,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true}],"rising":null,"risingEligible":true,"risingComponents":{"growth":85.7142857143},"mc":{"medianRank":11,"rankP10":9,"rankP90":14,"pTop8":0.0468,"pSelected":0.3281,"stability":"borderline"},"baselineRevenueRank":10},{"entityId":"h3c","name":"H3C","reachPct":1.31387,"importance":0.410342630716,"importanceRank":11,"functions":[{"id":"gpu-systems","weight":0.74,"sharePct":0.045,"shareLowPct":0,"shareHighPct":0.194,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true},{"id":"custom-accelerator-systems","weight":0.17,"sharePct":0.121,"shareLowPct":0,"shareHighPct":0.656,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true},{"id":"china","weight":0.09,"sharePct":14,"shareLowPct":6.2,"shareHighPct":22.8,"stale":true,"offBasis":false,"level":1.33333333333,"years":0.3149802624729905,"disagreement":false}],"rising":null,"risingEligible":true,"risingComponents":{},"mc":{"medianRank":13,"rankP10":10,"rankP90":17,"pTop8":0.0269,"pSelected":0.1837,"stability":"borderline"},"baselineRevenueRank":null},{"entityId":"xfusion","name":"xFusion","reachPct":1.11205,"importance":0.344021400028,"importanceRank":12,"functions":[{"id":"gpu-systems","weight":0.74,"sharePct":0.119,"shareLowPct":0.028,"shareHighPct":0.385,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true},{"id":"custom-accelerator-systems","weight":0.17,"sharePct":0.048,"shareLowPct":0,"shareHighPct":0.326,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true},{"id":"china","weight":0.09,"sharePct":11.287,"shareLowPct":7.328,"shareHighPct":18.645,"stale":false,"offBasis":false,"level":1.33333333333,"years":0.3149802624729905,"disagreement":false}],"rising":null,"risingEligible":true,"risingComponents":{},"mc":{"medianRank":14,"rankP10":11,"rankP90":17,"pTop8":0.0066,"pSelected":0.0723,"stability":"borderline"},"baselineRevenueRank":null},{"entityId":"hpe","name":"Hewlett Packard Enterprise","reachPct":1.18616,"importance":0.331610681396,"importanceRank":13,"functions":[{"id":"gpu-systems","weight":0.74,"sharePct":1.548,"shareLowPct":0.906,"shareHighPct":2.294,"stale":false,"offBasis":true,"level":1.16666666667,"years":0.2806155120779916,"disagreement":false},{"id":"custom-accelerator-systems","weight":0.17,"sharePct":0.121,"shareLowPct":0,"shareHighPct":0.656,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true},{"id":"china","weight":0.09,"sharePct":0.223,"shareLowPct":0,"shareHighPct":1.15,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true}],"rising":0,"risingEligible":true,"risingComponents":{"growth":0,"shareChange":0,"committed":0},"mc":{"medianRank":15,"rankP10":12,"rankP90":18,"pTop8":0.0023,"pSelected":0.0345,"stability":"borderline"},"baselineRevenueRank":11},{"entityId":"nvidia","name":"NVIDIA (DGX systems)","reachPct":1.11,"importance":0.311483218407,"importanceRank":14,"functions":[{"id":"gpu-systems","weight":0.74,"sharePct":1.5,"shareLowPct":0.283,"shareHighPct":5.404,"stale":false,"offBasis":false,"level":1.16666666667,"years":0.2806155120779916,"disagreement":false},{"id":"custom-accelerator-systems","weight":0.17,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"china","weight":0.09,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false}],"rising":null,"risingEligible":true,"risingComponents":{},"mc":{"medianRank":11,"rankP10":7,"rankP90":19,"pTop8":0.2485,"pSelected":0.433,"stability":"borderline"},"baselineRevenueRank":null},{"entityId":"celestica","name":"Celestica","reachPct":0.63337,"importance":0.28172,"importanceRank":15,"functions":[{"id":"gpu-systems","weight":0.74,"sharePct":0.18,"shareLowPct":0,"shareHighPct":0.583,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true},{"id":"custom-accelerator-systems","weight":0.17,"sharePct":2.903,"shareLowPct":1.2,"shareHighPct":6.522,"stale":false,"offBasis":false,"level":2,"years":0.5,"disagreement":false},{"id":"china","weight":0.09,"sharePct":0.074,"shareLowPct":0,"shareHighPct":0.383,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true}],"rising":null,"risingEligible":false,"risingComponents":{},"mc":{"medianRank":16,"rankP10":11,"rankP90":20,"pTop8":0.0128,"pSelected":0.0654,"stability":"borderline"},"baselineRevenueRank":null},{"entityId":"inventec","name":"Inventec","reachPct":0.814,"importance":0.2650825,"importanceRank":16,"functions":[{"id":"gpu-systems","weight":0.74,"sharePct":0.749,"shareLowPct":0.199,"shareHighPct":2.724,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true},{"id":"custom-accelerator-systems","weight":0.17,"sharePct":1.449,"shareLowPct":0.288,"shareHighPct":8.56,"stale":false,"offBasis":false,"level":2,"years":0.5,"disagreement":false},{"id":"china","weight":0.09,"sharePct":0.149,"shareLowPct":0,"shareHighPct":0.766,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true}],"rising":null,"risingEligible":false,"risingComponents":{},"mc":{"medianRank":11,"rankP10":8,"rankP90":17,"pTop8":0.229,"pSelected":0.4615,"stability":"borderline"},"baselineRevenueRank":9},{"entityId":"nettrix","name":"Nettrix","reachPct":0.77387,"importance":0.240253288981,"importanceRank":17,"functions":[{"id":"gpu-systems","weight":0.74,"sharePct":0.045,"shareLowPct":0,"shareHighPct":0.194,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true},{"id":"custom-accelerator-systems","weight":0.17,"sharePct":0.121,"shareLowPct":0,"shareHighPct":0.656,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true},{"id":"china","weight":0.09,"sharePct":8,"shareLowPct":3.4,"shareHighPct":16.6,"stale":true,"offBasis":false,"level":1.33333333333,"years":0.3149802624729905,"disagreement":false}],"rising":null,"risingEligible":false,"risingComponents":{},"mc":{"medianRank":17,"rankP10":13,"rankP90":21,"pTop8":0.0008,"pSelected":0.0121,"stability":"borderline"},"baselineRevenueRank":null},{"entityId":"zte","name":"ZTE (servers)","reachPct":0.72782,"importance":0.200702714976,"importanceRank":18,"functions":[{"id":"gpu-systems","weight":0.74,"sharePct":0.145,"shareLowPct":0.032,"shareHighPct":0.421,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true},{"id":"custom-accelerator-systems","weight":0.17,"sharePct":0.048,"shareLowPct":0,"shareHighPct":0.435,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true},{"id":"china","weight":0.09,"sharePct":6.804,"shareLowPct":2.967,"shareHighPct":13.163,"stale":false,"offBasis":false,"level":1.16666666667,"years":0.2806155120779916,"disagreement":false}],"rising":null,"risingEligible":false,"risingComponents":{},"mc":{"medianRank":18,"rankP10":15,"rankP90":21,"pTop8":0,"pSelected":0.0017,"stability":"borderline"},"baselineRevenueRank":null},{"entityId":"jabil","name":"Jabil","reachPct":0.7794,"importance":0.19485,"importanceRank":19,"functions":[{"id":"gpu-systems","weight":0.74,"sharePct":0.829,"shareLowPct":0.274,"shareHighPct":2.129,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true},{"id":"custom-accelerator-systems","weight":0.17,"sharePct":0.929,"shareLowPct":0.077,"shareHighPct":3.757,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true},{"id":"china","weight":0.09,"sharePct":0.089,"shareLowPct":0,"shareHighPct":0.385,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true}],"rising":null,"risingEligible":false,"risingComponents":{},"mc":{"medianRank":17,"rankP10":12,"rankP90":21,"pTop8":0.004,"pSelected":0.0334,"stability":"borderline"},"baselineRevenueRank":null},{"entityId":"sanmina","name":"Sanmina (incl. ZT Systems manufacturing)","reachPct":0.61187,"importance":0.16848,"importanceRank":20,"functions":[{"id":"gpu-systems","weight":0.74,"sharePct":0.734,"shareLowPct":0.195,"shareHighPct":1.436,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true},{"id":"custom-accelerator-systems","weight":0.17,"sharePct":0.365,"shareLowPct":0,"shareHighPct":2.407,"stale":false,"offBasis":false,"level":2,"years":0.5,"disagreement":false},{"id":"china","weight":0.09,"sharePct":0.074,"shareLowPct":0,"shareHighPct":0.385,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true}],"rising":null,"risingEligible":false,"risingComponents":{},"mc":{"medianRank":19,"rankP10":15,"rankP90":22,"pTop8":0.0001,"pSelected":0.0023,"stability":"borderline"},"baselineRevenueRank":null},{"entityId":"mitac","name":"MiTAC (TYAN)","reachPct":0.56057,"importance":0.1401425,"importanceRank":21,"functions":[{"id":"gpu-systems","weight":0.74,"sharePct":0.684,"shareLowPct":0.34,"shareHighPct":1.17,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true},{"id":"custom-accelerator-systems","weight":0.17,"sharePct":0.202,"shareLowPct":0,"shareHighPct":1.09,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true},{"id":"china","weight":0.09,"sharePct":0.223,"shareLowPct":0,"shareHighPct":1.15,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true}],"rising":null,"risingEligible":false,"risingComponents":{},"mc":{"medianRank":21,"rankP10":18,"rankP90":23,"pTop8":0,"pSelected":0,"stability":"borderline"},"baselineRevenueRank":null},{"entityId":"huakun-zhenyu","name":"Huakun Zhenyu","reachPct":0.38043,"importance":0.11850039449,"importanceRank":22,"functions":[{"id":"gpu-systems","weight":0.74,"sharePct":0.009,"shareLowPct":0,"shareHighPct":0.039,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true},{"id":"custom-accelerator-systems","weight":0.17,"sharePct":0.081,"shareLowPct":0,"shareHighPct":0.438,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true},{"id":"china","weight":0.09,"sharePct":4,"shareLowPct":1.5,"shareHighPct":8,"stale":false,"offBasis":false,"level":1.33333333333,"years":0.3149802624729905,"disagreement":false}],"rising":null,"risingEligible":false,"risingComponents":{},"mc":{"medianRank":23,"rankP10":21,"rankP90":25,"pTop8":0,"pSelected":0,"stability":"borderline"},"baselineRevenueRank":null},{"entityId":"flex","name":"Flex","reachPct":0.42942,"importance":0.107355,"importanceRank":23,"functions":[{"id":"gpu-systems","weight":0.74,"sharePct":0.453,"shareLowPct":0.106,"shareHighPct":1.183,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true},{"id":"custom-accelerator-systems","weight":0.17,"sharePct":0.507,"shareLowPct":0.032,"shareHighPct":2.435,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true},{"id":"china","weight":0.09,"sharePct":0.089,"shareLowPct":0,"shareHighPct":0.385,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true}],"rising":null,"risingEligible":false,"risingComponents":{},"mc":{"medianRank":21,"rankP10":18,"rankP90":24,"pTop8":0,"pSelected":0,"stability":"borderline"},"baselineRevenueRank":null},{"entityId":"sugon","name":"Sugon","reachPct":0.26073,"importance":0.0910822385213,"importanceRank":24,"functions":[{"id":"gpu-systems","weight":0.74,"sharePct":0.007,"shareLowPct":0,"shareHighPct":0.012,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true},{"id":"custom-accelerator-systems","weight":0.17,"sharePct":0.032,"shareLowPct":0,"shareHighPct":0.217,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true},{"id":"china","weight":0.09,"sharePct":2.779,"shareLowPct":1.153,"shareHighPct":5.586,"stale":false,"offBasis":false,"level":1.5,"years":0.3535533905932738,"disagreement":false}],"rising":null,"risingEligible":false,"risingComponents":{},"mc":{"medianRank":24,"rankP10":22,"rankP90":26,"pTop8":0,"pSelected":0,"stability":"borderline"},"baselineRevenueRank":null},{"entityId":"pegatron","name":"Pegatron","reachPct":0.28721,"importance":0.0718025,"importanceRank":25,"functions":[{"id":"gpu-systems","weight":0.74,"sharePct":0.358,"shareLowPct":0.139,"shareHighPct":0.694,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true},{"id":"custom-accelerator-systems","weight":0.17,"sharePct":0.084,"shareLowPct":0,"shareHighPct":0.581,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true},{"id":"china","weight":0.09,"sharePct":0.089,"shareLowPct":0,"shareHighPct":0.385,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true}],"rising":null,"risingEligible":false,"risingComponents":{},"mc":{"medianRank":24,"rankP10":23,"rankP90":26,"pTop8":0,"pSelected":0,"stability":"borderline"},"baselineRevenueRank":null},{"entityId":"compal","name":"Compal","reachPct":0.21075,"importance":0.0526875,"importanceRank":26,"functions":[{"id":"gpu-systems","weight":0.74,"sharePct":0.248,"shareLowPct":0.183,"shareHighPct":0.39,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true},{"id":"custom-accelerator-systems","weight":0.17,"sharePct":0.121,"shareLowPct":0,"shareHighPct":0.656,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true},{"id":"china","weight":0.09,"sharePct":0.074,"shareLowPct":0,"shareHighPct":0.383,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true}],"rising":null,"risingEligible":false,"risingComponents":{},"mc":{"medianRank":26,"rankP10":24,"rankP90":26,"pTop8":0,"pSelected":0,"stability":"borderline"},"baselineRevenueRank":null},{"entityId":"cerebras","name":"Cerebras (CS-3 systems)","reachPct":0.10259,"importance":0.0256475,"importanceRank":27,"functions":[{"id":"gpu-systems","weight":0.74,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"custom-accelerator-systems","weight":0.17,"sharePct":0.577,"shareLowPct":0.453,"shareHighPct":0.919,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true},{"id":"china","weight":0.09,"sharePct":0.05,"shareLowPct":0,"shareHighPct":0.15,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true}],"rising":null,"risingEligible":false,"risingComponents":{},"mc":{"medianRank":27,"rankP10":27,"rankP90":27,"pTop8":0,"pSelected":0,"stability":"borderline"},"baselineRevenueRank":null}],"selection":{"importanceIds":["hon-hai","quanta-computer","dell","wistron","super-micro","inspur","huawei","lenovo"],"risingIds":[],"filledFromImportance":["gigabyte","asus"]},"risingPools":{"growth":8,"shareChange":6,"committed":3},"weightSensitivity":[{"functionId":"gpu-systems","end":"low","entered":["h3c"],"left":["asus"]},{"functionId":"gpu-systems","end":"high","entered":[],"left":[]},{"functionId":"custom-accelerator-systems","end":"low","entered":[],"left":[]},{"functionId":"custom-accelerator-systems","end":"high","entered":[],"left":[]},{"functionId":"china","end":"low","entered":[],"left":[]},{"functionId":"china","end":"high","entered":["h3c"],"left":["asus"]}],"validation":[],"baseline":{"spearman":0.945454545455,"baselineTop10NotSelected":["inventec"],"selectedNotInBaselineTop10":["huawei"]},"raterAgreement":{"within":0.740740740741,"weightedKappa":0.793893129771,"disagreements":["hon-hai:gpu-systems","huawei:china"]}},"inputs":{"hon-hai":{"schema":"5lc-index-inputs/1","key":"infrastructure-servers-systems","entityId":"hon-hai","name":"Hon Hai (Foxconn)","kind":"public","commercialActivityVerified":true,"shares":{"gpu-systems":{"low":9.456,"central":21.632,"high":37.543,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"LTM to 2026-06-30 (3Q25-2Q26)","periodEnd":"2026-06-30","source":{"url":"https://focustaiwan.tw/business/202608120018","title":"Hon Hai reports highest-ever Q2 net profit on AI boom","publisher":"Focus Taiwan (CNA)","publishedDate":"2026-08-12","accessedDate":"2026-10-22","tier":"T5","locator":"Q2 2026 consolidated sales NT$2.53tn (+41% y/y, +19% q/q); cloud and networking 51% of Q2 revenue, up from 48% in Q1 2026"},"note":"Cloud and networking products (servers and networking, the segment that holds AI servers and racks) by quarter: Q3 2025 NT$2.06tn x 42% (Foxconn Q3 release as printed by I-Connect007, https://iconnect007.com/article/147725/foxconn-posts-record-q3-results-projects-strong-q4-and-optimistic-2026-outlook/147722/flex), Q4 2025 NT$2.6tn x 42% (Taipei Times, https://www.taipeitimes.com/News/biz/archives/2026/03/17/2003853939), Q1 2026 NT$2.12tn x 48% (Focus Taiwan, https://focustaiwan.tw/business/202605140017), Q2 2026 NT$2.53tn x 51% (the source above). Converted quarter by quarter at the Federal Reserve's monthly-average NT$ per US$ (https://fred.stlouisfed.org/data/EXTAUS): 3Q24 32.28, 4Q24 32.34, 1Q25 32.89, 2Q25 30.83, 3Q25 29.95, 4Q25 31.03, 1Q26 31.62, 2Q26 31.60. Segment LTM US$137.1bn. AI servers taken as 80% of the segment (65-92%): the company says about 80% of its servers use AI GPUs and about 20% AI ASICs (Taipei Times, 17 Mar 2026) and that AI servers and racks drove Q2 2026 growth; the rest of the segment is switches, general servers and telecom gear. AI revenue $109.7bn ($89.1-126.1bn). Credit rule: only systems sold under the company's own name directly to data-centre operators count; contract assembly for other brands, and boards and modules sold to NVIDIA or other designers, earn no credit. Part of Hon Hai's rack and board output is contract assembly for other brands and for NVIDIA, which the company does not split out, so the credited part is taken as 65% (40-85%) = $71.3bn (the Morgan Stanley estimate of NVL72 rack assembly shares, Foxconn 46% in 2025 and 39% in 2026, https://finance.technews.tw/2026/09/10/morgan-stanley-is-bullish-on-nvidias-rack-server-shipments-expecting-them-to-double-this-year/, is a unit share of assemblers, not of sellers). Of the credited total, 15% (10-20%) is ASIC servers (custom function) and about 0.7% China, leaving GPU systems $60.1bn ($27.8-96.5bn). Tier T5 (derived from company segment shares and statements). The earlier screening figure was 16.5%. Pool: GPU servers and racks outside mainland China, US$277.8bn over 3Q25-2Q26 (IDC figures as used in the approved card: GPU $301.3bn of $373.4bn accelerated servers, less China at 9% and 70% GPU); range US$257-294bn. Share = US$60.1bn / 277.8bn (range US$27.8-96.5bn against the pool range)."},"custom-accelerator-systems":{"low":4.752,"central":17.247,"high":46.611,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"LTM to 2026-06-30 (3Q25-2Q26)","periodEnd":"2026-06-30","source":{"url":"https://focustaiwan.tw/business/202608120018","title":"Hon Hai reports highest-ever Q2 net profit on AI boom","publisher":"Focus Taiwan (CNA)","publishedDate":"2026-08-12","accessedDate":"2026-10-22","tier":"T5","locator":"Q2 2026 consolidated sales NT$2.53tn (+41% y/y, +19% q/q); cloud and networking 51% of Q2 revenue, up from 48% in Q1 2026"},"note":"ASIC-based AI servers (cloud companies' own chips) = 15% (10-20%) of the credited AI total of $71.3bn, $10.7bn; the company says about 20% of its servers use AI ASICs (Taipei Times, 17 Mar 2026), by count, so a lower value share is used because ASIC servers cost less. See the GPU note for the derivation. Tier T5. Pool: servers for non-GPU (other) accelerators outside mainland China, US$62.0bn over 3Q25-2Q26 (IDC 'other accelerated' servers $72.1bn less China's share, as in the approved card); range US$46-75bn. Share = US$10.7bn / 62bn (range US$3.6-21.4bn against the pool range)."},"china":{"low":0,"central":1.04,"high":6.51,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"LTM to Jun 2026 (IDC 3Q25-2Q26)","periodEnd":"2026-06-30","source":{"url":"https://iconnect007.com/article/151169/foxconn-announces-q2-2026-financial-results/151166/ein","title":"Foxconn Announces Q2 2026 Financial Results (reprint of the Foxconn release)","publisher":"I-Connect007 (Foxconn / Hon Hai Technology Group release)","publishedDate":"2026-08-12","accessedDate":"2026-10-22","tier":"T5","locator":"Q2 2026 revenue NT$2.53tn (+41%), H1 NT$4.65tn; AI server racks: full-year 2026 shipments expected to more than double"},"note":"Foxconn Industrial Internet sells servers in China but export controls limit AI-server work there; placeholder US$0.35bn (0-1.7), tier T5. The register treats Fii as part of Hon Hai. Pool: AI servers sold in mainland China, about US$33.6bn over 3Q25-2Q26 (9% of IDC's $373.4bn accelerated servers, as in the approved card); range US$26-45bn. Share = US$0.35bn / 33.6bn (range US$0-1.7bn against the pool range)."}},"priorShares":{"gpu-systems":{"low":5.341,"central":12.428,"high":24.481,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"LTM to 2025-06-30 (3Q24-2Q25)","periodEnd":"2025-06-30","source":{"url":"https://www.taipeitimes.com/News/biz/archives/2026/03/17/2003853939","title":"AI server demand driving revenue, Hon Hai says","publisher":"Taipei Times","publishedDate":"2026-03-17","accessedDate":"2026-10-22","tier":"T5","locator":"Q4 2025 cloud and networking 42% of revenue, up from 26% a year earlier; full year 2025 40%"},"note":"Same method for 3Q24-2Q25. Revenue (NT$bn) and cloud and networking shares: Q3 2024 1,856 (Q3 2025 NT$2.06tn was +11% y/y, I-Connect007) x 31% (estimate, 28-34%, between the neighbouring quarters); Q4 2024 2,130 x 26% (Taipei Times, 17 Mar 2026); Q1 2025 1,643 (Q1 2026 NT$2.12tn was +29% y/y, Focus Taiwan) x 34% (implied by full-year 2025 at 40%, 31-36%); Q2 2025 1,794 (Q2 2026 +41% y/y) x 41% (Q3 release: 'from 41% in the previous quarter'). Converted quarter by quarter at the Federal Reserve's monthly-average NT$ per US$ (https://fred.stlouisfed.org/data/EXTAUS): 3Q24 32.28, 4Q24 32.34, 1Q25 32.89, 2Q25 30.83, 3Q25 29.95, 4Q25 31.03, 1Q26 31.62, 2Q26 31.60. Segment LTM US$75.8bn. AI servers taken as 55% (45-70%) of it: the company said AI servers were over 40% of server revenue in the first nine months of 2024 and would exceed half in 2025 (Omdia/Reuters via Yicai, https://yicaiglobal.com/news/foxconn-to-rank-as-worlds-largest-server-vendor-report-says). Credit 65% (40-85%), ASIC share 10% (5-15%). GPU systems $24.2bn ($11.3-42.8bn). Tier T5. Prior pool, 3Q24-2Q25 (5LayerCake derivation, T5). Worldwide GPU-server revenue for the four quarters = $217.5bn: 3Q24 $47.1bn and 4Q24 $47.0bn are the approved card's 3Q25 ($70.3bn) and 4Q25 ($74.7bn) GPU-server figures divided by IDC's embedded-GPU server growth of 49.4% (IDC 3Q25 release, 11 Dec 2025) and 59.1% (IDC 4Q25 release, 12 Mar 2026); 1Q25 $55.2bn and 2Q25 $68.3bn are the card's 1Q25 and 2Q25 totals less other-accelerated servers. China is IDC China's own accelerated-server value for the same window, $33.1bn (2H24 = 2024 $22.1bn less 1H24 $5.0bn = $17.1bn, plus 1H25 $16.0bn), of which GPU 69% = $22.8bn. GPU servers outside China = $217.5bn - $22.8bn = $194.7bn (range 175-212). Other-accelerated outside China = about $30bn worldwide (1Q25 $8.0bn and 2Q25 $8.2bn from the card's growth rates; 3Q24 and 4Q24 assumed about $7bn and $7.5bn) less China's non-GPU $10.3bn = $19.7bn (range 15-26). Share = US$24.2bn / 194.7bn."},"custom-accelerator-systems":{"low":2.624,"central":13.754,"high":45.096,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"LTM to 2025-06-30 (3Q24-2Q25)","periodEnd":"2025-06-30","source":{"url":"https://www.taipeitimes.com/News/biz/archives/2026/03/17/2003853939","title":"AI server demand driving revenue, Hon Hai says","publisher":"Taipei Times","publishedDate":"2026-03-17","accessedDate":"2026-10-22","tier":"T5","locator":"Q4 2025 cloud and networking 42% of revenue, up from 26% a year earlier; full year 2025 40%"},"note":"ASIC servers taken as 10% (5-15%) of the credited AI total of $27.1bn = $2.7bn; see the GPU prior note. Tier T5. Prior pool for this function: other-accelerated servers outside China about US$19.7bn (range 15-26), derived as described in the GPU-systems prior note of the same file."},"china":{"low":0,"central":1.04,"high":6.51,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"LTM to 2025-06-30","periodEnd":"2025-06-30","source":{"url":"https://www.taipeitimes.com/News/biz/archives/2026/03/17/2003853939","title":"AI server demand driving revenue, Hon Hai says","publisher":"Taipei Times","publishedDate":"2026-03-17","accessedDate":"2026-10-22","tier":"T5","locator":"Q4 2025 cloud and networking 42% of revenue, up from 26% a year earlier; full year 2025 40%"},"note":"Placeholder carried unchanged from the current period: no AI-server sales in China appear in the sources opened for either period, so it adds no share change."}},"activityGrowth":{"current":{"low":35644000000,"central":71287000000,"high":107205000000,"unit":"USD","basis":"AI server sales credited to the company as seller of the finished system, latest four quarters, US dollars","period":"LTM to 2026-06-30","periodEnd":"2026-06-30","source":{"url":"https://focustaiwan.tw/business/202608120018","title":"Hon Hai reports highest-ever Q2 net profit on AI boom","publisher":"Focus Taiwan (CNA)","publishedDate":"2026-08-12","accessedDate":"2026-10-22","tier":"T5","locator":"Q2 2026 consolidated sales NT$2.53tn (+41% y/y, +19% q/q); cloud and networking 51% of Q2 revenue, up from 48% in Q1 2026"},"note":"Derived as in the share notes (cloud and networking revenue x AI share x credited share)."},"prior":{"low":13643000000,"central":27096000000,"high":45096000000,"unit":"USD","basis":"AI server sales credited to the company as seller of the finished system, latest four quarters, US dollars","period":"LTM to 2025-06-30","periodEnd":"2025-06-30","source":{"url":"https://www.taipeitimes.com/News/biz/archives/2026/03/17/2003853939","title":"AI server demand driving revenue, Hon Hai says","publisher":"Taipei Times","publishedDate":"2026-03-17","accessedDate":"2026-10-22","tier":"T5","locator":"Q4 2025 cloud and networking 42% of revenue, up from 26% a year earlier; full year 2025 40%"},"note":"Derived as in the prior-share note."}},"baselineRevenueUsd":{"low":89109000000,"central":109672000000,"high":126123000000,"unit":"USD","basis":"AI server revenue (all AI servers and racks, before the credit rule), latest four quarters, US dollars","period":"LTM to 2026-06-30","periodEnd":"2026-06-30","source":{"url":"https://focustaiwan.tw/business/202608120018","title":"Hon Hai reports highest-ever Q2 net profit on AI boom","publisher":"Focus Taiwan (CNA)","publishedDate":"2026-08-12","accessedDate":"2026-10-22","tier":"T5","locator":"Q2 2026 consolidated sales NT$2.53tn (+41% y/y, +19% q/q); cloud and networking 51% of Q2 revenue, up from 48% in Q1 2026"},"note":"Estimate, not a reported figure: cloud and networking revenue US$137.1bn x 80% AI share (65-92%)."},"screening":{"gpu-systems":{"levelLow":1,"levelHigh":3},"custom-accelerator-systems":{"levelLow":1.5,"levelHigh":3.5},"china":{"levelLow":1,"levelHigh":2}},"checks":{"sourceSupport":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-22","passed":true,"sha256":"2b43d57fa309461b41c4572d73c0a5e043305784f9698ad450697863d5378797"},"consistency":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-22","passed":true,"sha256":"2b43d57fa309461b41c4572d73c0a5e043305784f9698ad450697863d5378797"}}},"quanta-computer":{"schema":"5lc-index-inputs/1","key":"infrastructure-servers-systems","entityId":"quanta-computer","name":"Quanta Computer (incl. QCT)","kind":"public","commercialActivityVerified":true,"shares":{"gpu-systems":{"low":7.259,"central":14.62,"high":21.634,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"LTM to 2026-06-30 (3Q25-2Q26)","periodEnd":"2026-06-30","source":{"url":"https://www.taipeitimes.com/News/biz/archives/2026/07/08/2003860386","title":"Quanta revenue: second quarter NT$1.04 trillion, first half NT$1.85 trillion (Taipei Times, 8 Jul 2026)","publisher":"Taipei Times","publishedDate":"2026-07-08","accessedDate":"2026-10-22","tier":"T5","locator":"Q2 2026 revenue NT$1.04tn; first half NT$1.85tn, +86.5%; AI-related revenue expected to double in 2026"},"note":"Quanta revenue (NT$bn): Q3 2025 495.26 (Taiwan News, 13 Nov 2025, https://www.taiwannews.com.tw/en/news/6243309); Q4 2025 about 634.8 (full-year 2025 NT$2.12tn less the first nine months NT$1.485tn: Q1 485.67 + Q2 504.12 + Q3 495.26, with Q1 from Taipei Times 26 Apr 2025 and the first half NT$989.79bn from Taipei Times 9 Jul 2025, https://www.taipeitimes.com/News/biz/archives/2025/07/09/2003839962); Q1 2026 809.2 and Q2 2026 1,040.8 (first half NT$1.85tn, Taipei Times 8 Jul 2026, less Q1; Q1 +27% q/q per KGI, May 2026). Converted quarter by quarter at the Federal Reserve's monthly-average NT$ per US$ (https://fred.stlouisfed.org/data/EXTAUS): 3Q24 32.28, 4Q24 32.34, 1Q25 32.89, 2Q25 30.83, 3Q25 29.95, 4Q25 31.03, 1Q26 31.62, 2Q26 31.60. Revenue LTM US$95.5bn. Server share of revenue: about 70% in Q3 2025 (Taipei Times 14 Nov 2025, https://www.taipeitimes.com/News/biz/archives/2025/11/14/2003847162), over 80% in Q1 2026 and about 85% expected in Q2 (KGI Securities, 14 May 2026, quoting management, https://www.kgi.com.hk/en/-/media/files/kgishk/research-reports/tw-reports/quanta-computer_2382-tt_14052026e.pdf); Q4 2025 taken as 75%. AI share of server revenue: 70% in Q3 2025, 'over 75%' in Q1 2026 (KGI), expected above 80% for 2026; Q4 2025 taken as 72%, Q2 2026 as 78%. AI revenue $56.5bn (+/-15%). Credit rule: only systems sold under the company's own name directly to data-centre operators count; contract assembly for other brands, and boards and modules sold to NVIDIA or other designers, earn no credit. QCT sells under its own name to cloud operators (none are named in the sources opened), so 85% (60-95%) is credited; management is moving GPU servers from a buy-and-sell to a consignment model with several clients (KGI), which lowers Quanta's revenue but not the system value. = $48.1bn. ASIC servers 15% (10-25%) of that (management expects ASIC server shipments to double in 2026), China about 0.5%, so GPU systems $40.6bn ($21.3-55.6bn). Tier T5. Cross-check: Morgan Stanley puts Quanta at 23% of 2025 and 26% of 2026 NVL72 rack assembly (technews.tw, 10 Sep 2026). The screening figure was 14.4%. Pool: GPU servers and racks outside mainland China, US$277.8bn over 3Q25-2Q26 (IDC figures as used in the approved card: GPU $301.3bn of $373.4bn accelerated servers, less China at 9% and 70% GPU); range US$257-294bn. Share = US$40.6bn / 277.8bn (range US$21.3-55.6bn against the pool range)."},"custom-accelerator-systems":{"low":3.845,"central":11.628,"high":33.574,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"LTM to 2026-06-30 (3Q25-2Q26)","periodEnd":"2026-06-30","source":{"url":"https://www.taipeitimes.com/News/biz/archives/2026/07/08/2003860386","title":"Quanta revenue: second quarter NT$1.04 trillion, first half NT$1.85 trillion (Taipei Times, 8 Jul 2026)","publisher":"Taipei Times","publishedDate":"2026-07-08","accessedDate":"2026-10-22","tier":"T5","locator":"Q2 2026 revenue NT$1.04tn; first half NT$1.85tn, +86.5%; AI-related revenue expected to double in 2026"},"note":"ASIC-based AI servers taken as 15% (10-25%) of the credited AI total of $48.1bn = $7.2bn; management says ASIC-based AI server shipments will double in 2026 (Taipei Times, 14 Nov 2025). See the GPU note. Tier T5. Pool: servers for non-GPU (other) accelerators outside mainland China, US$62.0bn over 3Q25-2Q26 (IDC 'other accelerated' servers $72.1bn less China's share, as in the approved card); range US$46-75bn. Share = US$7.2bn / 62bn."},"china":{"low":0,"central":0.223,"high":1.15,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"LTM to Jun 2026 (IDC 3Q25-2Q26)","periodEnd":"2026-06-30","source":{"url":"https://www.taipeitimes.com/News/biz/archives/2026/07/08/2003860386","title":"Taipei Times, 8 July 2026, p. 9 (Meryl Kao): Advantech and Quanta Computer second-quarter revenue","publisher":"Taipei Times","publishedDate":"2026-07-08","accessedDate":"2026-10-22","tier":"T5","locator":"Quanta Q2 2026 revenue NT$1.04tn, H1 NT$1.85tn (+86.5%); 'robust AI server shipments'"},"note":"Wide placeholder estimate, tier T5: no sale of AI servers inside mainland China is shown in the sources opened. Upper end set at US$0.3bn; central at a quarter of that. This function is not material for the company. Pool: AI servers sold in mainland China, about US$33.6bn over 3Q25-2Q26 (9% of IDC's $373.4bn accelerated servers, as in the approved card); range US$26-45bn. Share = US$0.075bn / 33.6bn (range US$0-0.3bn against the pool range)."}},"priorShares":{"gpu-systems":{"low":4.406,"central":9.052,"high":14.337,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"LTM to 2025-06-30 (3Q24-2Q25)","periodEnd":"2025-06-30","source":{"url":"https://www.taipeitimes.com/News/biz/archives/2025/07/09/2003839962","title":"Quanta revenue jumps 70.6 percent on server demand","publisher":"Taipei Times","publishedDate":"2025-07-09","accessedDate":"2026-10-22","tier":"T5","locator":"First half 2025 NT$989.79bn (+73.99%); servers about 70% of Q1 2025 sales; AI servers expected 65-70% of Q2 server sales"},"note":"Revenue (NT$bn): Q3 2024 424.4 (Q3 2025 was +16.7% y/y, Taiwan News); Q4 2024 about 416.5 (2024 total NT$1.41tn, Taipei Times 26 Apr 2025, less Q1 259.0 + Q2 309.9 + Q3 424.4; Q1 2024 is Q1 2025's NT$485.67bn / 1.8756 and the first half of 2024 is the first half of 2025's NT$989.79bn / 1.7399); Q1 2025 485.67; Q2 2025 504.12 (first half NT$989.79bn). Converted quarter by quarter at the Federal Reserve's monthly-average NT$ per US$ (https://fred.stlouisfed.org/data/EXTAUS): 3Q24 32.28, 4Q24 32.34, 1Q25 32.89, 2Q25 30.83, 3Q25 29.95, 4Q25 31.03, 1Q26 31.62, 2Q26 31.60. LTM US$57.1bn. Server share of revenue 55% (Q3 2024, when servers overtook notebooks, which fell to about 30% of revenue; Taipei Times 26 Apr 2025, https://www.taipeitimes.com/News/biz/archives/2025/04/26/2003835826), 60% (Q4), 70% (Q1 2025, 'about 70% of overall sales', Taipei Times 9 Jul 2025), 70%; AI share of server revenue 60%, 60%, 62% (first-half 2025 'over 60%'), 68% (Q2 2025 guidance 65-70%). AI revenue $23.2bn (+/-20%), credit 85% (60-95%), ASIC 10%. GPU systems $17.6bn. Tier T5. Prior pool, 3Q24-2Q25 (5LayerCake derivation, T5). Worldwide GPU-server revenue for the four quarters = $217.5bn: 3Q24 $47.1bn and 4Q24 $47.0bn are the approved card's 3Q25 ($70.3bn) and 4Q25 ($74.7bn) GPU-server figures divided by IDC's embedded-GPU server growth of 49.4% (IDC 3Q25 release, 11 Dec 2025) and 59.1% (IDC 4Q25 release, 12 Mar 2026); 1Q25 $55.2bn and 2Q25 $68.3bn are the card's 1Q25 and 2Q25 totals less other-accelerated servers. China is IDC China's own accelerated-server value for the same window, $33.1bn (2H24 = 2024 $22.1bn less 1H24 $5.0bn = $17.1bn, plus 1H25 $16.0bn), of which GPU 69% = $22.8bn. GPU servers outside China = $217.5bn - $22.8bn = $194.7bn (range 175-212). Other-accelerated outside China = about $30bn worldwide (1Q25 $8.0bn and 2Q25 $8.2bn from the card's growth rates; 3Q24 and 4Q24 assumed about $7bn and $7.5bn) less China's non-GPU $10.3bn = $19.7bn (range 15-26). Share = US$17.6bn / 194.7bn."},"custom-accelerator-systems":{"low":2.139,"central":9.996,"high":26.411,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"LTM to 2025-06-30 (3Q24-2Q25)","periodEnd":"2025-06-30","source":{"url":"https://www.taipeitimes.com/News/biz/archives/2025/07/09/2003839962","title":"Quanta revenue jumps 70.6 percent on server demand","publisher":"Taipei Times","publishedDate":"2025-07-09","accessedDate":"2026-10-22","tier":"T5","locator":"First half 2025 NT$989.79bn (+73.99%); servers about 70% of Q1 2025 sales; AI servers expected 65-70% of Q2 server sales"},"note":"ASIC servers taken as 10% (5-15%) of the credited AI total of $19.7bn = $2.0bn. Tier T5. Prior pool for this function: other-accelerated servers outside China about US$19.7bn (range 15-26), derived as described in the GPU-systems prior note of the same file."},"china":{"low":0,"central":0.223,"high":1.15,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"LTM to 2025-06-30","periodEnd":"2025-06-30","source":{"url":"https://www.taipeitimes.com/News/biz/archives/2025/07/09/2003839962","title":"Quanta revenue jumps 70.6 percent on server demand","publisher":"Taipei Times","publishedDate":"2025-07-09","accessedDate":"2026-10-22","tier":"T5","locator":"First half 2025 NT$989.79bn (+73.99%); servers about 70% of Q1 2025 sales; AI servers expected 65-70% of Q2 server sales"},"note":"Placeholder carried unchanged from the current period, so it adds no share change."}},"activityGrowth":{"current":{"low":28838000000,"central":48064000000,"high":61776000000,"unit":"USD","basis":"AI server sales credited to the company as seller of the finished system, latest four quarters, US dollars","period":"LTM to 2026-06-30","periodEnd":"2026-06-30","source":{"url":"https://www.taipeitimes.com/News/biz/archives/2026/07/08/2003860386","title":"Quanta revenue: second quarter NT$1.04 trillion, first half NT$1.85 trillion (Taipei Times, 8 Jul 2026)","publisher":"Taipei Times","publishedDate":"2026-07-08","accessedDate":"2026-10-22","tier":"T5","locator":"Q2 2026 revenue NT$1.04tn; first half NT$1.85tn, +86.5%; AI-related revenue expected to double in 2026"},"note":"Derived as in the share notes."},"prior":{"low":11120000000,"central":19692000000,"high":26411000000,"unit":"USD","basis":"AI server sales credited to the company as seller of the finished system, latest four quarters, US dollars","period":"LTM to 2025-06-30","periodEnd":"2025-06-30","source":{"url":"https://www.taipeitimes.com/News/biz/archives/2025/07/09/2003839962","title":"Quanta revenue jumps 70.6 percent on server demand","publisher":"Taipei Times","publishedDate":"2025-07-09","accessedDate":"2026-10-22","tier":"T5","locator":"First half 2025 NT$989.79bn (+73.99%); servers about 70% of Q1 2025 sales; AI servers expected 65-70% of Q2 server sales"},"note":"Derived as in the prior-share note."}},"baselineRevenueUsd":{"low":48064000000,"central":56545000000,"high":65027000000,"unit":"USD","basis":"AI server revenue (before the credit rule), latest four quarters, US dollars","period":"LTM to 2026-06-30","periodEnd":"2026-06-30","source":{"url":"https://www.taipeitimes.com/News/biz/archives/2026/07/08/2003860386","title":"Quanta revenue: second quarter NT$1.04 trillion, first half NT$1.85 trillion (Taipei Times, 8 Jul 2026)","publisher":"Taipei Times","publishedDate":"2026-07-08","accessedDate":"2026-10-22","tier":"T5","locator":"Q2 2026 revenue NT$1.04tn; first half NT$1.85tn, +86.5%; AI-related revenue expected to double in 2026"},"note":"Estimate, not a reported figure: revenue US$95.5bn x server share x AI share by quarter."},"screening":{"gpu-systems":{"levelLow":1,"levelHigh":3},"custom-accelerator-systems":{"levelLow":1.5,"levelHigh":3.5},"china":{"levelLow":1,"levelHigh":2}},"checks":{"sourceSupport":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-22","passed":true,"sha256":"ff5b0d4b39bb3596ff7e86f7fc698feccf8f84e0c698ed1858f480e288ae5a8e"},"consistency":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-22","passed":true,"sha256":"ff5b0d4b39bb3596ff7e86f7fc698feccf8f84e0c698ed1858f480e288ae5a8e"}}},"dell":{"schema":"5lc-index-inputs/1","key":"infrastructure-servers-systems","entityId":"dell","name":"Dell Technologies","kind":"public","commercialActivityVerified":true,"shares":{"gpu-systems":{"low":15.2,"central":17,"high":19.2,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"LTM to 2026-07-31 (FY2026 + H1 FY2027 - H1 FY2026), against IDC 3Q25-2Q26","periodEnd":"2026-07-31","source":{"url":"https://www.dell.com/en-us/dt/corporate/newsroom/announcements/detailpage.press-releases~usa~2026~09~dell-technologies-delivers-second-quarter-fiscal-2027-financial-results.htm","title":"Dell Technologies Delivers Second Quarter Fiscal 2027 Financial Results","publisher":"Dell Technologies","publishedDate":"2026-09-01","accessedDate":"2026-10-22","tier":"T4","locator":"AI-optimized servers: Q2 FY2027 $16,401M (vs $8,208M); first half FY2027 $32,533M (vs $10,090M)"},"note":"Dell's own AI-optimized server revenue: FY2026 $24,683M (Form 10-K, MD&A ISG table, read from the local copy .cache/sec/dell; URL https://www.sec.gov/Archives/edgar/data/1571996/000157199626000008/dell-20260130.htm) plus H1 FY2027 $32,533M minus H1 FY2026 $10,090M (release) = $47,126M. Tier T4: the numerator is a company figure (T1/T2), the pool is an IDC estimate (T4), so the lower tier applies; range is the pool range and +/-5% on the numerator. Dell's fiscal LTM ends one month after IDC's window. Pool: GPU servers and racks outside mainland China, US$277.8bn over 3Q25-2Q26 (IDC figures as used in the approved card: GPU $301.3bn of $373.4bn accelerated servers, less China at 9% and 70% GPU); range US$257-294bn. Share = US$47.1bn / 277.8bn (range US$44.8-49.5bn against the pool range)."},"custom-accelerator-systems":{"low":0,"central":0.121,"high":0.656,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"LTM to 2026-07-31","periodEnd":"2026-07-31","source":{"url":"https://www.dell.com/en-us/dt/corporate/newsroom/announcements/detailpage.press-releases~usa~2026~09~dell-technologies-delivers-second-quarter-fiscal-2027-financial-results.htm","title":"Dell Technologies Delivers Second Quarter Fiscal 2027 Financial Results","publisher":"Dell Technologies","publishedDate":"2026-09-01","accessedDate":"2026-10-22","tier":"T5","locator":"AI-optimized servers: Q2 FY2027 $16,401M (vs $8,208M); first half FY2027 $32,533M (vs $10,090M)"},"note":"Wide placeholder estimate, tier T5: no system sales built on cloud companies' own accelerators are shown in the sources opened. Upper end set at US$0.3bn; central at a quarter of that. This function is not material for the company. Pool: servers for non-GPU (other) accelerators outside mainland China, US$62.0bn over 3Q25-2Q26 (IDC 'other accelerated' servers $72.1bn less China's share, as in the approved card); range US$46-75bn. Share = US$0.075bn / 62bn (range US$0-0.3bn against the pool range)."},"china":{"low":0,"central":0.372,"high":1.92,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"LTM to 2026-07-31","periodEnd":"2026-07-31","source":{"url":"https://www.dell.com/en-us/dt/corporate/newsroom/announcements/detailpage.press-releases~usa~2026~09~dell-technologies-delivers-second-quarter-fiscal-2027-financial-results.htm","title":"Dell Technologies Delivers Second Quarter Fiscal 2027 Financial Results","publisher":"Dell Technologies","publishedDate":"2026-09-01","accessedDate":"2026-10-22","tier":"T5","locator":"AI-optimized servers: Q2 FY2027 $16,401M (vs $8,208M); first half FY2027 $32,533M (vs $10,090M)"},"note":"Wide placeholder estimate, tier T5: no sale of AI servers inside mainland China is shown in the sources opened. Upper end set at US$0.5bn; central at a quarter of that. This function is not material for the company. Pool: AI servers sold in mainland China, about US$33.6bn over 3Q25-2Q26 (9% of IDC's $373.4bn accelerated servers, as in the approved card); range US$26-45bn. Share = US$0.125bn / 33.6bn (range US$0-0.5bn against the pool range)."}},"priorShares":{"gpu-systems":{"low":6.792,"central":7.64,"high":8.8,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"LTM to 2025-07-31 (FY2025 + H1 FY2026 - H1 FY2025), against 3Q24-2Q25","periodEnd":"2025-07-31","source":{"url":"https://www.dell.com/en-us/dt/corporate/newsroom/announcements/detailpage.press-releases~usa~2026~09~dell-technologies-delivers-second-quarter-fiscal-2027-financial-results.htm","title":"Dell Technologies Delivers Second Quarter Fiscal 2027 Financial Results","publisher":"Dell Technologies","publishedDate":"2026-09-01","accessedDate":"2026-10-22","tier":"T5","locator":"Prior-year columns: Q2 FY2026 AI-optimized servers $8,208M, first half FY2026 $10,090M; FY2025 $9,286M is from the FY2026 10-K"},"note":"Dell's own AI-optimized server revenue for the four fiscal quarters to 31 Jul 2025 = FY2025 $9,286M (Form 10-K FY2026, MD&A ISG table, prior-year column; https://www.sec.gov/Archives/edgar/data/1571996/000157199626000008/dell-20260130.htm) + H1 FY2026 $10,090M (Q2 FY2027 release, prior-year column) - H1 FY2025 estimated at $4.5bn (range 4.0-5.0), about 48% of FY2025: Dell first reported AI-optimized servers as a product category in Q4 FY2026, and no quarterly FY2025 split appears in the sources opened. = $14.9bn (14.4-15.4). Tier T5 (derived; the H1 FY2025 split is an estimate). Prior pool, 3Q24-2Q25 (5LayerCake derivation, T5). Worldwide GPU-server revenue for the four quarters = $217.5bn: 3Q24 $47.1bn and 4Q24 $47.0bn are the approved card's 3Q25 ($70.3bn) and 4Q25 ($74.7bn) GPU-server figures divided by IDC's embedded-GPU server growth of 49.4% (IDC 3Q25 release, 11 Dec 2025) and 59.1% (IDC 4Q25 release, 12 Mar 2026); 1Q25 $55.2bn and 2Q25 $68.3bn are the card's 1Q25 and 2Q25 totals less other-accelerated servers. China is IDC China's own accelerated-server value for the same window, $33.1bn (2H24 = 2024 $22.1bn less 1H24 $5.0bn = $17.1bn, plus 1H25 $16.0bn), of which GPU 69% = $22.8bn. GPU servers outside China = $217.5bn - $22.8bn = $194.7bn (range 175-212). Other-accelerated outside China = about $30bn worldwide (1Q25 $8.0bn and 2Q25 $8.2bn from the card's growth rates; 3Q24 and 4Q24 assumed about $7bn and $7.5bn) less China's non-GPU $10.3bn = $19.7bn (range 15-26). Share = US$14.9bn / 194.7bn (range US$14.4-15.4bn against the prior pool range)."},"custom-accelerator-systems":{"low":0,"central":0.121,"high":0.656,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"LTM to 2025-07-31","periodEnd":"2025-07-31","source":{"url":"https://www.dell.com/en-us/dt/corporate/newsroom/announcements/detailpage.press-releases~usa~2026~09~dell-technologies-delivers-second-quarter-fiscal-2027-financial-results.htm","title":"Dell Technologies Delivers Second Quarter Fiscal 2027 Financial Results","publisher":"Dell Technologies","publishedDate":"2026-09-01","accessedDate":"2026-10-22","tier":"T5","locator":"Prior-year columns: Q2 FY2026 AI-optimized servers $8,208M, first half FY2026 $10,090M; FY2025 $9,286M is from the FY2026 10-K"},"note":"Placeholder carried unchanged from the current period: no AI-server sales in this function appear in the sources opened for either period, so it adds no share change."},"china":{"low":0,"central":0.372,"high":1.92,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"LTM to 2025-07-31","periodEnd":"2025-07-31","source":{"url":"https://www.dell.com/en-us/dt/corporate/newsroom/announcements/detailpage.press-releases~usa~2026~09~dell-technologies-delivers-second-quarter-fiscal-2027-financial-results.htm","title":"Dell Technologies Delivers Second Quarter Fiscal 2027 Financial Results","publisher":"Dell Technologies","publishedDate":"2026-09-01","accessedDate":"2026-10-22","tier":"T5","locator":"Prior-year columns: Q2 FY2026 AI-optimized servers $8,208M, first half FY2026 $10,090M; FY2025 $9,286M is from the FY2026 10-K"},"note":"Placeholder carried unchanged from the current period: no AI-server sales in this function appear in the sources opened for either period, so it adds no share change."}},"activityGrowth":{"current":{"low":32533000000,"central":32533000000,"high":32533000000,"unit":"USD","basis":"AI-optimized server revenue, first six months of the fiscal year, US dollars","period":"H1 FY2027 (six months to 2026-07-31)","periodEnd":"2026-07-31","source":{"url":"https://www.dell.com/en-us/dt/corporate/newsroom/announcements/detailpage.press-releases~usa~2026~09~dell-technologies-delivers-second-quarter-fiscal-2027-financial-results.htm","title":"Dell Technologies Delivers Second Quarter Fiscal 2027 Financial Results","publisher":"Dell Technologies","publishedDate":"2026-09-01","accessedDate":"2026-10-22","tier":"T2","locator":"Q2 FY2027 AI-optimized server revenue $16.4bn; six-month $32.5bn (+222%); record $95bn AI server backlog; FY2027 AI-optimized server revenue outlook $74.0bn"},"note":"Company-reported: 'six-month AI-optimized server revenue was $32.5 billion, up 222%' ($32,533M)."},"prior":{"low":10090000000,"central":10090000000,"high":10090000000,"unit":"USD","basis":"AI-optimized server revenue, first six months of the fiscal year, US dollars","period":"H1 FY2026 (six months to 2025-07-31)","periodEnd":"2025-07-31","source":{"url":"https://www.dell.com/en-us/dt/corporate/newsroom/announcements/detailpage.press-releases~usa~2026~09~dell-technologies-delivers-second-quarter-fiscal-2027-financial-results.htm","title":"Dell Technologies Delivers Second Quarter Fiscal 2027 Financial Results","publisher":"Dell Technologies","publishedDate":"2026-09-01","accessedDate":"2026-10-22","tier":"T2","locator":"Q2 FY2027 AI-optimized server revenue $16.4bn; six-month $32.5bn (+222%); record $95bn AI server backlog; FY2027 AI-optimized server revenue outlook $74.0bn"},"note":"Prior-year comparative in the same release ($10,090M)."}},"committedExpansion":{"low":2.016,"central":2.016,"high":2.016,"unit":"ratio","basis":"Record AI server backlog at the end of the quarter ÷ AI-optimized server revenue, latest four fiscal quarters","period":"AI backlog at 2026-07-31","periodEnd":"2026-07-31","source":{"url":"https://www.dell.com/en-us/dt/corporate/newsroom/announcements/detailpage.press-releases~usa~2026~09~dell-technologies-delivers-second-quarter-fiscal-2027-financial-results.htm","title":"Dell Technologies Delivers Second Quarter Fiscal 2027 Financial Results","publisher":"Dell Technologies","publishedDate":"2026-09-01","accessedDate":"2026-10-22","tier":"T2","locator":"Q2 FY2027 AI-optimized server revenue $16.4bn; six-month $32.5bn (+222%); record $95bn AI server backlog; FY2027 AI-optimized server revenue outlook $74.0bn"},"note":"The company says it exited Q2 FY2027 'with a record $95 billion backlog' for AI servers (CEO Jeff Clarke) and booked $60.9bn of AI orders in the quarter; it raised its FY2027 AI-optimized server revenue outlook to $74.0bn. Current annual activity: $47.126bn = FY2026 $24,683M (10-K) + H1 FY2027 $32,533M - H1 FY2026 $10,090M (release). 95.0 ÷ 47.126 = 2.02. Tier T2 (the company's own statement)."},"baselineRevenueUsd":{"low":47126000000,"central":47126000000,"high":47126000000,"unit":"USD","basis":"AI-optimized server revenue, latest four fiscal quarters to 2026-07-31, US dollars","period":"LTM to 2026-07-31","periodEnd":"2026-07-31","source":{"url":"https://www.dell.com/en-us/dt/corporate/newsroom/announcements/detailpage.press-releases~usa~2026~09~dell-technologies-delivers-second-quarter-fiscal-2027-financial-results.htm","title":"Dell Technologies Delivers Second Quarter Fiscal 2027 Financial Results","publisher":"Dell Technologies","publishedDate":"2026-09-01","accessedDate":"2026-10-22","tier":"T2","locator":"Q2 FY2027 AI-optimized server revenue $16.4bn; six-month $32.5bn (+222%); record $95bn AI server backlog; FY2027 AI-optimized server revenue outlook $74.0bn"},"note":"AI-optimized server revenue only (not Dell's total revenue): FY2026 $24,683M (Form 10-K) + H1 FY2027 $32,533M - H1 FY2026 $10,090M = $47,126M."},"screening":{"gpu-systems":{"levelLow":1,"levelHigh":3},"custom-accelerator-systems":{"levelLow":1,"levelHigh":2},"china":{"levelLow":1,"levelHigh":2}},"checks":{"sourceSupport":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-22","passed":true,"sha256":"4c743b37eb679dbd154501ec3f6d515cac28f0937eb29f06dec6ed978b3f8dc5"},"consistency":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-22","passed":true,"sha256":"4c743b37eb679dbd154501ec3f6d515cac28f0937eb29f06dec6ed978b3f8dc5"}}},"wistron":{"schema":"5lc-index-inputs/1","key":"infrastructure-servers-systems","entityId":"wistron","name":"Wistron (group, incl. Wiwynn)","kind":"public","commercialActivityVerified":true,"shares":{"gpu-systems":{"low":4.355,"central":9.342,"high":18.615,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"LTM to 2026-06-30 (3Q25-2Q26)","periodEnd":"2026-06-30","source":{"url":"https://www.taipeitimes.com/News/biz/archives/2026/07/11/2003860550","title":"Wistron sales hit record (Taipei Times, 11 Jul 2026)","publisher":"Taipei Times","publishedDate":"2026-07-11","accessedDate":"2026-10-22","tier":"T5","locator":"Q2 revenue NT$895.44bn; first half NT$1.74tn, +94.1%; AI servers about 70% of revenue in the first two months of 2026 (company, March)"},"note":"Wistron group revenue (NT$bn): Q4 2025 720.94 (Taipei Times 14 Mar 2026, https://www.taipeitimes.com/News/biz/archives/2026/03/14/2003853775), 2025 NT$2.18tn, +108.4% (Taipei Times 9 Jan 2026, https://www.taipeitimes.com/News/biz/archives/2026/01/09/2003850283), Q2 2026 895.44 and first half NT$1.74tn (Taipei Times 11 Jul 2026; Reuters via The Star, 4 Aug 2026), so Q1 2026 about 844.6; Q3 2025 about 562.6 (2025 NT$2.18tn, +108.4%, less Q4 and a first half of about 896.5). Converted quarter by quarter at the Federal Reserve's monthly-average NT$ per US$ (https://fred.stlouisfed.org/data/EXTAUS): 3Q24 32.28, 4Q24 32.34, 1Q25 32.89, 2Q25 30.83, 3Q25 29.95, 4Q25 31.03, 1Q26 31.62, 2Q26 31.60. Group revenue LTM US$97.1bn, of which Wiwynn (consolidated) US$35.9bn (NT$ 266.8bn, 292.4bn, 276.5bn and 278.2bn by quarter, Wiwynn results releases: Q3 2025 https://www.wiwynn.com/news/wiwynn-reports-third-quarter-2025-financial-results, Q4 2025, Q1 2026 deck https://www.wiwynn.com/hubfs/Investors/Shareholders_Services/Investor_Conference/26Q1_ENG_20260507.pdf, Q2 2026 https://www.wiwynn.com/news/wiwynn-reports-second-quarter-2026-financial-results). AI servers: the company said about 70% of revenue in January-February 2026 (Taipei Times 14 Mar 2026); the LTM average, which includes the lower 2025 quarters, is taken as 62% (55-72%) = US$60.2bn. Wiwynn: 'AI-related products contributing to more than half of the revenue' in 2025 (Wiwynn release, https://www.wiwynn.com/news/wiwynn-reports-fourth-quarter-2025-unaudited-financial-results), taken as 60% (50-70%) = US$21.5bn; the rest, US$38.7bn, is Wistron's own business (GPU baseboards and servers built largely for NVIDIA, Dell and others). Credit rule: only systems sold under the company's own name directly to data-centre operators count; contract assembly for other brands, and boards and modules sold to NVIDIA or other designers, earn no credit. Wiwynn sells to hyperscale cloud operators under its own name (customers are not named in the sources opened): 90% (80-100%) credited; Wistron's own business 40% (20-60%) credited because much of it is boards and assembly for other brands. Credited US$34.8bn. ASIC servers 25% (15-35%; assumed, no split is published), China ~0.5%, so GPU systems US$26.0bn (US$12.8-47.8bn). Tier T5. Cross-check: Morgan Stanley puts Wistron (with Wiwynn) at 22% of 2025 and 2026 NVL72 rack assembly (technews.tw, 10 Sep 2026). The screening figure was 8.6%. Pool: GPU servers and racks outside mainland China, US$277.8bn over 3Q25-2Q26 (IDC figures as used in the approved card: GPU $301.3bn of $373.4bn accelerated servers, less China at 9% and 70% GPU); range US$257-294bn. Share = US$26.0bn / 277.8bn (range US$12.8-47.8bn against the pool range)."},"custom-accelerator-systems":{"low":4.001,"central":14.047,"high":42.824,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"LTM to 2026-06-30 (3Q25-2Q26)","periodEnd":"2026-06-30","source":{"url":"https://www.taipeitimes.com/News/biz/archives/2026/07/11/2003860550","title":"Wistron sales hit record (Taipei Times, 11 Jul 2026)","publisher":"Taipei Times","publishedDate":"2026-07-11","accessedDate":"2026-10-22","tier":"T5","locator":"Q2 revenue NT$895.44bn; first half NT$1.74tn, +94.1%; AI servers about 70% of revenue in the first two months of 2026 (company, March)"},"note":"ASIC servers taken as 25% (15-35%) of the credited AI total of US$34.8bn = US$8.7bn. See the GPU note. Tier T5. Pool: servers for non-GPU (other) accelerators outside mainland China, US$62.0bn over 3Q25-2Q26 (IDC 'other accelerated' servers $72.1bn less China's share, as in the approved card); range US$46-75bn. Share = US$8.7bn / 62bn."},"china":{"low":0,"central":0.223,"high":1.15,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"LTM to Jun 2026 (IDC 3Q25-2Q26)","periodEnd":"2026-06-30","source":{"url":"https://www.thestar.com.my/tech/tech-news/2026/08/04/taiwan039s-wistron-says-ai-server-demand-remains-strong-quarterly-profit-up-128","title":"Taiwan's Wistron says AI server demand remains strong, quarterly profit up 128% (Reuters, Wen-Yee Lee)","publisher":"Reuters via The Star","publishedDate":"2026-08-04","accessedDate":"2026-10-22","tier":"T5","locator":"Q2 2026 revenue T$895.44bn (+64%); AI server demand exceeds supply; US$1 = T$32.414"},"note":"Wide placeholder estimate, tier T5: no sale of AI servers inside mainland China is shown in the sources opened. Upper end set at US$0.3bn; central at a quarter of that. This function is not material for the company. Pool: AI servers sold in mainland China, about US$33.6bn over 3Q25-2Q26 (9% of IDC's $373.4bn accelerated servers, as in the approved card); range US$26-45bn. Share = US$0.075bn / 33.6bn (range US$0-0.3bn against the pool range)."}},"priorShares":{"gpu-systems":{"low":2.446,"central":5.645,"high":11.556,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"LTM to 2025-06-30 (3Q24-2Q25)","periodEnd":"2025-06-30","source":{"url":"https://www.wiwynn.com/news/wiwynn-reports-fourth-quarter-2025-unaudited-financial-results","title":"Wiwynn Reports Fourth Quarter 2025 Unaudited Financial Results","publisher":"Wiwynn Corporation","publishedDate":"undated","accessedDate":"2026-10-22","tier":"T5","locator":"FY2025 consolidated revenue NT$950.663bn, +163.7%; Q4 2025 NT$292.440bn, +152.9%; AI-related products more than half of revenue"},"note":"Group revenue (NT$bn): Q4 2024 297.2 (Q4 2025 NT$720.94bn was +142.6% y/y); Q1 2025 345.1 and Q2 2025 551.4 (Q2 2026 NT$895.44bn was +62.4% y/y; first half 2025 about 896.5); Q3 2024 estimated at 270 (250-290; 2024 total about NT$1,046bn = 2025's NT$2.18tn / 2.084). Wiwynn: Q3 2024 97.8, Q4 2024 115.6, Q1 2025 170.7, Q2 2025 220.8 (from its releases' growth rates). Converted quarter by quarter at the Federal Reserve's monthly-average NT$ per US$ (https://fred.stlouisfed.org/data/EXTAUS): 3Q24 32.28, 4Q24 32.34, 1Q25 32.89, 2Q25 30.83, 3Q25 29.95, 4Q25 31.03, 1Q26 31.62, 2Q26 31.60. Group LTM US$45.9bn, Wiwynn US$19.0bn. AI share taken as 45% (35-55%) of group revenue and 50% (40-60%) of Wiwynn's. Credit 90% for Wiwynn, 40% for Wistron's own business. Credited US$13.0bn; ASIC 15% (8-25%). GPU systems US$11.0bn. Tier T5. Prior pool, 3Q24-2Q25 (5LayerCake derivation, T5). Worldwide GPU-server revenue for the four quarters = $217.5bn: 3Q24 $47.1bn and 4Q24 $47.0bn are the approved card's 3Q25 ($70.3bn) and 4Q25 ($74.7bn) GPU-server figures divided by IDC's embedded-GPU server growth of 49.4% (IDC 3Q25 release, 11 Dec 2025) and 59.1% (IDC 4Q25 release, 12 Mar 2026); 1Q25 $55.2bn and 2Q25 $68.3bn are the card's 1Q25 and 2Q25 totals less other-accelerated servers. China is IDC China's own accelerated-server value for the same window, $33.1bn (2H24 = 2024 $22.1bn less 1H24 $5.0bn = $17.1bn, plus 1H25 $16.0bn), of which GPU 69% = $22.8bn. GPU servers outside China = $217.5bn - $22.8bn = $194.7bn (range 175-212). Other-accelerated outside China = about $30bn worldwide (1Q25 $8.0bn and 2Q25 $8.2bn from the card's growth rates; 3Q24 and 4Q24 assumed about $7bn and $7.5bn) less China's non-GPU $10.3bn = $19.7bn (range 15-26). Share = US$11.0bn / 194.7bn."},"custom-accelerator-systems":{"low":2.156,"central":9.904,"high":36.636,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"LTM to 2025-06-30 (3Q24-2Q25)","periodEnd":"2025-06-30","source":{"url":"https://www.wiwynn.com/news/wiwynn-reports-fourth-quarter-2025-unaudited-financial-results","title":"Wiwynn Reports Fourth Quarter 2025 Unaudited Financial Results","publisher":"Wiwynn Corporation","publishedDate":"undated","accessedDate":"2026-10-22","tier":"T5","locator":"FY2025 consolidated revenue NT$950.663bn, +163.7%; Q4 2025 NT$292.440bn, +152.9%; AI-related products more than half of revenue"},"note":"ASIC servers taken as 15% (8-25%) of the credited AI total of US$13.0bn = US$2.0bn. Tier T5. Prior pool for this function: other-accelerated servers outside China about US$19.7bn (range 15-26), derived as described in the GPU-systems prior note of the same file."},"china":{"low":0,"central":0.223,"high":1.15,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"LTM to 2025-06-30","periodEnd":"2025-06-30","source":{"url":"https://www.wiwynn.com/news/wiwynn-reports-fourth-quarter-2025-unaudited-financial-results","title":"Wiwynn Reports Fourth Quarter 2025 Unaudited Financial Results","publisher":"Wiwynn Corporation","publishedDate":"undated","accessedDate":"2026-10-22","tier":"T5","locator":"FY2025 consolidated revenue NT$950.663bn, +163.7%; Q4 2025 NT$292.440bn, +152.9%; AI-related products more than half of revenue"},"note":"Placeholder carried unchanged from the current period, so it adds no share change."}},"activityGrowth":{"current":{"low":20006000000,"central":34836000000,"high":56283000000,"unit":"USD","basis":"AI server sales credited to the company as seller of the finished system, latest four quarters, US dollars","period":"LTM to 2026-06-30","periodEnd":"2026-06-30","source":{"url":"https://www.taipeitimes.com/News/biz/archives/2026/07/11/2003860550","title":"Wistron sales hit record (Taipei Times, 11 Jul 2026)","publisher":"Taipei Times","publishedDate":"2026-07-11","accessedDate":"2026-10-22","tier":"T5","locator":"Q2 revenue NT$895.44bn; first half NT$1.74tn, +94.1%; AI servers about 70% of revenue in the first two months of 2026 (company, March)"},"note":"Derived as in the share notes."},"prior":{"low":7006000000,"central":13007000000,"high":21982000000,"unit":"USD","basis":"AI server sales credited to the company as seller of the finished system, latest four quarters, US dollars","period":"LTM to 2025-06-30","periodEnd":"2025-06-30","source":{"url":"https://www.wiwynn.com/news/wiwynn-reports-fourth-quarter-2025-unaudited-financial-results","title":"Wiwynn Reports Fourth Quarter 2025 Unaudited Financial Results","publisher":"Wiwynn Corporation","publishedDate":"undated","accessedDate":"2026-10-22","tier":"T5","locator":"FY2025 consolidated revenue NT$950.663bn, +163.7%; Q4 2025 NT$292.440bn, +152.9%; AI-related products more than half of revenue"},"note":"Derived as in the prior-share note."}},"baselineRevenueUsd":{"low":53385000000,"central":60179000000,"high":69886000000,"unit":"USD","basis":"AI server revenue (group, before the credit rule), latest four quarters, US dollars","period":"LTM to 2026-06-30","periodEnd":"2026-06-30","source":{"url":"https://www.taipeitimes.com/News/biz/archives/2026/07/11/2003860550","title":"Wistron sales hit record (Taipei Times, 11 Jul 2026)","publisher":"Taipei Times","publishedDate":"2026-07-11","accessedDate":"2026-10-22","tier":"T5","locator":"Q2 revenue NT$895.44bn; first half NT$1.74tn, +94.1%; AI servers about 70% of revenue in the first two months of 2026 (company, March)"},"note":"Estimate: group revenue US$97.1bn x 62% (55-72%)."},"screening":{"gpu-systems":{"levelLow":1,"levelHigh":3},"custom-accelerator-systems":{"levelLow":1.5,"levelHigh":3.5},"china":{"levelLow":1,"levelHigh":2}},"checks":{"sourceSupport":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-22","passed":true,"sha256":"2752d661ab8aa17bc7c87285d2b61747ba20c1aa20d42331fa7920f93c47bd9c"},"consistency":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-22","passed":true,"sha256":"2752d661ab8aa17bc7c87285d2b61747ba20c1aa20d42331fa7920f93c47bd9c"}}},"super-micro":{"schema":"5lc-index-inputs/1","key":"infrastructure-servers-systems","entityId":"super-micro","name":"Supermicro","kind":"public","commercialActivityVerified":true,"shares":{"gpu-systems":{"low":9.838,"central":10.793,"high":12.277,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"FY2026 (four quarters to 2026-06-30), against IDC 3Q25-2Q26","periodEnd":"2026-06-30","source":{"url":"https://s204.q4cdn.com/707617056/files/doc_financials/2026/q4/V2/FQ426-SMCI-Prepared-Remarks.pdf","title":"Super Micro Computer, Inc. Fourth Quarter Fiscal 2026 Earnings Call, August 11, 2026 (prepared remarks)","publisher":"Super Micro Computer, Inc.","publishedDate":"2026-08-11","accessedDate":"2026-10-22","tier":"T5","locator":"CFO: FY26 revenue $39.1bn vs $22bn; Q4 revenue $11.1bn, AI solutions about 60% of Q4 vs over 80% in Q3; over $60bn of new orders in Q4"},"note":"Supermicro's AI revenue, derived from company statements: net sales per quarter Q1 FY26 $5.0bn, Q2 $12.7bn, Q3 $10.2bn, Q4 $11.1bn (Supermicro IR releases, https://ir.supermicro.com/news/news-details/2026/Supermicro-Announces-Fourth-Quarter-and-Full-Fiscal-Year-2026-Financial-Results/default.aspx; FY26 $39.1bn, Form 10-K FY2026). AI share of each quarter's revenue: Q1 not stated, taken as 75% (70-80%; Q4 FY25 was over 70% and Q2 over 90%); Q2 'AI GPU platforms ... over 90% of Q2 revenue' (Q2 FY26 prepared remarks, https://s204.q4cdn.com/707617056/files/doc_financials/2026/q2/SMCI-Q226-Prepared-Remarks-UPDATED.pdf), taken as 90% (88-92%); Q3 'AI GPU related platforms contributing over 80%' (Q3 FY26 prepared remarks, https://s204.q4cdn.com/707617056/files/doc_financials/2026/q3/SMCI-Q326-Prepared-Remarks.pdf), taken as 80% (80-85%); Q4 'AI solutions contributed approximately 60%' (Q4 FY26 prepared remarks), taken as 60% (55-65%). AI revenue = 5.0x0.75 + 12.68x0.90 + 10.2x0.80 + 11.1x0.60 = $29.98bn (range $28.9-31.6bn). Tier T5 (derived from company percentages; 'over X%' statements are floors, so the central is not a ceiling). The company sells finished racks and servers under its own name, so no credit adjustment is made. Pool: GPU servers and racks outside mainland China, US$277.8bn over 3Q25-2Q26 (IDC figures as used in the approved card: GPU $301.3bn of $373.4bn accelerated servers, less China at 9% and 70% GPU); range US$257-294bn. Share = US$30.0bn / 277.8bn (range US$28.9-31.6bn against the pool range)."},"custom-accelerator-systems":{"low":0,"central":0.121,"high":0.656,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"LTM to Jun 2026 (IDC 3Q25-2Q26)","periodEnd":"2026-06-30","source":{"url":"https://s204.q4cdn.com/707617056/files/doc_financials/2026/q4/V2/FQ426-SMCI-Prepared-Remarks.pdf","title":"Super Micro Computer, Inc. Fourth Quarter Fiscal 2026 Earnings Call, August 11, 2026 (prepared remarks)","publisher":"Super Micro Computer, Inc.","publishedDate":"2026-08-11","accessedDate":"2026-10-22","tier":"T5","locator":"CFO: FY26 revenue $39.1bn; AI solutions about 60% of Q4 revenue versus over 80% in Q3; 'one large data center/CSP customer representing 28% of revenue'"},"note":"Wide placeholder estimate, tier T5: no system sales built on cloud companies' own accelerators are shown in the sources opened. Upper end set at US$0.3bn; central at a quarter of that. This function is not material for the company. Pool: servers for non-GPU (other) accelerators outside mainland China, US$62.0bn over 3Q25-2Q26 (IDC 'other accelerated' servers $72.1bn less China's share, as in the approved card); range US$46-75bn. Share = US$0.075bn / 62bn (range US$0-0.3bn against the pool range)."},"china":{"low":0,"central":0.298,"high":1.53,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"LTM to Jun 2026 (IDC 3Q25-2Q26)","periodEnd":"2026-06-30","source":{"url":"https://s204.q4cdn.com/707617056/files/doc_financials/2026/q4/V2/FQ426-SMCI-Prepared-Remarks.pdf","title":"Super Micro Computer, Inc. Fourth Quarter Fiscal 2026 Earnings Call, August 11, 2026 (prepared remarks)","publisher":"Super Micro Computer, Inc.","publishedDate":"2026-08-11","accessedDate":"2026-10-22","tier":"T5","locator":"CFO: FY26 revenue $39.1bn; AI solutions about 60% of Q4 revenue versus over 80% in Q3; 'one large data center/CSP customer representing 28% of revenue'"},"note":"Wide placeholder estimate, tier T5: no sale of AI servers inside mainland China is shown in the sources opened. Upper end set at US$0.4bn; central at a quarter of that. This function is not material for the company. Pool: AI servers sold in mainland China, about US$33.6bn over 3Q25-2Q26 (9% of IDC's $373.4bn accelerated servers, as in the approved card); range US$26-45bn. Share = US$0.1bn / 33.6bn (range US$0-0.4bn against the pool range)."}},"priorShares":{"gpu-systems":{"low":6.226,"central":7.91,"high":10.057,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"FY2025 (four quarters to 2025-06-30), against 3Q24-2Q25","periodEnd":"2025-06-30","source":{"url":"https://s204.q4cdn.com/707617056/files/doc_financials/2025/q4/Q4FY25-Script.pdf","title":"Super Micro Computer, Inc. Fourth Quarter Fiscal 2025 Earnings Call script (FQ425)","publisher":"Super Micro Computer, Inc.","publishedDate":"2025-08-05","accessedDate":"2026-10-22","tier":"T5","locator":"CFO: Q4 FY25 revenue $5.8bn; AI GPU platforms over 70% of Q4 revenues; FY25 revenue $22bn, +47% over FY24 $15bn"},"note":"FY2025 net sales $22bn (Q4 FY25 script) x AI GPU platforms share taken as 70% (60-80%): the script says AI GPU platforms were over 70% of Q4 FY25 revenue, and the Q2 FY25 call put AI-related platforms above 70%; the FY25 share is not stated. = $15.4bn (13.2-17.6). Tier T5. Prior pool, 3Q24-2Q25 (5LayerCake derivation, T5). Worldwide GPU-server revenue for the four quarters = $217.5bn: 3Q24 $47.1bn and 4Q24 $47.0bn are the approved card's 3Q25 ($70.3bn) and 4Q25 ($74.7bn) GPU-server figures divided by IDC's embedded-GPU server growth of 49.4% (IDC 3Q25 release, 11 Dec 2025) and 59.1% (IDC 4Q25 release, 12 Mar 2026); 1Q25 $55.2bn and 2Q25 $68.3bn are the card's 1Q25 and 2Q25 totals less other-accelerated servers. China is IDC China's own accelerated-server value for the same window, $33.1bn (2H24 = 2024 $22.1bn less 1H24 $5.0bn = $17.1bn, plus 1H25 $16.0bn), of which GPU 69% = $22.8bn. GPU servers outside China = $217.5bn - $22.8bn = $194.7bn (range 175-212). Other-accelerated outside China = about $30bn worldwide (1Q25 $8.0bn and 2Q25 $8.2bn from the card's growth rates; 3Q24 and 4Q24 assumed about $7bn and $7.5bn) less China's non-GPU $10.3bn = $19.7bn (range 15-26). Share = US$15.4bn / 194.7bn (range US$13.2-17.6bn against the prior pool range)."},"custom-accelerator-systems":{"low":0,"central":0.121,"high":0.656,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"FY2025","periodEnd":"2025-06-30","source":{"url":"https://s204.q4cdn.com/707617056/files/doc_financials/2025/q4/Q4FY25-Script.pdf","title":"Super Micro Computer, Inc. Fourth Quarter Fiscal 2025 Earnings Call script (FQ425)","publisher":"Super Micro Computer, Inc.","publishedDate":"2025-08-05","accessedDate":"2026-10-22","tier":"T5","locator":"CFO: Q4 FY25 revenue $5.8bn; AI GPU platforms over 70% of Q4 revenues; FY25 revenue $22bn, +47% over FY24 $15bn"},"note":"Placeholder carried unchanged from the current period: no AI-server sales in this function appear in the sources opened for either period, so it adds no share change."},"china":{"low":0,"central":0.298,"high":1.53,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"FY2025","periodEnd":"2025-06-30","source":{"url":"https://s204.q4cdn.com/707617056/files/doc_financials/2025/q4/Q4FY25-Script.pdf","title":"Super Micro Computer, Inc. Fourth Quarter Fiscal 2025 Earnings Call script (FQ425)","publisher":"Super Micro Computer, Inc.","publishedDate":"2025-08-05","accessedDate":"2026-10-22","tier":"T5","locator":"CFO: Q4 FY25 revenue $5.8bn; AI GPU platforms over 70% of Q4 revenues; FY25 revenue $22bn, +47% over FY24 $15bn"},"note":"Placeholder carried unchanged from the current period: no AI-server sales in this function appear in the sources opened for either period, so it adds no share change."}},"activityGrowth":{"current":{"low":28925160000,"central":29983800000,"high":31552440000,"unit":"USD","basis":"AI GPU platform revenue, fiscal year, US dollars (company-stated shares of net sales)","period":"FY2026 (to 2026-06-30)","periodEnd":"2026-06-30","source":{"url":"https://s204.q4cdn.com/707617056/files/doc_financials/2026/q4/V2/FQ426-SMCI-Prepared-Remarks.pdf","title":"Super Micro Computer, Inc. Fourth Quarter Fiscal 2026 Earnings Call, August 11, 2026 (prepared remarks)","publisher":"Super Micro Computer, Inc.","publishedDate":"2026-08-11","accessedDate":"2026-10-22","tier":"T5","locator":"CFO: FY26 revenue $39.1bn vs $22bn; Q4 revenue $11.1bn, AI solutions about 60% of Q4 vs over 80% in Q3; over $60bn of new orders in Q4"},"note":"Derived as in the share note: $29.98bn."},"prior":{"low":13200000000,"central":15400000000,"high":17600000000,"unit":"USD","basis":"AI GPU platform revenue, fiscal year, US dollars (company-stated shares of net sales)","period":"FY2025 (to 2025-06-30)","periodEnd":"2025-06-30","source":{"url":"https://s204.q4cdn.com/707617056/files/doc_financials/2025/q4/Q4FY25-Script.pdf","title":"Super Micro Computer, Inc. Fourth Quarter Fiscal 2025 Earnings Call script (FQ425)","publisher":"Super Micro Computer, Inc.","publishedDate":"2025-08-05","accessedDate":"2026-10-22","tier":"T5","locator":"CFO: Q4 FY25 revenue $5.8bn; AI GPU platforms over 70% of Q4 revenues; FY25 revenue $22bn, +47% over FY24 $15bn"},"note":"Derived: $22bn x 70% (60-80%)."}},"committedExpansion":{"low":2,"central":2,"high":2,"unit":"ratio","basis":"New orders received in Q4 FY2026 (at least $60bn) ÷ AI GPU platform revenue, FY2026","period":"Orders received in Q4 FY2026","periodEnd":"2026-06-30","source":{"url":"https://ir.supermicro.com/news/news-details/2026/Supermicro-Provides-Fourth-Quarter-of-Fiscal-Year-2026-Preliminary-Business-Update/default.aspx","title":"Supermicro Provides Fourth Quarter of Fiscal Year 2026 Preliminary Business Update","publisher":"Super Micro Computer, Inc.","publishedDate":"2026-07-21","accessedDate":"2026-10-22","tier":"T5","locator":"Backlog rose to record levels; more than $60 billion of new orders received during the fourth quarter of fiscal 2026, expected to be delivered over future quarters"},"note":"The company says it received more than $60bn of new orders in Q4 FY2026, to be delivered over future quarters, and that backlog rose to record levels; the release warns some orders may not be firm commitments and may be cancelled or delayed. Backlog beyond these orders is not stated, so the ratio is a floor. Current annual activity $29.98bn (derived, see the share note). 60 ÷ 29.98 = 2.00. Tier T5 (the lower of the company's order statement, T2, and the derived denominator, T5)."},"baselineRevenueUsd":{"low":28925160000,"central":29983800000,"high":31552440000,"unit":"USD","basis":"AI GPU platform revenue, fiscal 2026, US dollars","period":"FY2026 (to 2026-06-30)","periodEnd":"2026-06-30","source":{"url":"https://s204.q4cdn.com/707617056/files/doc_financials/2026/q4/V2/FQ426-SMCI-Prepared-Remarks.pdf","title":"Super Micro Computer, Inc. Fourth Quarter Fiscal 2026 Earnings Call, August 11, 2026 (prepared remarks)","publisher":"Super Micro Computer, Inc.","publishedDate":"2026-08-11","accessedDate":"2026-10-22","tier":"T5","locator":"CFO: FY26 revenue $39.1bn vs $22bn; Q4 revenue $11.1bn, AI solutions about 60% of Q4 vs over 80% in Q3; over $60bn of new orders in Q4"},"note":"AI platform revenue only, derived from company-stated shares of net sales ($39.1bn FY26): $29.98bn."},"screening":{"gpu-systems":{"levelLow":1,"levelHigh":3},"custom-accelerator-systems":{"levelLow":1,"levelHigh":2},"china":{"levelLow":1,"levelHigh":2}},"checks":{"sourceSupport":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-22","passed":true,"sha256":"0a4784a6cae8e7f29c6f062afb4edf50ffb2bdbbab18d82a6657e38de662b478"},"consistency":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-22","passed":true,"sha256":"0a4784a6cae8e7f29c6f062afb4edf50ffb2bdbbab18d82a6657e38de662b478"}}},"inspur":{"schema":"5lc-index-inputs/1","key":"infrastructure-servers-systems","entityId":"inspur","name":"Inspur Information (IEIT Systems)","kind":"public","commercialActivityVerified":true,"shares":{"gpu-systems":{"low":0.802,"central":2.09,"high":3.577,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"LTM to 2026-06-30 (3Q25-2Q26)","periodEnd":"2026-06-30","source":{"url":"https://static.cninfo.com.cn/finalpage/2025-03-29/1222950880.PDF","title":"浪潮电子信息产业股份有限公司 2024 年年度报告全文 (IEIT SYSTEMS 2024 annual report)","publisher":"IEIT SYSTEMS Co., Ltd. (Inspur Information)","publishedDate":"2025-03-29","accessedDate":"2026-10-22","tier":"T5","locator":"Section 4.2(1) revenue composition: overseas RMB 3,408,139.09 万元 (29.70%), +256.98%; domestic RMB 8,068,597.60 万元 (70.30%); total RMB 114.77bn"},"note":"IEIT SYSTEMS (Inspur Information) sells overseas under its own name: in 2024 overseas revenue was RMB 34.08bn, 29.7% of RMB 114.77bn and up 257% (2024 annual report, section 4.2). The 2025 annual report (https://static.cninfo.com.cn/finalpage/2026-04-11/1225096638.PDF) gives only a product split (servers RMB 154.61bn of RMB 164.78bn, +43.25%) and no region split, and the H1 2026 report summary (http://static.cninfo.com.cn/finalpage/2026-08-29/1225530924.PDF) shows revenue RMB 84.38bn (+5.22%). LTM revenue = H2 2025 RMB 84.59bn (2025 less H1 2025 of 84.38 / 1.0522 = 80.19) + Q1 2026 RMB 35.47bn + Q2 2026 RMB 48.91bn (Q2 +46.7%; the H1 2026 results as summarised by Huachuang Securities, https://m.gelonghui.com/news/5305021) = US$24.2bn. Converted at the Federal Reserve's monthly-average CNY per US$ (https://fred.stlouisfed.org/data/EXCHUS): 3Q25 7.157, 4Q25 7.090, 1Q26 6.923, 2Q26 6.804. Overseas share taken as 30% (15-40%, the 2024 level) = US$7.3bn; AI servers 80% (65-95%) of it = US$5.8bn (US$2.4-9.2bn). Tier T5 (2024 regional mix carried to the LTM). The register separates Aivres and KAYTUS; IEIT's reports list only 'other overseas sibling companies' (兄弟公司, common control by Inspur Group) with RMB 8.0m of related-party sales in 2024, so these overseas sales are IEIT's own and do not include them. Pool: GPU servers and racks outside mainland China, US$277.8bn over 3Q25-2Q26 (IDC figures as used in the approved card: GPU $301.3bn of $373.4bn accelerated servers, less China at 9% and 70% GPU); range US$257-294bn. Share = US$5.8bn / 277.8bn (range US$2.4-9.2bn against the pool range). The screening figure was 0.144%."},"custom-accelerator-systems":{"low":0,"central":0.202,"high":1.09,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"LTM to Jun 2026 (IDC 3Q25-2Q26)","periodEnd":"2026-06-30","source":{"url":"https://www.idc.com/resource-center/press-releases/2q2026-wwserver-tracker/","title":"Worldwide Server Market Revenue Hits Record $166.3 Billion in the Second Quarter of 2026","publisher":"IDC","publishedDate":"2026-09-10","accessedDate":"2026-10-22","tier":"T5","locator":"Vendor table, 2Q26: Dell $22,240.6M, Supermicro $10,177.0M, Lenovo $8,411.3M, HPE $5,868.4M, IEIT Systems $4,001.5M (-8.1% YoY), ODM Direct $89,660.3M, Rest of Market $25,960.1M; GPU-accelerated $87.4B, other accelerated $27.5B"},"note":"Wide placeholder estimate, tier T5: no system sales built on cloud companies' own accelerators are shown in the sources opened. Upper end set at US$0.5bn; central at a quarter of that. This function is not material for the company. Pool: servers for non-GPU (other) accelerators outside mainland China, US$62.0bn over 3Q25-2Q26 (IDC 'other accelerated' servers $72.1bn less China's share, as in the approved card); range US$46-75bn. Share = US$0.125bn / 62bn (range US$0-0.5bn against the pool range)."},"china":{"low":14,"central":22,"high":32,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"1H25 (IDC China, six months to 2025-06-30)","periodEnd":"2025-06-30","source":{"url":"https://www.idc.com/resource-center/press-releases/idc%ef%bc%9aai%e6%9c%8d%e5%8a%a1%e5%99%a8%e5%90%8c%e6%af%94%e5%a2%9e%e9%95%bf%e8%b6%85%e4%b8%80%e5%80%8d%ef%bc%8c%e5%b8%82%e5%9c%ba%e7%8e%af%e5%a2%83%e6%b7%b1%e5%8c%96%e6%9c%ac%e5%9c%9f%e5%8c%96/","title":"IDC：AI服务器同比增长超一倍，市场环境深化本土化 (IDC: AI servers more than double year on year, 1H25)","publisher":"IDC","publishedDate":"2025-10-08","accessedDate":"2026-10-22","tier":"T5","locator":"By sales Inspur, H3C and Lenovo rank first to third with nearly 50% combined; China accelerated servers US$16.0bn in 1H25"},"note":"IDC China ranks Inspur first by accelerated-server sales in 1H25, 1H24 and 2024; in 1H25 Inspur, H3C and Lenovo together held nearly 50% (they held over 70% with Nettrix in 1H24, https://www.idc.com/resource-center/press-releases/idc%ef%bc%9aai%e6%9c%8d%e5%8a%a1%e5%99%a8%e5%b8%82%e5%9c%ba%e7%a8%b3%e6%ad%a5%e5%a2%9e%e9%95%bf%ef%bc%8c%e9%9d%9egpu%e6%9c%8d%e5%8a%a1%e5%99%a8%e5%b8%82%e5%9c%ba%e5%a2%9e%e9%80%9f%e8%b6%852%e5%80%8d/), but no vendor share is published. Inspur taken at 22% (14-32%): about half of the roughly 50% top-three total, with H3C 14% and Lenovo about 10.5% (Lenovo, Nettrix and Huawei were re-set in October 2026 to follow the 1H25 ranking). Cross-check: IEIT's domestic revenue was about 70% of LTM revenue of US$24bn, US$17bn, of which 44% would be the US$7.4bn this share implies; IEIT's first-half 2026 revenue grew only 5% against a faster-growing market. IDC's 1H25 data is more than 12 months old, so the engine widens the range. Tier T5 (allocation by 5LayerCake). Pool: AI servers sold in mainland China, about US$33.6bn over 3Q25-2Q26 (9% of IDC's $373.4bn accelerated servers, as in the approved card); range US$26-45bn. Share = 22% central of the pool."}},"baselineRevenueUsd":{"low":6040000000,"central":13317000000,"high":22741420000,"unit":"USD","basis":"AI server revenue, latest four quarters, US dollars","period":"LTM to 2026-06-30","periodEnd":"2026-06-30","source":{"url":"https://static.cninfo.com.cn/finalpage/2025-03-29/1222950880.PDF","title":"浪潮电子信息产业股份有限公司 2024 年年度报告全文 (IEIT SYSTEMS 2024 annual report)","publisher":"IEIT SYSTEMS Co., Ltd. (Inspur Information)","publishedDate":"2025-03-29","accessedDate":"2026-10-22","tier":"T5","locator":"Section 4.2(1) revenue composition: overseas RMB 3,408,139.09 万元 (29.70%), +256.98%; domestic RMB 8,068,597.60 万元 (70.30%); total RMB 114.77bn"},"note":"Estimate, not a reported figure: IEIT does not report AI server revenue separately (its 2025 annual report gives servers RMB 154.6bn of RMB 164.8bn and no AI split). AI server revenue = the three function amounts in this file: GPU systems outside China US$5.8bn (2.4-9.2), own-chip systems US$0.125bn (0-0.5) and China 22% of the US$33.6bn pool = US$7.4bn (14% of US$26bn to 32% of US$45bn: US$3.6-14.4bn), total US$13.3bn (US$6.0-24.1bn, capped at the server revenue of 94% x US$24.2bn = US$22.7bn). Total company revenue for the four quarters was US$24.2bn at the Federal Reserve's monthly-average CNY per US$ (https://fred.stlouisfed.org/data/EXCHUS) of 7.157, 7.090, 6.923 and 6.804 for 3Q25 to 2Q26; the earlier version of this file used that total as the baseline."},"screening":{"gpu-systems":{"levelLow":1,"levelHigh":2},"custom-accelerator-systems":{"levelLow":1,"levelHigh":2},"china":{"levelLow":1,"levelHigh":3}},"checks":{"sourceSupport":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-22","passed":true,"sha256":"fefe80e6ebcc51692db0bf5fc941c1e2a4c6c38acdfd1c19a45bc1ece0eba83f"},"consistency":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-22","passed":true,"sha256":"fefe80e6ebcc51692db0bf5fc941c1e2a4c6c38acdfd1c19a45bc1ece0eba83f"}}},"huawei":{"schema":"5lc-index-inputs/1","key":"infrastructure-servers-systems","entityId":"huawei","name":"Huawei (Atlas and Ascend systems)","kind":"public","commercialActivityVerified":true,"shares":{"gpu-systems":{"low":0,"central":0.09,"high":0.389,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"LTM to Jun 2026 (IDC 3Q25-2Q26)","periodEnd":"2026-06-30","source":{"url":"https://www.idc.com/?p=104362","title":"IDC press release on China's accelerated server market in 2024 (Chinese-language; page metadata dated 30 March 2025, dateline Beijing 31 March 2025)","publisher":"IDC","publishedDate":"2025-03-31","accessedDate":"2026-10-22","tier":"T5","locator":"China accelerated server market 2024 US$22.1bn (+134%); top three by revenue Inspur, Nettrix, H3C (over 50% together); top three by shipments Inspur, Nettrix, Huawei (nearly 55%); GPU servers 69%"},"note":"Wide placeholder estimate, tier T5: Atlas systems sold outside China are small and not documented in the sources opened. Upper end set at US$1bn; central at a quarter of that. This function is not material for the company. Pool: GPU servers and racks outside mainland China, US$277.8bn over 3Q25-2Q26 (IDC figures as used in the approved card: GPU $301.3bn of $373.4bn accelerated servers, less China at 9% and 70% GPU); range US$257-294bn. Share = US$0.25bn / 277.8bn (range US$0-1bn against the pool range)."},"custom-accelerator-systems":{"low":0,"central":0.484,"high":3.28,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"LTM to Jun 2026 (IDC 3Q25-2Q26)","periodEnd":"2026-06-30","source":{"url":"https://www.idc.com/?p=104362","title":"IDC press release on China's accelerated server market in 2024 (Chinese-language; page metadata dated 30 March 2025, dateline Beijing 31 March 2025)","publisher":"IDC","publishedDate":"2025-03-31","accessedDate":"2026-10-22","tier":"T5","locator":"China accelerated server market 2024 US$22.1bn (+134%); top three by revenue Inspur, Nettrix, H3C (over 50% together); top three by shipments Inspur, Nettrix, Huawei (nearly 55%); GPU servers 69%"},"note":"Ascend systems outside China are non-GPU accelerator systems; placeholder US$0.3bn (0-1.5), tier T5. Pool: servers for non-GPU (other) accelerators outside mainland China, US$62.0bn over 3Q25-2Q26 (IDC 'other accelerated' servers $72.1bn less China's share, as in the approved card); range US$46-75bn. Share = US$0.3bn / 62bn (range US$0-1.5bn against the pool range)."},"china":{"low":5,"central":9.5,"high":17,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"1H25 (IDC China, six months to 2025-06-30)","periodEnd":"2025-06-30","source":{"url":"https://www.idc.com/resource-center/press-releases/idc%ef%bc%9aai%e6%9c%8d%e5%8a%a1%e5%99%a8%e5%90%8c%e6%af%94%e5%a2%9e%e9%95%bf%e8%b6%85%e4%b8%80%e5%80%8d%ef%bc%8c%e5%b8%82%e5%9c%ba%e7%8e%af%e5%a2%83%e6%b7%b1%e5%8c%96%e6%9c%ac%e5%9c%9f%e5%8c%96/","title":"IDC：AI服务器同比增长超一倍，市场环境深化本土化 (IDC: AI servers more than double year on year, 1H25)","publisher":"IDC","publishedDate":"2025-10-08","accessedDate":"2026-10-22","tier":"T5","locator":"China accelerated servers US$16.0bn in 1H25 (more than double); by sales Inspur, H3C and Lenovo top three, nearly 50%; by shipments Inspur, H3C and Nettrix, about 43%"},"note":"IDC China names Huawei in the top three by accelerated-server shipments in 2024 (with Inspur and Nettrix, nearly 55%) but not by sales, and Huawei is in neither top three in the 1H25 release (by sales Inspur, H3C and Lenovo, nearly 50%; by shipments Inspur, H3C and Nettrix), so in 1H25 its sales share was below Lenovo's. Later evidence points up: IDC reports Huawei shipped 572,000 AI chips in China in the first half of 2026 against more than 300,000 a year earlier, a 28% share of China's AI accelerators (IDC via Reuters, TechNode 14 Oct 2026, https://technode.com/2026/10/14/nvidias-china-ai-chip-share-falls-to-34-as-huawei-reaches-28/), a growth of about 90% in cards. Part of these ship inside other vendors' servers (xFusion is the leading Ascend partner), so Huawei's own system sales are taken at 9.5% (5-17%) of the pool, a little below Lenovo's 10.5% because the only vendor-level ranking by sales (1H25) puts Lenovo third and Huawei outside the top three; the 2026 growth above is covered by the high end of the range, and the two ranges overlap widely. Tier T5 (allocation by 5LayerCake). IDC's 1H25 data is more than 12 months old, so the engine widens the range. Pool: AI servers sold in mainland China, about US$33.6bn over 3Q25-2Q26 (9% of IDC's $373.4bn accelerated servers, as in the approved card); range US$26-45bn. Share = 9.5% central of the pool."}},"activityGrowth":{"current":{"low":572000,"central":572000,"high":572000,"unit":"cards","basis":"AI accelerator cards shipped in China (all vendors' servers), first half of the year, units","period":"H1 2026 (six months to 2026-06-30)","periodEnd":"2026-06-30","source":{"url":"https://technode.com/2026/10/14/nvidias-china-ai-chip-share-falls-to-34-as-huawei-reaches-28/","title":"Nvidia's China AI chip share falls to 34% as Huawei reaches 28%","publisher":"TechNode (citing IDC via Reuters)","publishedDate":"2026-10-14","accessedDate":"2026-10-22","tier":"T4","locator":"Huawei 572,000 units in H1 2026, 28% share"},"note":"Huawei Ascend accelerator cards shipped in China (IDC): a proxy for Huawei's AI system activity; includes cards sold into other vendors' servers."},"prior":{"low":300000,"central":300000,"high":300000,"unit":"cards","basis":"AI accelerator cards shipped in China (all vendors' servers), first half of the year, units","period":"H1 2025 (six months to 2025-06-30)","periodEnd":"2025-06-30","source":{"url":"https://technode.com/2026/10/14/nvidias-china-ai-chip-share-falls-to-34-as-huawei-reaches-28/","title":"Nvidia's China AI chip share falls to 34% as Huawei reaches 28%","publisher":"TechNode (citing IDC via Reuters)","publishedDate":"2026-10-14","accessedDate":"2026-10-22","tier":"T4","locator":"Huawei 'more than 300,000' in the same period of 2025"},"note":"Reported as 'more than 300,000', so 300,000 is a floor and the growth rate is a ceiling-biased estimate."}},"screening":{"gpu-systems":{"levelLow":1,"levelHigh":2},"custom-accelerator-systems":{"levelLow":1.5,"levelHigh":3.5},"china":{"levelLow":1.5,"levelHigh":3.5}},"checks":{"sourceSupport":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-22","passed":true,"sha256":"07d23e00968216c2ae6380b2514baf26659038a6946c9fe1fbb9210f5da8d7bc"},"consistency":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-22","passed":true,"sha256":"07d23e00968216c2ae6380b2514baf26659038a6946c9fe1fbb9210f5da8d7bc"}}},"lenovo":{"schema":"5lc-index-inputs/1","key":"infrastructure-servers-systems","entityId":"lenovo","name":"Lenovo","kind":"public","commercialActivityVerified":true,"shares":{"gpu-systems":{"low":1.02,"central":1.502,"high":2.412,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"LTM to 2026-06-30 (ISG four quarters), against IDC 3Q25-2Q26","periodEnd":"2026-06-30","source":{"url":"https://doc.irasia.com/listco/hk/lenovo/interim/2027/int1q.pdf","title":"Lenovo Group FY2026/27 First Quarter Results Announcement","publisher":"Lenovo Group","publishedDate":"2026-08-13","accessedDate":"2026-10-22","tier":"T5","locator":"ISG revenue record US$8.5bn, +98% year on year; AI server pipeline US$54.0bn (pipeline, not backlog); AI server revenue triple-digit growth internationally"},"note":"Lenovo does not disclose AI server revenue in dollars. ISG revenue for the four quarters to June 2026 = FY2025/26 $19.2bn (Lenovo FY2025/26 results as reported by CRN Asia, an undated page, https://www.crnasia.com/news/2026/vendor/lenovo-closes-its-best-year-ever-and-asia-pacific-led-the-ch; the same result gives group revenue $83.1bn) - Q1 FY25/26 $4.3bn (= $8.5bn / 1.98) + Q1 FY26/27 $8.5bn (this announcement) = $23.4bn. AI servers taken as 35% of ISG (25-50%) = $8.2bn (5.9-11.7); the company says only that AI server revenue grew at triple digits internationally and that CSP and Enterprise & SMB revenue each nearly doubled. Tier T5. Split: China $3.5bn (10.5% of the China pool; IDC China 1H25 ranks Lenovo third by sales, with Inspur and H3C nearly 50% together, and Lenovo's announcement says its prior-year quarter included international GPU sales into China that did not recur), custom accelerators $0.5bn, GPU outside China $4.2bn. Pool: GPU servers and racks outside mainland China, US$277.8bn over 3Q25-2Q26 (IDC figures as used in the approved card: GPU $301.3bn of $373.4bn accelerated servers, less China at 9% and 70% GPU); range US$257-294bn. Share = US$4.2bn / 277.8bn (range US$3.0-6.2bn against the pool range). The earlier split put China at US$1.7bn and GPU outside China at US$6.0bn; it moves with the China share below."},"custom-accelerator-systems":{"low":0,"central":0.806,"high":2.609,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"LTM to 2026-06-30","periodEnd":"2026-06-30","source":{"url":"https://doc.irasia.com/listco/hk/lenovo/interim/2027/int1q.pdf","title":"Lenovo Group FY2026/27 First Quarter Results Announcement","publisher":"Lenovo Group","publishedDate":"2026-08-13","accessedDate":"2026-10-22","tier":"T5","locator":"ISG revenue record US$8.5bn, +98% year on year; AI server pipeline US$54.0bn (pipeline, not backlog); AI server revenue triple-digit growth internationally"},"note":"Wide placeholder estimate, tier T5: no sale of servers built on cloud companies' own accelerators is shown in the sources opened; central US$0.5bn, upper end US$1.2bn. Pool: servers for non-GPU (other) accelerators outside mainland China, US$62.0bn over 3Q25-2Q26 (IDC 'other accelerated' servers $72.1bn less China's share, as in the approved card); range US$46-75bn. Share = US$0.5bn / 62bn."},"china":{"low":4,"central":10.5,"high":16,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"LTM to 2026-06-30, judged from IDC China 1H25","periodEnd":"2026-06-30","source":{"url":"https://www.idc.com/resource-center/press-releases/idc%ef%bc%9aai%e6%9c%8d%e5%8a%a1%e5%99%a8%e5%90%8c%e6%af%94%e5%a2%9e%e9%95%bf%e8%b6%85%e4%b8%80%e5%80%8d%ef%bc%8c%e5%b8%82%e5%9c%ba%e7%8e%af%e5%a2%83%e6%b7%b1%e5%8c%96%e6%9c%ac%e5%9c%9f%e5%8c%96/","title":"IDC：AI服务器同比增长超一倍，市场环境深化本土化 (IDC: AI servers more than double year on year, 1H25)","publisher":"IDC","publishedDate":"2025-10-08","accessedDate":"2026-10-22","tier":"T5","locator":"By sales: Inspur, H3C and Lenovo rank first to third, nearly 50% combined (1H25)"},"note":"IDC China's 1H25 release ranks Inspur, H3C and Lenovo as the top three accelerated-server vendors by sales with nearly 50% of a US$16.0bn half-year, and Nettrix and Huawei outside the top three by sales (Nettrix third by shipments only); only the total of the three is given. With Inspur at 22% and H3C at 14% in their own files, the third seat is about 12-13% in 1H25. Lenovo is taken at 10.5% of the LTM pool (4-16%): a little below that seat because its Q1 FY26/27 announcement says AI server growth excludes 'international GPU sales in China in the prior-year quarter', that is Lenovo's 1H25 sales into China from outside China that did not recur (https://doc.irasia.com/listco/hk/lenovo/interim/2027/int1q.pdf), and well above its earlier 5% because the 1H25 ranking has Lenovo ahead of Nettrix and Huawei, now set at 8% and 9.5% in their files (the only vendor-level ranking by sales available is this one; Huawei's 2026 growth is covered by the high end of its range). Sanity check: Lenovo's whole China business had revenue of US$5.86bn in the June 2026 quarter, so US$3.5bn a year of AI servers would be about 15% of it. Tier T5 (allocation by 5LayerCake). IDC's 1H25 data is more than 12 months old, so the engine widens the range. Pool: AI servers sold in mainland China, about US$33.6bn over 3Q25-2Q26 (9% of IDC's $373.4bn accelerated servers, as in the approved card); range US$26-45bn. Share = US$3.5bn / 33.6bn (range US$1.3-5.4bn against the pool range)."}},"baselineRevenueUsd":{"low":5900000000,"central":8200000000,"high":11700000000,"unit":"USD","basis":"AI server revenue, latest four quarters to 2026-06-30, US dollars","period":"LTM to 2026-06-30","periodEnd":"2026-06-30","source":{"url":"https://doc.irasia.com/listco/hk/lenovo/interim/2027/int1q.pdf","title":"Lenovo Group FY2026/27 First Quarter Results Announcement","publisher":"Lenovo Group","publishedDate":"2026-08-13","accessedDate":"2026-10-22","tier":"T5","locator":"ISG revenue record US$8.5bn, +98% year on year; AI server pipeline US$54.0bn (pipeline, not backlog); AI server revenue triple-digit growth internationally"},"note":"Estimate, not a reported figure: 35% (25-50%) of ISG revenue of $23.4bn. Lenovo reports ISG revenue, not AI server revenue."},"screening":{"gpu-systems":{"levelLow":1,"levelHigh":3},"custom-accelerator-systems":{"levelLow":1,"levelHigh":2},"china":{"levelLow":1,"levelHigh":3}},"checks":{"sourceSupport":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-22","passed":true,"sha256":"4cb6fa2eb8ed8875c8214ce6ae44dd065b140f579f92d2b356bad64dbc454249"},"consistency":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-22","passed":true,"sha256":"4cb6fa2eb8ed8875c8214ce6ae44dd065b140f579f92d2b356bad64dbc454249"}}},"gigabyte":{"schema":"5lc-index-inputs/1","key":"infrastructure-servers-systems","entityId":"gigabyte","name":"GIGABYTE (Giga Computing)","kind":"public","commercialActivityVerified":true,"shares":{"gpu-systems":{"low":1.453,"central":2.213,"high":3.145,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"LTM to 2026-06-30 (3Q25-2Q26)","periodEnd":"2026-06-30","source":{"url":"https://www.taipeitimes.com/News/biz/archives/2026/08/18/2003862657","title":"Gigabyte expects second half of year to be prosperous (Taipei Times, 18 Aug 2026)","publisher":"Taipei Times","publishedDate":"2026-08-18","accessedDate":"2026-10-22","tier":"T5","locator":"Q2 2026 revenue NT$144.21bn; first half NT$249.27bn, +48.4%; GB-series AI servers a larger share of shipments"},"note":"Revenue (NT$bn): Q2 2026 144.21 and the first half NT$249.27bn, +48.4% (Taipei Times, 18 Aug 2026), so Q1 2026 105.06; Q3 2025 about 79.5 (first nine months NT$247.5bn, Taiwan News, 17 Nov 2025, https://www.taiwannews.com.tw/en/news/6245735, less a first half of about 168.0 = 249.27 / 1.484); Q4 2025 about 89.4 (2025 revenue about NT$336.9bn, implied by Citi's 2026 forecast of NT$485.1bn at +44%, TechNews 7 May 2026, https://finance.technews.tw/2026/05/07/gigabyte-ai-server/, less nine months). Converted quarter by quarter at the Federal Reserve's monthly-average NT$ per US$ (https://fred.stlouisfed.org/data/EXTAUS): 3Q24 32.28, 4Q24 32.34, 1Q25 32.89, 2Q25 30.83, 3Q25 29.95, 4Q25 31.03, 1Q26 31.62, 2Q26 31.60. LTM US$13.4bn. Server share: about 60% in Q3 2025 with AI over 80% of it (Taiwan News); brokers put server-related revenue above 70% of Q1 2026 revenue and 73% for 2026 (TechNews/Citi); taken as 60%, 65%, 71% and 75% by quarter. Server revenue US$9.2bn x AI 85% (75-92%) = US$7.9bn. Credit rule: only systems sold under the company's own name directly to data-centre operators count; contract assembly for other brands, and boards and modules sold to NVIDIA or other designers, earn no credit. Gigabyte sells servers under its own name (Giga Computing); 85% (70-95%) is credited = US$6.7bn; ASIC servers 4%, China 4%; GPU systems US$6.1bn (US$4.3-8.1bn). Tier T5. Gigabyte's own investor deck returned 403 and was not retried. Pool: GPU servers and racks outside mainland China, US$277.8bn over 3Q25-2Q26 (IDC figures as used in the approved card: GPU $301.3bn of $373.4bn accelerated servers, less China at 9% and 70% GPU); range US$257-294bn. Share = US$6.1bn / 277.8bn (range US$4.3-8.1bn against the pool range)."},"custom-accelerator-systems":{"low":0,"central":0.323,"high":1.75,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"LTM to Jun 2026 (IDC 3Q25-2Q26)","periodEnd":"2026-06-30","source":{"url":"https://www.taiwannews.com.tw/en/news/6245735","title":"Taiwan News, 17 November 2025 (Carol Yang): Gigabyte third-quarter results","publisher":"Taiwan News","publishedDate":"2025-11-17","accessedDate":"2026-10-22","tier":"T5","locator":"Server sales about 60% of Q3 2025 revenue, AI servers over 80% of that; NT$247.5bn revenue for the first three quarters of 2025"},"note":"Wide placeholder estimate, tier T5: no system sales built on cloud companies' own accelerators are shown in the sources opened. Upper end set at US$0.8bn; central at a quarter of that. This function is not material for the company. Pool: servers for non-GPU (other) accelerators outside mainland China, US$62.0bn over 3Q25-2Q26 (IDC 'other accelerated' servers $72.1bn less China's share, as in the approved card); range US$46-75bn. Share = US$0.2bn / 62bn (range US$0-0.8bn against the pool range)."},"china":{"low":0,"central":0.744,"high":3.83,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"LTM to Jun 2026 (IDC 3Q25-2Q26)","periodEnd":"2026-06-30","source":{"url":"https://www.taiwannews.com.tw/en/news/6245735","title":"Taiwan News, 17 November 2025 (Carol Yang): Gigabyte third-quarter results","publisher":"Taiwan News","publishedDate":"2025-11-17","accessedDate":"2026-10-22","tier":"T5","locator":"Server sales about 60% of Q3 2025 revenue, AI servers over 80% of that; NT$247.5bn revenue for the first three quarters of 2025"},"note":"Wide placeholder estimate, tier T5: no sale of AI servers inside mainland China is shown in the sources opened. Upper end set at US$1bn; central at a quarter of that. This function is not material for the company. Pool: AI servers sold in mainland China, about US$33.6bn over 3Q25-2Q26 (9% of IDC's $373.4bn accelerated servers, as in the approved card); range US$26-45bn. Share = US$0.25bn / 33.6bn (range US$0-1bn against the pool range)."}},"baselineRevenueUsd":{"low":6935000000,"central":7860000000,"high":8507000000,"unit":"USD","basis":"AI server revenue (before the credit rule), latest four quarters, US dollars","period":"LTM to 2026-06-30","periodEnd":"2026-06-30","source":{"url":"https://www.taipeitimes.com/News/biz/archives/2026/08/18/2003862657","title":"Gigabyte expects second half of year to be prosperous (Taipei Times, 18 Aug 2026)","publisher":"Taipei Times","publishedDate":"2026-08-18","accessedDate":"2026-10-22","tier":"T5","locator":"Q2 2026 revenue NT$144.21bn; first half NT$249.27bn, +48.4%; GB-series AI servers a larger share of shipments"},"note":"Estimate, not a reported figure: revenue US$13.4bn x server share x 85% AI."},"screening":{"gpu-systems":{"levelLow":1,"levelHigh":2.5},"custom-accelerator-systems":{"levelLow":1,"levelHigh":2},"china":{"levelLow":1,"levelHigh":2}},"checks":{"sourceSupport":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-22","passed":true,"sha256":"ba550af34c29c7a623b50674e86b7974f7040be9e707f5d7625f5adfc24d3226"},"consistency":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-22","passed":true,"sha256":"ba550af34c29c7a623b50674e86b7974f7040be9e707f5d7625f5adfc24d3226"}}},"asus":{"schema":"5lc-index-inputs/1","key":"infrastructure-servers-systems","entityId":"asus","name":"ASUS","kind":"public","commercialActivityVerified":true,"shares":{"gpu-systems":{"low":1.468,"central":2.075,"high":2.756,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"LTM to 2026-06-30 (3Q25-2Q26)","periodEnd":"2026-06-30","source":{"url":"https://www.taiwannews.com.tw/news/6420610","title":"Taiwan's Asus lifts full-year AI server revenue growth target to 150%","publisher":"Taiwan News","publishedDate":"2026-08-13","accessedDate":"2026-10-22","tier":"T5","locator":"ASUS server business more than NT$120bn in 2025; Q2 2026 AI server sales +200% y/y; AI servers about 90% of the server business; 2026 AI server growth target raised from 100% to at least 150%"},"note":"ASUS's server business made more than NT$120bn in 2025 and AI servers are about 90% of it; Q2 2026 AI server sales rose 200% year on year and the 2026 growth target was raised to at least 150% (Taiwan News, 13 Aug 2026). The Infrastructure Solutions Business Group (ISBG) was 36% of Q2 2026 revenue of NT$241bn = NT$86.8bn, with HGX-type systems about 60% and GB300 NVL72 racks about 30% of server mix (https://futurumgroup.com/insights/asus-q2-fy-2026-earnings-hit-record-revenue-on-ai-servers/). Earlier quarters are not reported: Q3 2025 about NT$32bn, Q4 2025 about 40 and Q1 2026 about 55 (assumed; they sum with Q1-Q2 2025 to the reported 2025 total of more than NT$120bn, and Q1 2026 is consistent with the company's guidance that first-quarter server revenue would about double). Converted quarter by quarter at the Federal Reserve's monthly-average NT$ per US$ (https://fred.stlouisfed.org/data/EXTAUS): 3Q24 32.28, 4Q24 32.34, 1Q25 32.89, 2Q25 30.83, 3Q25 29.95, 4Q25 31.03, 1Q26 31.62, 2Q26 31.60. Server business LTM US$6.8bn (+/-15%). AI servers 90% of it = US$6.2bn. Credit rule: only systems sold under the company's own name directly to data-centre operators count; contract assembly for other brands, and boards and modules sold to NVIDIA or other designers, earn no credit. ASUS sells servers under its own brand; 95% (85-100%) is credited. GPU systems US$5.8bn (US$4.3-7.1bn). Tier T5. The screening figure was 1.91%. Pool: GPU servers and racks outside mainland China, US$277.8bn over 3Q25-2Q26 (IDC figures as used in the approved card: GPU $301.3bn of $373.4bn accelerated servers, less China at 9% and 70% GPU); range US$257-294bn. Share = US$5.8bn / 277.8bn (range US$4.3-7.1bn against the pool range)."},"custom-accelerator-systems":{"low":0,"central":0.202,"high":1.09,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"LTM to Jun 2026 (IDC 3Q25-2Q26)","periodEnd":"2026-06-30","source":{"url":"https://futurumgroup.com/insights/asus-q2-fy-2026-earnings-hit-record-revenue-on-ai-servers/","title":"ASUS Q2 FY 2026 Earnings Hit Record Revenue on AI Servers (Olivier Blanchard)","publisher":"Futurum Group","publishedDate":"2026-08-14","accessedDate":"2026-10-22","tier":"T5","locator":"Q2 2026 revenue TWD 241bn; Infrastructure Solutions (server) group 36% of revenue; server mix about 60% HGX-type, 30% GB300 NVL72 racks, 10% other"},"note":"Wide placeholder estimate, tier T5: no system sales built on cloud companies' own accelerators are shown in the sources opened. Upper end set at US$0.5bn; central at a quarter of that. This function is not material for the company. Pool: servers for non-GPU (other) accelerators outside mainland China, US$62.0bn over 3Q25-2Q26 (IDC 'other accelerated' servers $72.1bn less China's share, as in the approved card); range US$46-75bn. Share = US$0.125bn / 62bn (range US$0-0.5bn against the pool range)."},"china":{"low":0,"central":0.446,"high":2.3,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"LTM to Jun 2026 (IDC 3Q25-2Q26)","periodEnd":"2026-06-30","source":{"url":"https://futurumgroup.com/insights/asus-q2-fy-2026-earnings-hit-record-revenue-on-ai-servers/","title":"ASUS Q2 FY 2026 Earnings Hit Record Revenue on AI Servers (Olivier Blanchard)","publisher":"Futurum Group","publishedDate":"2026-08-14","accessedDate":"2026-10-22","tier":"T5","locator":"Q2 2026 revenue TWD 241bn; Infrastructure Solutions (server) group 36% of revenue; server mix about 60% HGX-type, 30% GB300 NVL72 racks, 10% other"},"note":"Wide placeholder estimate, tier T5: no sale of AI servers inside mainland China is shown in the sources opened. Upper end set at US$0.6bn; central at a quarter of that. This function is not material for the company. Pool: AI servers sold in mainland China, about US$33.6bn over 3Q25-2Q26 (9% of IDC's $373.4bn accelerated servers, as in the approved card); range US$26-45bn. Share = US$0.15bn / 33.6bn (range US$0-0.6bn against the pool range)."}},"activityGrowth":{"current":{"low":2274000000,"central":2472000000,"high":2670000000,"unit":"USD","basis":"AI server revenue, single quarter, US dollars","period":"Q2 2026 (three months to 2026-06-30)","periodEnd":"2026-06-30","source":{"url":"https://www.taiwannews.com.tw/news/6420610","title":"Taiwan's Asus lifts full-year AI server revenue growth target to 150%","publisher":"Taiwan News","publishedDate":"2026-08-13","accessedDate":"2026-10-22","tier":"T5","locator":"ASUS server business more than NT$120bn in 2025; Q2 2026 AI server sales +200% y/y; AI servers about 90% of the server business; 2026 AI server growth target raised from 100% to at least 150%"},"note":"ISBG revenue NT$86.8bn (36% of NT$241bn, Futurum) x 90% AI at NT$31.60 per US$ (Federal Reserve 2Q26 average)."},"prior":{"low":778900000,"central":846600000,"high":914300000,"unit":"USD","basis":"AI server revenue, single quarter, US dollars","period":"Q2 2025 (three months to 2025-06-30)","periodEnd":"2025-06-30","source":{"url":"https://www.taiwannews.com.tw/news/6420610","title":"Taiwan's Asus lifts full-year AI server revenue growth target to 150%","publisher":"Taiwan News","publishedDate":"2026-08-13","accessedDate":"2026-10-22","tier":"T5","locator":"ASUS server business more than NT$120bn in 2025; Q2 2026 AI server sales +200% y/y; AI servers about 90% of the server business; 2026 AI server growth target raised from 100% to at least 150%"},"note":"One third of Q2 2026 in NT$ (Taiwan News: Q2 2026 AI server sales +200% year on year; Futurum, which describes ISBG revenue as having doubled, would give a higher prior value and a growth rate near +100%), NT$26.1bn at NT$30.83 per US$."}},"baselineRevenueUsd":{"low":5235000000,"central":6159000000,"high":7083000000,"unit":"USD","basis":"AI server revenue, latest four quarters, US dollars","period":"LTM to 2026-06-30","periodEnd":"2026-06-30","source":{"url":"https://www.taiwannews.com.tw/news/6420610","title":"Taiwan's Asus lifts full-year AI server revenue growth target to 150%","publisher":"Taiwan News","publishedDate":"2026-08-13","accessedDate":"2026-10-22","tier":"T5","locator":"ASUS server business more than NT$120bn in 2025; Q2 2026 AI server sales +200% y/y; AI servers about 90% of the server business; 2026 AI server growth target raised from 100% to at least 150%"},"note":"Estimate: server business US$6.8bn x 90% AI."},"screening":{"gpu-systems":{"levelLow":1,"levelHigh":2.5},"custom-accelerator-systems":{"levelLow":1,"levelHigh":2},"china":{"levelLow":1,"levelHigh":2}},"checks":{"sourceSupport":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-22","passed":true,"sha256":"049faef37cd99702e2f52c146a92d42bdb6ec3c93ed027a233c864e31aed89db"},"consistency":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-22","passed":true,"sha256":"049faef37cd99702e2f52c146a92d42bdb6ec3c93ed027a233c864e31aed89db"}}},"h3c":{"schema":"5lc-index-inputs/1","key":"infrastructure-servers-systems","entityId":"h3c","name":"H3C","kind":"unit","parentId":"unisplendour","commercialActivityVerified":true,"shares":{"gpu-systems":{"low":0,"central":0.045,"high":0.194,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"LTM to Jun 2026 (IDC 3Q25-2Q26)","periodEnd":"2026-06-30","source":{"url":"https://www.idc.com/?p=104362","title":"IDC press release on China's accelerated server market in 2024 (Chinese-language; page metadata dated 30 March 2025, dateline Beijing 31 March 2025)","publisher":"IDC","publishedDate":"2025-03-31","accessedDate":"2026-10-22","tier":"T5","locator":"China accelerated server market 2024 US$22.1bn (+134%); top three by revenue Inspur, Nettrix, H3C (over 50% together); top three by shipments Inspur, Nettrix, Huawei (nearly 55%); GPU servers 69%"},"note":"Wide placeholder estimate, tier T5: Sales outside China are not documented in the sources opened. Upper end set at US$0.5bn; central at a quarter of that. This function is not material for the company. Pool: GPU servers and racks outside mainland China, US$277.8bn over 3Q25-2Q26 (IDC figures as used in the approved card: GPU $301.3bn of $373.4bn accelerated servers, less China at 9% and 70% GPU); range US$257-294bn. Share = US$0.125bn / 277.8bn (range US$0-0.5bn against the pool range)."},"custom-accelerator-systems":{"low":0,"central":0.121,"high":0.656,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"LTM to Jun 2026 (IDC 3Q25-2Q26)","periodEnd":"2026-06-30","source":{"url":"https://www.idc.com/?p=104362","title":"IDC press release on China's accelerated server market in 2024 (Chinese-language; page metadata dated 30 March 2025, dateline Beijing 31 March 2025)","publisher":"IDC","publishedDate":"2025-03-31","accessedDate":"2026-10-22","tier":"T5","locator":"China accelerated server market 2024 US$22.1bn (+134%); top three by revenue Inspur, Nettrix, H3C (over 50% together); top three by shipments Inspur, Nettrix, Huawei (nearly 55%); GPU servers 69%"},"note":"Wide placeholder estimate, tier T5: no system sales built on cloud companies' own accelerators are shown in the sources opened. Upper end set at US$0.3bn; central at a quarter of that. This function is not material for the company. Pool: servers for non-GPU (other) accelerators outside mainland China, US$62.0bn over 3Q25-2Q26 (IDC 'other accelerated' servers $72.1bn less China's share, as in the approved card); range US$46-75bn. Share = US$0.075bn / 62bn (range US$0-0.3bn against the pool range)."},"china":{"low":9,"central":14,"high":20,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"1H25 (IDC China, six months to 2025-06-30)","periodEnd":"2025-06-30","source":{"url":"https://www.idc.com/resource-center/press-releases/idc%ef%bc%9aai%e6%9c%8d%e5%8a%a1%e5%99%a8%e5%90%8c%e6%af%94%e5%a2%9e%e9%95%bf%e8%b6%85%e4%b8%80%e5%80%8d%ef%bc%8c%e5%b8%82%e5%9c%ba%e7%8e%af%e5%a2%83%e6%b7%b1%e5%8c%96%e6%9c%ac%e5%9c%9f%e5%8c%96/","title":"IDC：AI服务器同比增长超一倍，市场环境深化本土化 (IDC: AI servers more than double year on year, 1H25)","publisher":"IDC","publishedDate":"2025-10-08","accessedDate":"2026-10-22","tier":"T5","locator":"China accelerated servers US$16.0bn in 1H25 (more than double); by sales Inspur, H3C and Lenovo top three, nearly 50%; by shipments Inspur, H3C and Nettrix, about 43%"},"note":"IDC China ranks H3C second by accelerated-server sales in 1H25 (Inspur, H3C and Lenovo together nearly 50%), second in 1H24 and third in 2024 (Inspur, Nettrix and H3C over 50%). H3C's total revenue rose 37.3% to RMB 49.98bn in H1 2026, domestic enterprise RMB 43.87bn (Ping An Securities note on Unisplendour's H1 2026 results, https://m.gelonghui.com/news/5306718; IDC's Q1 2026 China market shares in the same note: x86 servers 14.9%, No. 2). No vendor share is published; H3C taken at 14% (9-20%). Tier T5 (allocation by 5LayerCake). IDC's 1H25 data is more than 12 months old, so the engine widens the range. Pool: AI servers sold in mainland China, about US$33.6bn over 3Q25-2Q26 (9% of IDC's $373.4bn accelerated servers, as in the approved card); range US$26-45bn. Share = 14% central of the pool."}},"screening":{"gpu-systems":{"levelLow":1,"levelHigh":2},"custom-accelerator-systems":{"levelLow":1,"levelHigh":2},"china":{"levelLow":1,"levelHigh":3}},"checks":{"sourceSupport":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-22","passed":true,"sha256":"ea2a22437a02be73fb789f14af347690143b8ac4d728d1ae78ef516aae357f88"},"consistency":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-22","passed":true,"sha256":"ea2a22437a02be73fb789f14af347690143b8ac4d728d1ae78ef516aae357f88"}}},"xfusion":{"schema":"5lc-index-inputs/1","key":"infrastructure-servers-systems","entityId":"xfusion","name":"xFusion","kind":"private","commercialActivityVerified":true,"shares":{"gpu-systems":{"low":0.028,"central":0.119,"high":0.385,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"CY2025 (prospectus), against IDC 3Q25-2Q26","periodEnd":"2025-12-31","source":{"url":"https://news.10jqka.com.cn/20260810/c678822279.shtml","title":"超聚变：以全栈算力底座叩响资本市场，赋能智能体时代 (xFusion: knocking on the capital market with a full-stack computing base), via 同花顺 citing 中国上市公司网","publisher":"同花顺 (10jqka)","publishedDate":"2026-08-10","accessedDate":"2026-10-22","tier":"T5","locator":"AI server revenue RMB 296.29亿 (RMB 29.63bn) in 2025, 50.91% of main business revenue, against RMB 62.78亿 in 2023; general-purpose servers RMB 263.42亿; order backlog RMB 226.85亿 at 31 Dec 2025"},"note":"xFusion is private; its figures come from its ChiNext IPO application (accepted by the Shenzhen exchange on 22 May 2026), as reported by the press (the prospectus itself could not be located on the exchange site). Revenue was RMB 58.246bn in 2025 (RMB 25.092bn in 2023, 44.267bn in 2024; 华尔街见闻 report of the filing, https://wallstreetcn.com/articles/3772946; DIGITIMES, 21 May 2026, gives CNY58.2bn, https://www.digitimes.com/news/a20260521VL224.html). AI servers were RMB 29.629bn (50.91%), the page cited here. Converted at FRED's CY2025 mean of CNY 7.188 per US$ (https://fred.stlouisfed.org/data/EXCHUS) = US$4.12bn. This is calendar 2025, against an LTM pool to June 2026; xFusion's AI revenue is growing, so the high end adds 20%. The filing does not split revenue by region (the company says it serves over 110 countries); its AI customers are mainly large domestic internet companies and Ascend partners, so 92% is taken as sold in mainland China (80-98%) and 8% abroad (2-20%). An earlier version of this file rested on the 2024 IDC China release, which does not name xFusion, and on xFusion's homepage, which gives no figures. Sales abroad = US$0.33bn (US$0.08-0.99bn), all taken as GPU systems. Pool: GPU servers and racks outside mainland China, US$277.8bn over 3Q25-2Q26 (IDC figures as used in the approved card: GPU $301.3bn of $373.4bn accelerated servers, less China at 9% and 70% GPU); range US$257-294bn. Share = US$0.33bn / 277.8bn (range US$0.08-0.99bn against the pool range)."},"custom-accelerator-systems":{"low":0,"central":0.048,"high":0.326,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"CY2025 (prospectus), against IDC 3Q25-2Q26","periodEnd":"2025-12-31","source":{"url":"https://news.10jqka.com.cn/20260810/c678822279.shtml","title":"超聚变：以全栈算力底座叩响资本市场，赋能智能体时代 (xFusion: knocking on the capital market with a full-stack computing base), via 同花顺 citing 中国上市公司网","publisher":"同花顺 (10jqka)","publishedDate":"2026-08-10","accessedDate":"2026-10-22","tier":"T5","locator":"AI server revenue RMB 296.29亿 (RMB 29.63bn) in 2025, 50.91% of main business revenue, against RMB 62.78亿 in 2023; general-purpose servers RMB 263.42亿; order backlog RMB 226.85亿 at 31 Dec 2025"},"note":"T5 allowance, not a documented sale: nothing in the sources opened shows xFusion building servers around a cloud company's own accelerator outside China. Upper end set at US$0.15bn; central at US$0.03bn. Pool: servers for non-GPU (other) accelerators outside mainland China, US$62.0bn over 3Q25-2Q26 (IDC 'other accelerated' servers $72.1bn less China's share, as in the approved card); range US$46-75bn. Share = US$0.03bn / 62bn (range US$0-0.15bn against the pool range)."},"china":{"low":7.328,"central":11.287,"high":18.645,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"CY2025 (prospectus), against IDC 3Q25-2Q26","periodEnd":"2025-12-31","source":{"url":"https://news.10jqka.com.cn/20260810/c678822279.shtml","title":"超聚变：以全栈算力底座叩响资本市场，赋能智能体时代 (xFusion: knocking on the capital market with a full-stack computing base), via 同花顺 citing 中国上市公司网","publisher":"同花顺 (10jqka)","publishedDate":"2026-08-10","accessedDate":"2026-10-22","tier":"T5","locator":"AI server revenue RMB 296.29亿 (RMB 29.63bn) in 2025, 50.91% of main business revenue, against RMB 62.78亿 in 2023; general-purpose servers RMB 263.42亿; order backlog RMB 226.85亿 at 31 Dec 2025"},"note":"xFusion is private; its figures come from its ChiNext IPO application (accepted by the Shenzhen exchange on 22 May 2026), as reported by the press (the prospectus itself could not be located on the exchange site). Revenue was RMB 58.246bn in 2025 (RMB 25.092bn in 2023, 44.267bn in 2024; 华尔街见闻 report of the filing, https://wallstreetcn.com/articles/3772946; DIGITIMES, 21 May 2026, gives CNY58.2bn, https://www.digitimes.com/news/a20260521VL224.html). AI servers were RMB 29.629bn (50.91%), the page cited here. Converted at FRED's CY2025 mean of CNY 7.188 per US$ (https://fred.stlouisfed.org/data/EXCHUS) = US$4.12bn. This is calendar 2025, against an LTM pool to June 2026; xFusion's AI revenue is growing, so the high end adds 20%. The filing does not split revenue by region (the company says it serves over 110 countries); its AI customers are mainly large domestic internet companies and Ascend partners, so 92% is taken as sold in mainland China (80-98%) and 8% abroad (2-20%). An earlier version of this file rested on the 2024 IDC China release, which does not name xFusion, and on xFusion's homepage, which gives no figures. China AI servers = US$3.79bn (US$3.30-4.85bn). IDC China's 1H25 release ranks Inspur, H3C and Lenovo as the top three by sales and does not name xFusion. xFusion's calendar-2025 AI sales equal about 11% of the pool, about a point above Lenovo's 10.5% in its own file; the two are close and their ranges overlap. The order can differ from 1H25 because Lenovo's 1H25 position included international GPU sales into China that its Q1 FY26/27 announcement says did not recur, and xFusion's AI sales probably weighted to the second half of 2025. Pool: AI servers sold in mainland China, about US$33.6bn over 3Q25-2Q26 (9% of IDC's $373.4bn accelerated servers, as in the approved card); range US$26-45bn. Share = US$3.79bn / 33.6bn (range US$3.30-4.85bn against the pool range)."}},"screening":{"gpu-systems":{"levelLow":1,"levelHigh":2},"custom-accelerator-systems":{"levelLow":1,"levelHigh":2},"china":{"levelLow":1,"levelHigh":3}},"checks":{"sourceSupport":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-22","passed":true,"sha256":"6a7090e51ab82e50a042e41992c21321739744571b8df0b0a01f169a338d2049"},"consistency":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-22","passed":true,"sha256":"6a7090e51ab82e50a042e41992c21321739744571b8df0b0a01f169a338d2049"}}},"hpe":{"schema":"5lc-index-inputs/1","key":"infrastructure-servers-systems","entityId":"hpe","name":"Hewlett Packard Enterprise","kind":"public","commercialActivityVerified":true,"shares":{"gpu-systems":{"low":1.293,"central":1.548,"high":1.907,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system; the numerator is HPE's AI Systems revenue, which includes services as well as products","period":"Four fiscal quarters to 2026-07-31 (Q4 FY25-Q3 FY26), against IDC 3Q25-2Q26","periodEnd":"2026-07-31","source":{"url":"https://investors.hpe.com/~/media/Files/H/HP-Enterprise-IR/documents/q3-2026/q3-2026-earnings-presentation.pdf","title":"HPE Q3 FY26 Earnings (presentation, September 2, 2026)","publisher":"Hewlett Packard Enterprise","publishedDate":"2026-09-02","accessedDate":"2026-10-22","tier":"T5","locator":"Slide 9, 'Record AI backlog': AI Systems backlog ($B) by quarter Q4 FY24 2.4, 3.1, 3.2, 3.7, 4.7, 5.0, 5.9, Q3 FY26 6.8; AI Systems orders ($B) 0.5, 1.6, 1.1, 2.1, 1.9, 1.2, 1.8, 2.4"},"note":"HPE does not report AI systems revenue by quarter. Derived from the company's order and backlog chart (HPE Q3 FY26 presentation, slide 9, footnote 4: AI Systems include products and services, so the figure is a little off the card's server-sales basis and is flagged off basis, which widens its range by 25 points): revenue = opening backlog + orders - closing backlog. Q4 FY25 3.7+1.9-4.7 = $0.9bn; Q1 FY26 4.7+1.2-5.0 = $0.9bn; Q2 FY26 5.0+1.8-5.9 = $0.9bn; Q3 FY26 5.9+2.4-6.8 = $1.5bn, which matches the CEO's 'AI systems revenue for the quarter was almost $1.6 billion' (Q3 FY26 call, https://investors.hpe.com/~/media/Files/H/HP-Enterprise-IR/documents/q3-2026/q3-2026-transcript.pdf); Q2 FY26 orders of $1.8bn match the Q2 FY26 call (https://investors.hpe.com/~/media/Files/H/HP-Enterprise-IR/documents/q2-2026/q2-2026-transcript.pdf). Total $4.2bn; Q3 taken at $1.6bn gives $4.3bn (3.8-4.9 allowing for cancellations and rounding). Tier T5 (derived). The earlier screening figure ($6.0bn) was higher than this roll-forward supports. Pool: GPU servers and racks outside mainland China, US$277.8bn over 3Q25-2Q26 (IDC figures as used in the approved card: GPU $301.3bn of $373.4bn accelerated servers, less China at 9% and 70% GPU); range US$257-294bn. Share = US$4.3bn / 277.8bn (range US$3.8-4.9bn against the pool range).","offBasis":true},"custom-accelerator-systems":{"low":0,"central":0.121,"high":0.656,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"LTM to 2026-07-31","periodEnd":"2026-07-31","source":{"url":"https://investors.hpe.com/~/media/Files/H/HP-Enterprise-IR/documents/q3-2026/q3-2026-transcript.pdf","title":"Hewlett Packard Enterprise Third Quarter 2026 Earnings Conference Call (transcript)","publisher":"Hewlett Packard Enterprise","publishedDate":"2026-09-02","accessedDate":"2026-10-22","tier":"T5","locator":"Opening remarks: 'AI systems revenue for the quarter was almost $1.6 billion'"},"note":"Wide placeholder estimate, tier T5: no system sales built on cloud companies' own accelerators are shown in the sources opened. Upper end set at US$0.3bn; central at a quarter of that. This function is not material for the company. Pool: servers for non-GPU (other) accelerators outside mainland China, US$62.0bn over 3Q25-2Q26 (IDC 'other accelerated' servers $72.1bn less China's share, as in the approved card); range US$46-75bn. Share = US$0.075bn / 62bn (range US$0-0.3bn against the pool range)."},"china":{"low":0,"central":0.223,"high":1.15,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"LTM to 2026-07-31","periodEnd":"2026-07-31","source":{"url":"https://investors.hpe.com/~/media/Files/H/HP-Enterprise-IR/documents/q3-2026/q3-2026-transcript.pdf","title":"Hewlett Packard Enterprise Third Quarter 2026 Earnings Conference Call (transcript)","publisher":"Hewlett Packard Enterprise","publishedDate":"2026-09-02","accessedDate":"2026-10-22","tier":"T5","locator":"Opening remarks: 'AI systems revenue for the quarter was almost $1.6 billion'"},"note":"Wide placeholder estimate, tier T5: no sale of AI servers inside mainland China is shown in the sources opened. Upper end set at US$0.3bn; central at a quarter of that. This function is not material for the company. Pool: AI servers sold in mainland China, about US$33.6bn over 3Q25-2Q26 (9% of IDC's $373.4bn accelerated servers, as in the approved card); range US$26-45bn. Share = US$0.075bn / 33.6bn (range US$0-0.3bn against the pool range)."}},"priorShares":{"gpu-systems":{"low":1.887,"central":2.311,"high":2.914,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system; the numerator is HPE's AI Systems revenue, which includes services as well as products","period":"Four fiscal quarters to 2025-07-31 (Q4 FY24-Q3 FY25), against 3Q24-2Q25","periodEnd":"2025-07-31","source":{"url":"https://investors.hpe.com/~/media/Files/H/HP-Enterprise-IR/documents/q3-2026/q3-2026-earnings-presentation.pdf","title":"HPE Q3 FY26 Earnings (presentation, September 2, 2026)","publisher":"Hewlett Packard Enterprise","publishedDate":"2026-09-02","accessedDate":"2026-10-22","tier":"T5","locator":"Slide 9, 'Record AI backlog': AI Systems backlog ($B) by quarter Q4 FY24 2.4, 3.1, 3.2, 3.7, 4.7, 5.0, 5.9, Q3 FY26 6.8; AI Systems orders ($B) 0.5, 1.6, 1.1, 2.1, 1.9, 1.2, 1.8, 2.4"},"note":"Same roll-forward for Q1-Q3 FY25, computed from the slide: Q1 FY25 2.4+1.6-3.1 = $0.9bn; Q2 3.1+1.1-3.2 = $1.0bn; Q3 3.2+2.1-3.7 = $1.6bn) plus Q4 FY24 estimated at $1.0bn (0.8-1.2; the Q3 FY24 backlog is not on the slide) = $4.5bn (4.0-5.1). Tier T5. Prior pool, 3Q24-2Q25 (5LayerCake derivation, T5). Worldwide GPU-server revenue for the four quarters = $217.5bn: 3Q24 $47.1bn and 4Q24 $47.0bn are the approved card's 3Q25 ($70.3bn) and 4Q25 ($74.7bn) GPU-server figures divided by IDC's embedded-GPU server growth of 49.4% (IDC 3Q25 release, 11 Dec 2025) and 59.1% (IDC 4Q25 release, 12 Mar 2026); 1Q25 $55.2bn and 2Q25 $68.3bn are the card's 1Q25 and 2Q25 totals less other-accelerated servers. China is IDC China's own accelerated-server value for the same window, $33.1bn (2H24 = 2024 $22.1bn less 1H24 $5.0bn = $17.1bn, plus 1H25 $16.0bn), of which GPU 69% = $22.8bn. GPU servers outside China = $217.5bn - $22.8bn = $194.7bn (range 175-212). Other-accelerated outside China = about $30bn worldwide (1Q25 $8.0bn and 2Q25 $8.2bn from the card's growth rates; 3Q24 and 4Q24 assumed about $7bn and $7.5bn) less China's non-GPU $10.3bn = $19.7bn (range 15-26). Share = US$4.5bn / 194.7bn (range US$4.0-5.1bn against the prior pool range).","offBasis":true},"custom-accelerator-systems":{"low":0,"central":0.121,"high":0.656,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"Four fiscal quarters to 2025-07-31","periodEnd":"2025-07-31","source":{"url":"https://investors.hpe.com/~/media/Files/H/HP-Enterprise-IR/documents/q3-2026/q3-2026-earnings-presentation.pdf","title":"HPE Q3 FY26 Earnings (presentation, September 2, 2026)","publisher":"Hewlett Packard Enterprise","publishedDate":"2026-09-02","accessedDate":"2026-10-22","tier":"T5","locator":"Slide 9, 'Record AI backlog': AI Systems backlog ($B) by quarter Q4 FY24 2.4, 3.1, 3.2, 3.7, 4.7, 5.0, 5.9, Q3 FY26 6.8; AI Systems orders ($B) 0.5, 1.6, 1.1, 2.1, 1.9, 1.2, 1.8, 2.4"},"note":"Placeholder carried unchanged from the current period: no AI-server sales in this function appear in the sources opened for either period, so it adds no share change."},"china":{"low":0,"central":0.223,"high":1.15,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"Four fiscal quarters to 2025-07-31","periodEnd":"2025-07-31","source":{"url":"https://investors.hpe.com/~/media/Files/H/HP-Enterprise-IR/documents/q3-2026/q3-2026-earnings-presentation.pdf","title":"HPE Q3 FY26 Earnings (presentation, September 2, 2026)","publisher":"Hewlett Packard Enterprise","publishedDate":"2026-09-02","accessedDate":"2026-10-22","tier":"T5","locator":"Slide 9, 'Record AI backlog': AI Systems backlog ($B) by quarter Q4 FY24 2.4, 3.1, 3.2, 3.7, 4.7, 5.0, 5.9, Q3 FY26 6.8; AI Systems orders ($B) 0.5, 1.6, 1.1, 2.1, 1.9, 1.2, 1.8, 2.4"},"note":"Placeholder carried unchanged from the current period: no AI-server sales in this function appear in the sources opened for either period, so it adds no share change."}},"activityGrowth":{"current":{"low":3800000000,"central":4300000000,"high":4900000000,"unit":"USD","basis":"AI systems revenue (products and services), four fiscal quarters, US dollars","period":"Q4 FY25-Q3 FY26 (to 2026-07-31)","periodEnd":"2026-07-31","source":{"url":"https://investors.hpe.com/~/media/Files/H/HP-Enterprise-IR/documents/q3-2026/q3-2026-earnings-presentation.pdf","title":"HPE Q3 FY26 Earnings (presentation, September 2, 2026)","publisher":"Hewlett Packard Enterprise","publishedDate":"2026-09-02","accessedDate":"2026-10-22","tier":"T5","locator":"Slide 9, 'Record AI backlog': AI Systems backlog ($B) by quarter Q4 FY24 2.4, 3.1, 3.2, 3.7, 4.7, 5.0, 5.9, Q3 FY26 6.8; AI Systems orders ($B) 0.5, 1.6, 1.1, 2.1, 1.9, 1.2, 1.8, 2.4"},"note":"Derived from the backlog roll-forward; see the share note.","offBasis":true},"prior":{"low":4000000000,"central":4500000000,"high":5100000000,"unit":"USD","basis":"AI systems revenue (products and services), four fiscal quarters, US dollars","period":"Q4 FY24-Q3 FY25 (to 2025-07-31)","periodEnd":"2025-07-31","source":{"url":"https://investors.hpe.com/~/media/Files/H/HP-Enterprise-IR/documents/q3-2026/q3-2026-earnings-presentation.pdf","title":"HPE Q3 FY26 Earnings (presentation, September 2, 2026)","publisher":"Hewlett Packard Enterprise","publishedDate":"2026-09-02","accessedDate":"2026-10-22","tier":"T5","locator":"Slide 9, 'Record AI backlog': AI Systems backlog ($B) by quarter Q4 FY24 2.4, 3.1, 3.2, 3.7, 4.7, 5.0, 5.9, Q3 FY26 6.8; AI Systems orders ($B) 0.5, 1.6, 1.1, 2.1, 1.9, 1.2, 1.8, 2.4"},"note":"Derived from the backlog roll-forward; Q4 FY24 estimated.","offBasis":true}},"committedExpansion":{"low":1.58,"central":1.58,"high":1.58,"unit":"ratio","basis":"AI Systems backlog at the end of Q3 FY26 ÷ AI systems revenue, latest four fiscal quarters","period":"AI Systems backlog at 2026-07-31","periodEnd":"2026-07-31","source":{"url":"https://investors.hpe.com/~/media/Files/H/HP-Enterprise-IR/documents/q3-2026/q3-2026-earnings-presentation.pdf","title":"HPE Q3 FY26 Earnings (presentation, September 2, 2026)","publisher":"Hewlett Packard Enterprise","publishedDate":"2026-09-02","accessedDate":"2026-10-22","tier":"T5","locator":"Slide 9, 'Record AI backlog': AI Systems backlog ($B) by quarter Q4 FY24 2.4, 3.1, 3.2, 3.7, 4.7, 5.0, 5.9, Q3 FY26 6.8; AI Systems orders ($B) 0.5, 1.6, 1.1, 2.1, 1.9, 1.2, 1.8, 2.4"},"note":"The company's presentation shows AI Systems backlog (products and services) of $6.8bn at the end of Q3 FY26 (slide 9) and a further $3.5bn hyperscaler inferencing award after quarter end (not included). Current annual activity $4.3bn (derived, see the share note). 6.8 ÷ 4.3 = 1.58. Tier T5 (the lower of the company's backlog, T2, and the derived revenue, T5)."},"baselineRevenueUsd":{"low":3800000000,"central":4300000000,"high":4900000000,"unit":"USD","basis":"AI systems revenue (products and services), four fiscal quarters, US dollars","period":"Q4 FY25-Q3 FY26 (to 2026-07-31)","periodEnd":"2026-07-31","source":{"url":"https://investors.hpe.com/~/media/Files/H/HP-Enterprise-IR/documents/q3-2026/q3-2026-earnings-presentation.pdf","title":"HPE Q3 FY26 Earnings (presentation, September 2, 2026)","publisher":"Hewlett Packard Enterprise","publishedDate":"2026-09-02","accessedDate":"2026-10-22","tier":"T5","locator":"Slide 9, 'Record AI backlog': AI Systems backlog ($B) by quarter Q4 FY24 2.4, 3.1, 3.2, 3.7, 4.7, 5.0, 5.9, Q3 FY26 6.8; AI Systems orders ($B) 0.5, 1.6, 1.1, 2.1, 1.9, 1.2, 1.8, 2.4"},"note":"AI systems revenue only, derived (see the share note).","offBasis":true},"screening":{"gpu-systems":{"levelLow":1,"levelHigh":3},"custom-accelerator-systems":{"levelLow":1,"levelHigh":2},"china":{"levelLow":1,"levelHigh":2}},"checks":{"sourceSupport":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-22","passed":true,"sha256":"5b2670bb193a4273d68eea9ba13c0d0b33c526acd731b7fdc03f79c8e8a05c83"},"consistency":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-22","passed":true,"sha256":"5b2670bb193a4273d68eea9ba13c0d0b33c526acd731b7fdc03f79c8e8a05c83"}}},"nvidia":{"schema":"5lc-index-inputs/1","key":"infrastructure-servers-systems","entityId":"nvidia","name":"NVIDIA (DGX systems)","kind":"public","commercialActivityVerified":true,"shares":{"gpu-systems":{"low":0.283,"central":1.5,"high":5.404,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"LTM to 2026-07-26 (NVIDIA Q3 FY26-Q2 FY27)","periodEnd":"2026-07-26","source":{"url":"https://nvidianews.nvidia.com/news/nvidia-announces-financial-results-for-second-quarter-fiscal-2027","title":"NVIDIA Announces Financial Results for Second Quarter Fiscal 2027","publisher":"NVIDIA","publishedDate":"2026-08-26","accessedDate":"2026-10-22","tier":"T5","locator":"Data Center revenue $89.0bn in Q2 FY27, up 18% on the quarter and 117% on the year; used with the Q1 FY27 release (Data Center $75.2bn, up 92% on the year) and the FY2026 Form 10-K (Data Center $193,737m)"},"note":"NVIDIA does not report DGX or any finished-system revenue, so this is a judgment (T5) with stated anchors. Anchor 1, the size of the business: Data Center revenue for the four quarters to 26 Jul 2026 = US$277.8bn = FY2026 US$193.7bn (Form 10-K, https://www.sec.gov/Archives/edgar/data/1045810/000104581026000021/nvda-20260125.htm, local copy) - Q1 FY26 US$39.2bn (= Q1 FY27 US$75.2bn / 1.92, https://nvidianews.nvidia.com/news/nvidia-announces-financial-results-for-first-quarter-fiscal-2027) - Q2 FY26 US$41.0bn (= Q2 FY27 US$89.0bn / 2.17, the release cited here) + US$75.2bn + US$89.0bn. Anchor 2, how NVIDIA sells: the 10-K says Data Center products are 'typically delivered to customers as rack-scale systems, subsystems, or modules', that its direct customers include OEMs, ODMs, system integrators and distributors (two direct customers were 22% and 14% of FY2026 revenue) and that gross margin fell as NVIDIA moved 'from offering Hopper HGX systems to Blackwell full-scale datacenter solutions'. The data-centre operator then buys the finished rack from the OEM or ODM, which the card credits instead of NVIDIA. Only complete systems NVIDIA sells itself (DGX SuperPOD and BasePOD, which its DGX page says are 'available through a wide variety of partners', https://www.nvidia.com/en-us/data-center/dgx-platform/) count here; DGX Spark desktops are counted in Professional Visualization (10-K) and are excluded. Anchor 3, an outside size check: The Next Platform estimated in 2023 that DGX machines were about 20% of NVIDIA's systems revenue, about US$1.26bn, the rest being HGX boards sold to partners (https://www.nextplatform.com/2023/05/01/just-how-big-are-nvidias-server-and-networking-businesses/, T4, a published guess: 'Our guess is the DGX machines represent 20 percent of the revenues'). Judgment: NVIDIA's own finished systems are 1.5% of LTM Data Center revenue (range 0.3-5%) = US$4.2bn (US$0.8-13.9bn). The 5% high end is a quarter of the 2023 DGX proportion applied to all Data Center revenue, because Blackwell racks mostly pass through ODMs and OEMs; the 0.3% low end is a few SuperPOD deployments. Pool: GPU servers and racks outside mainland China, US$277.8bn over 3Q25-2Q26 (IDC figures as used in the approved card: GPU $301.3bn of $373.4bn accelerated servers, less China at 9% and 70% GPU); range US$257-294bn. Share = US$4.2bn / 277.8bn (range US$0.8-13.9bn against the pool range)."},"custom-accelerator-systems":{"participates":false,"evidence":{"url":"https://www.sec.gov/Archives/edgar/data/1045810/000104581026000021/nvda-20260125.htm","title":"NVIDIA Form 10-K for fiscal year ended January 25, 2026 (read from the local copy .cache/sec/nvidia)","publisher":"NVIDIA","publishedDate":"2026-02-25","accessedDate":"2026-10-22","tier":"T1","locator":"Item 1, Data Center: platform of NVIDIA GPUs, CPUs, interconnects and software; NVLink Fusion lets custom CPU and XPU designers connect to NVIDIA's platform (an interconnect, not a system sale); no system built around another company's accelerator is described"}},"china":{"participates":false,"evidence":{"url":"https://www.sec.gov/Archives/edgar/data/1045810/000104581026000021/nvda-20260125.htm","title":"NVIDIA Form 10-K for fiscal year ended January 25, 2026 (read from the local copy .cache/sec/nvidia)","publisher":"NVIDIA","publishedDate":"2026-02-25","accessedDate":"2026-10-22","tier":"T1","locator":"Item 1A: 'we are already effectively foreclosed from the China market by U.S. export controls'; export rules cover DGX and other systems with A100, H100, GB200 NVL72 and B200 chips"}}},"screening":{"gpu-systems":{"levelLow":1,"levelHigh":2.5}},"checks":{"sourceSupport":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-22","passed":true,"sha256":"1364585fb72b5cfebc0b0853523e895c14d41afeebdaa7f79601aba8672b326f"},"consistency":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-22","passed":true,"sha256":"1364585fb72b5cfebc0b0853523e895c14d41afeebdaa7f79601aba8672b326f"}}},"celestica":{"schema":"5lc-index-inputs/1","key":"infrastructure-servers-systems","entityId":"celestica","name":"Celestica","kind":"public","commercialActivityVerified":true,"shares":{"gpu-systems":{"low":0,"central":0.18,"high":0.583,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"Estimate to 2026-06-30","periodEnd":"2026-06-30","source":{"url":"https://www.sec.gov/Archives/edgar/data/1030894/000103089426000011/cls-20251231.htm","title":"Celestica Inc. Annual Report on Form 10-K for the fiscal year ended December 31, 2025 (read from the local copy .cache/sec/celestica)","publisher":"Celestica Inc.","publishedDate":"2026-02-27","accessedDate":"2026-10-22","tier":"T5","locator":"Item 1 / MD&A: CCS segment revenue $9.19bn in 2025 (74% of revenue); HPS revenue 41% of total revenue; three customers 32%, 14%, 12% of revenue"},"note":"No GPU-system sales identified in the sources opened; placeholder US$0.5bn (0-1.5), tier T5. Pool: GPU servers and racks outside mainland China, US$277.8bn over 3Q25-2Q26 (IDC figures as used in the approved card: GPU $301.3bn of $373.4bn accelerated servers, less China at 9% and 70% GPU); range US$257-294bn. Share = US$0.5bn / 277.8bn (range US$0-1.5bn against the pool range)."},"custom-accelerator-systems":{"low":1.2,"central":2.903,"high":6.522,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"LTM to 2026-06-30 (3Q25-2Q26)","periodEnd":"2026-06-30","source":{"url":"https://www.sec.gov/Archives/edgar/data/1030894/000103089426000011/cls-20251231.htm","title":"Celestica Inc. Form 10-K for the fiscal year ended 31 December 2025","publisher":"Celestica Inc. (SEC filing)","publishedDate":"2026-02-27","accessedDate":"2026-10-22","tier":"T5","locator":"MD&A: CCS segment revenue 2025 $9.19bn (+42%); Communications +81% (+$3,179.7M); Enterprise -19% on an AI/ML compute program transition; HPS 41% of revenue ($5.0bn)"},"note":"Celestica's Enterprise end market (servers, storage and AI/ML compute for hyperscalers) was about $2.09bn in 2025 and $2.57bn in 2024 (derived from the 10-K: CCS $9.19bn; Communications up 81%, +$3,179.7M, so $7.1bn; the rest is Enterprise, down 19% on 'a technology transition in an AI/machine learning (ML) compute program with one of our hyperscaler customers'). The company expects 'volume growth in 2026 in our AI/ML compute programs with our hyperscaler customers'. Q2 2026 revenue was $4.70bn (+62%), CCS $3.81bn (+84%), and 2026 revenue is guided to $20.5bn (8-K, https://www.sec.gov/Archives/edgar/data/0001030894/000103089426000043/cls-20260630prcondensedfs.htm), but no Enterprise or AI figure is given for 2026. AI compute for cloud companies' own accelerators taken as $1.8bn for the four quarters to June 2026 (0.9-3.0), most of Enterprise growth being AI; Celestica sells these systems under its own design and manufacturing contract, so no credit adjustment. Tier T5 (derived; no AI revenue is disclosed). The screening figure ($3.5bn, 45% of an assumed HPS total) is replaced because HPS is mainly networking switches. Pool: servers for non-GPU (other) accelerators outside mainland China, US$62.0bn over 3Q25-2Q26 (IDC 'other accelerated' servers $72.1bn less China's share, as in the approved card); range US$46-75bn. Share = US$1.8bn / 62bn (range US$0.9-3.0bn against the pool range)."},"china":{"low":0,"central":0.074,"high":0.383,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"Estimate to 2026-06-30","periodEnd":"2026-06-30","source":{"url":"https://www.sec.gov/Archives/edgar/data/1030894/000103089426000011/cls-20251231.htm","title":"Celestica Inc. Annual Report on Form 10-K for the fiscal year ended December 31, 2025 (read from the local copy .cache/sec/celestica)","publisher":"Celestica Inc.","publishedDate":"2026-02-27","accessedDate":"2026-10-22","tier":"T5","locator":"Item 1 / MD&A: CCS segment revenue $9.19bn in 2025 (74% of revenue); HPS revenue 41% of total revenue; three customers 32%, 14%, 12% of revenue"},"note":"Wide placeholder estimate, tier T5: no sale of AI servers inside mainland China is shown in the sources opened. Upper end set at US$0.1bn; central at a quarter of that. This function is not material for the company. Pool: AI servers sold in mainland China, about US$33.6bn over 3Q25-2Q26 (9% of IDC's $373.4bn accelerated servers, as in the approved card); range US$26-45bn. Share = US$0.025bn / 33.6bn (range US$0-0.1bn against the pool range)."}},"screening":{"gpu-systems":{"levelLow":1,"levelHigh":2},"custom-accelerator-systems":{"levelLow":1.5,"levelHigh":3.5},"china":{"levelLow":1,"levelHigh":2}},"checks":{"sourceSupport":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-22","passed":true,"sha256":"0245b76acc41d9abc50e60d5f8a1317c55aa57b8f245190d5b1a5e4a922f5fe8"},"consistency":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-22","passed":true,"sha256":"0245b76acc41d9abc50e60d5f8a1317c55aa57b8f245190d5b1a5e4a922f5fe8"}}},"inventec":{"schema":"5lc-index-inputs/1","key":"infrastructure-servers-systems","entityId":"inventec","name":"Inventec","kind":"public","commercialActivityVerified":true,"shares":{"gpu-systems":{"low":0.199,"central":0.749,"high":2.724,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"LTM to 2026-06-30 (3Q25-2Q26)","periodEnd":"2026-06-30","source":{"url":"https://www.taipeitimes.com/News/biz/archives/2025/11/12/2003847039","title":"Inventec expects revenue growth in 2026 (Taipei Times, 12 Nov 2025)","publisher":"Taipei Times","publishedDate":"2025-11-12","accessedDate":"2026-10-22","tier":"T5","locator":"Q3 2025 revenue NT$176.29bn; servers 41-45% of revenue; AI servers 40-50% of server revenue, above 50% expected in 2026; major assembly partner for Dell, HP and Lenovo servers"},"note":"Revenue (NT$bn): Q3 2025 176.29 (Taipei Times, 12 Nov 2025); Q1 2026 200.31, +16.9% q/q (Taiwan News, 9 Apr 2026, https://www.taiwannews.com.tw/news/6337378), so Q4 2025 about 171.4; Q2 2026 269.87, +44.65% y/y (Taipei Times, 11 Jul 2026, https://www.taipeitimes.com/News/biz/archives/2026/07/11/2003860550). Converted quarter by quarter at the Federal Reserve's monthly-average NT$ per US$ (https://fred.stlouisfed.org/data/EXTAUS): 3Q24 32.28, 4Q24 32.34, 1Q25 32.89, 2Q25 30.83, 3Q25 29.95, 4Q25 31.03, 1Q26 31.62, 2Q26 31.60. LTM US$26.3bn. Servers were 41-45% of Q3 2025 revenue and AI servers 40-50% of server revenue (Taipei Times); for later quarters the server share is taken as 45%, 50%, 58% and the AI share of server revenue as 48%, 52%, 54% (management expects AI above 50% of server revenue in 2026 and double-digit AI shipment growth in Q2 with faster growth in Q3-Q4). AI server revenue US$6.7bn (+/-35%). Credit rule: only systems sold under the company's own name directly to data-centre operators count; contract assembly for other brands, and boards and modules sold to NVIDIA or other designers, earn no credit. Inventec is 'a major assembly partner for level 6 and level 10 servers for Dell, HP and Lenovo' (Taipei Times), so only 45% (25-70%) is credited = US$3.0bn; ASIC servers (Google TPU programs) 30% (20-45%) of that; GPU systems US$2.1bn (US$0.6-5.0bn). Tier T5. A later Taipei Times line (11 Jul 2026) says AI servers should exceed 50% of total revenue in 2026 against over 40% in 2025; that reading conflicts with the company's Q3 2025 statement on server revenue and is not used. The screening figure was 1.73%. Pool: GPU servers and racks outside mainland China, US$277.8bn over 3Q25-2Q26 (IDC figures as used in the approved card: GPU $301.3bn of $373.4bn accelerated servers, less China at 9% and 70% GPU); range US$257-294bn. Share = US$2.1bn / 277.8bn (range US$0.6-5.0bn against the pool range). Conflicting statement and why the range is wider: Taipei Times (11 Jul 2026, https://www.taipeitimes.com/News/biz/archives/2026/07/11/2003860550) quotes the company as expecting AI server revenue above 50% of total revenue in 2026 against above 40% last year; applied to the quarterly revenue above (40% to 3Q25-4Q25, 50% to 1H26) that gives AI server revenue of about US$12.0bn, 1.8 times the central US$6.65bn. The central keeps the company's Nov 2025 breakdown (servers 41-45% of revenue, AI 40-50% of server revenue) because the July line would make AI almost all of server revenue, which conflicts with that breakdown and with the July article's own statement that Inventec also ships ASIC servers and mainly assembles Nvidia B300 systems; the high end of AI server revenue is raised from US$9.0bn to US$12.5bn so the range covers the July reading, and the high ends of both shares follow. "},"custom-accelerator-systems":{"low":0.288,"central":1.449,"high":8.56,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"LTM to 2026-06-30 (3Q25-2Q26)","periodEnd":"2026-06-30","source":{"url":"https://www.taipeitimes.com/News/biz/archives/2025/11/12/2003847039","title":"Inventec expects revenue growth in 2026 (Taipei Times, 12 Nov 2025)","publisher":"Taipei Times","publishedDate":"2025-11-12","accessedDate":"2026-10-22","tier":"T5","locator":"Q3 2025 revenue NT$176.29bn; servers 41-45% of revenue; AI servers 40-50% of server revenue, above 50% expected in 2026; major assembly partner for Dell, HP and Lenovo servers"},"note":"ASIC-based servers taken as 30% (20-45%) of the credited AI total of US$3.0bn = US$0.9bn (the DIGITIMES headline 'Google TPU stays strong, Inventec confident for 2026 server outlook', https://www.digitimes.com/news/a20251210PD211/asic-2026-google-tpu-inventec.html, was seen only in search results). Tier T5. Pool: servers for non-GPU (other) accelerators outside mainland China, US$62.0bn over 3Q25-2Q26 (IDC 'other accelerated' servers $72.1bn less China's share, as in the approved card); range US$46-75bn. Share = US$0.9bn / 62bn. The high end covers the July 2026 reading of AI server revenue (about US$12.0bn against the central US$6.65bn), as explained in the GPU-systems note: 12.5 x 0.70 x 0.45 = US$3.9bn against the pool low of US$46bn. Taipei Times (11 Jul 2026) also says Inventec ships ASIC-based servers, which supports a minority ASIC share; the 30% remains an assumption."},"china":{"low":0,"central":0.149,"high":0.766,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"LTM to Jun 2026 (IDC 3Q25-2Q26)","periodEnd":"2026-06-30","source":{"url":"https://www.taipeitimes.com/News/biz/archives/2026/07/11/2003860550","title":"Taipei Times, 11 July 2026, p. 9 (Meryl Kao): Wistron and Inventec second-quarter revenue","publisher":"Taipei Times","publishedDate":"2026-07-11","accessedDate":"2026-10-22","tier":"T5","locator":"Inventec Q2 2026 revenue NT$269.87bn (+44.65%); AI servers expected above 50% of total revenue this year, above 40% last year; Wistron AI servers about 70% of revenue in Jan-Feb 2026"},"note":"Wide placeholder estimate, tier T5: no sale of AI servers inside mainland China is shown in the sources opened. Upper end set at US$0.2bn; central at a quarter of that. This function is not material for the company. Pool: AI servers sold in mainland China, about US$33.6bn over 3Q25-2Q26 (9% of IDC's $373.4bn accelerated servers, as in the approved card); range US$26-45bn. Share = US$0.05bn / 33.6bn (range US$0-0.2bn against the pool range)."}},"baselineRevenueUsd":{"low":4325000000,"central":6654000000,"high":12500000000,"unit":"USD","basis":"AI server revenue (before the credit rule), latest four quarters, US dollars","period":"LTM to 2026-06-30","periodEnd":"2026-06-30","source":{"url":"https://www.taipeitimes.com/News/biz/archives/2025/11/12/2003847039","title":"Inventec expects revenue growth in 2026 (Taipei Times, 12 Nov 2025)","publisher":"Taipei Times","publishedDate":"2025-11-12","accessedDate":"2026-10-22","tier":"T5","locator":"Q3 2025 revenue NT$176.29bn; servers 41-45% of revenue; AI servers 40-50% of server revenue, above 50% expected in 2026; major assembly partner for Dell, HP and Lenovo servers"},"note":"Estimate, not a reported figure: revenue US$26.3bn x server share x AI share by quarter. High end raised from US$9.0bn to US$12.5bn to cover the 11 Jul 2026 Taipei Times reading (AI above 40% of 2025 and 50% of 2026 revenue, about US$12.0bn)."},"screening":{"gpu-systems":{"levelLow":1,"levelHigh":2.5},"custom-accelerator-systems":{"levelLow":1,"levelHigh":3},"china":{"levelLow":1,"levelHigh":2}},"checks":{"sourceSupport":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-22","passed":true,"sha256":"c519f8be1dd98b5e018e200ff6fcf4c4d21ba3ba86379ca5ef1054d789055eb4"},"consistency":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-22","passed":true,"sha256":"c519f8be1dd98b5e018e200ff6fcf4c4d21ba3ba86379ca5ef1054d789055eb4"}}},"nettrix":{"schema":"5lc-index-inputs/1","key":"infrastructure-servers-systems","entityId":"nettrix","name":"Nettrix","kind":"private","commercialActivityVerified":true,"shares":{"gpu-systems":{"low":0,"central":0.045,"high":0.194,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"LTM to Jun 2026 (IDC 3Q25-2Q26)","periodEnd":"2026-06-30","source":{"url":"https://www.idc.com/?p=104362","title":"IDC press release on China's accelerated server market in 2024 (Chinese-language; page metadata dated 30 March 2025, dateline Beijing 31 March 2025)","publisher":"IDC","publishedDate":"2025-03-31","accessedDate":"2026-10-22","tier":"T5","locator":"China accelerated server market 2024 US$22.1bn (+134%); top three by revenue Inspur, Nettrix, H3C (over 50% together); top three by shipments Inspur, Nettrix, Huawei (nearly 55%); GPU servers 69%"},"note":"Wide placeholder estimate, tier T5: Sales outside China are not documented in the sources opened. Upper end set at US$0.5bn; central at a quarter of that. This function is not material for the company. Pool: GPU servers and racks outside mainland China, US$277.8bn over 3Q25-2Q26 (IDC figures as used in the approved card: GPU $301.3bn of $373.4bn accelerated servers, less China at 9% and 70% GPU); range US$257-294bn. Share = US$0.125bn / 277.8bn (range US$0-0.5bn against the pool range)."},"custom-accelerator-systems":{"low":0,"central":0.121,"high":0.656,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"LTM to Jun 2026 (IDC 3Q25-2Q26)","periodEnd":"2026-06-30","source":{"url":"https://www.idc.com/?p=104362","title":"IDC press release on China's accelerated server market in 2024 (Chinese-language; page metadata dated 30 March 2025, dateline Beijing 31 March 2025)","publisher":"IDC","publishedDate":"2025-03-31","accessedDate":"2026-10-22","tier":"T5","locator":"China accelerated server market 2024 US$22.1bn (+134%); top three by revenue Inspur, Nettrix, H3C (over 50% together); top three by shipments Inspur, Nettrix, Huawei (nearly 55%); GPU servers 69%"},"note":"Wide placeholder estimate, tier T5: no system sales built on cloud companies' own accelerators are shown in the sources opened. Upper end set at US$0.3bn; central at a quarter of that. This function is not material for the company. Pool: servers for non-GPU (other) accelerators outside mainland China, US$62.0bn over 3Q25-2Q26 (IDC 'other accelerated' servers $72.1bn less China's share, as in the approved card); range US$46-75bn. Share = US$0.075bn / 62bn (range US$0-0.3bn against the pool range)."},"china":{"low":5,"central":8,"high":15,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"1H25 (IDC China, six months to 2025-06-30)","periodEnd":"2025-06-30","source":{"url":"https://www.idc.com/resource-center/press-releases/idc%ef%bc%9aai%e6%9c%8d%e5%8a%a1%e5%99%a8%e5%90%8c%e6%af%94%e5%a2%9e%e9%95%bf%e8%b6%85%e4%b8%80%e5%80%8d%ef%bc%8c%e5%b8%82%e5%9c%ba%e7%8e%af%e5%a2%83%e6%b7%b1%e5%8c%96%e6%9c%ac%e5%9c%9f%e5%8c%96/","title":"IDC：AI服务器同比增长超一倍，市场环境深化本土化 (IDC: AI servers more than double year on year, 1H25)","publisher":"IDC","publishedDate":"2025-10-08","accessedDate":"2026-10-22","tier":"T5","locator":"China accelerated servers US$16.0bn in 1H25 (more than double); by sales Inspur, H3C and Lenovo top three, nearly 50%; by shipments Inspur, H3C and Nettrix, about 43%"},"note":"IDC China ranks Nettrix second by accelerated-server sales in 2024 (with Inspur and H3C over 50%) and among the top three by shipments in 1H25 (Inspur, H3C and Nettrix, about 43%), but not among the top three by sales in 1H25 (Inspur, H3C and Lenovo), so in 1H25 its sales share was below the third seat (about 12-13%) and below Lenovo's. Taken at 8% (5-15%), down from 14% in 2024 and from 10% in the first version of this file, and set below Lenovo's 10.5% for that reason. Tier T5 (allocation by 5LayerCake). IDC's 1H25 data is more than 12 months old, so the engine widens the range. Pool: AI servers sold in mainland China, about US$33.6bn over 3Q25-2Q26 (9% of IDC's $373.4bn accelerated servers, as in the approved card); range US$26-45bn. Share = 8% central of the pool."}},"screening":{"gpu-systems":{"levelLow":1,"levelHigh":2},"custom-accelerator-systems":{"levelLow":1,"levelHigh":2},"china":{"levelLow":1,"levelHigh":3}},"checks":{"sourceSupport":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-22","passed":true,"sha256":"56d0b19cff1f01db34502658bbd022913d11fea4c19616937454442933cd3b2c"},"consistency":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-22","passed":true,"sha256":"56d0b19cff1f01db34502658bbd022913d11fea4c19616937454442933cd3b2c"}}},"zte":{"schema":"5lc-index-inputs/1","key":"infrastructure-servers-systems","entityId":"zte","name":"ZTE (servers)","kind":"public","commercialActivityVerified":true,"shares":{"gpu-systems":{"low":0.032,"central":0.145,"high":0.421,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"LTM to 2026-06-30","periodEnd":"2026-06-30","source":{"url":"https://www1.hkexnews.hk/listedco/listconews/sehk/2026/0821/2026082100843.pdf","title":"ZTE Corporation Interim Report 2026 (results announcement, 21 Aug 2026)","publisher":"ZTE Corporation","publishedDate":"2026-08-21","accessedDate":"2026-10-22","tier":"T5","locator":"Operating revenue by business: Government and corporate business RMB 27,631,306 thousand (+43.51%), 'reflecting mainly increase in revenue from server and storage'; carriers' network RMB 31,765,129 thousand"},"note":"ZTE does not report AI server revenue; the closest segment is Government and corporate business. Anchors (ZTE filings): that segment was RMB 27.631bn in H1 2026, +43.51% (the interim report cited here), and RMB 37.222bn in 2025, +100.49%, 'attributable mainly to the growth in revenue from server and storage' (2025 results announcement, 6 Mar 2026, https://www1.hkexnews.hk/listedco/listconews/sehk/2026/0306/2026030601381.pdf). H1 2025 = 27.631 / 1.4351 = RMB 19.25bn, so the four quarters to June 2026 = 37.222 - 19.25 + 27.631 = RMB 45.6bn. Server and storage is taken as 75% of the segment (65-85%): ZTE said H1 2025 server and storage revenue rose 200% while the segment rose 109.9%, which with the other lines growing 0-30% puts server and storage at about 67-79% of the segment. AI servers are 55% of server and storage (45-65%), the company's H1 2025 figure ('AI服务器营收占比55%', C114, 28 Aug 2025, https://m.c114.com.cn/w127-1295995.html; the base is most probably server and storage revenue). AI servers = RMB 18.8bn (RMB 13.3-25.2bn) at FRED's LTM mean of CNY 6.993 per US$ (https://fred.stlouisfed.org/data/EXCHUS) = US$2.69bn (US$1.91-3.60bn). Location: the segment's growth comes from leading domestic internet companies, banks and power companies, and the overseas business expansion of Chinese enterprises (2025 results; the 2026 interim report says ZTE 'deepened its involvement in the overseas business expansion of Chinese enterprises'), while ZTE's international revenue is 35.7% of the group (mostly carriers), so 85% (70-95%) is taken as sold in mainland China and 15% (5-30%) outside. IDC China's releases do not name ZTE among the top three. Outside China = US$0.40bn (US$0.10-1.08bn), all taken as GPU systems. Pool: GPU servers and racks outside mainland China, US$277.8bn over 3Q25-2Q26 (IDC figures as used in the approved card: GPU $301.3bn of $373.4bn accelerated servers, less China at 9% and 70% GPU); range US$257-294bn. Share = US$0.40bn / 277.8bn (range US$0.10-1.08bn against the pool range)."},"custom-accelerator-systems":{"low":0,"central":0.048,"high":0.435,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"LTM to 2026-06-30","periodEnd":"2026-06-30","source":{"url":"https://www1.hkexnews.hk/listedco/listconews/sehk/2026/0821/2026082100843.pdf","title":"ZTE Corporation Interim Report 2026 (results announcement, 21 Aug 2026)","publisher":"ZTE Corporation","publishedDate":"2026-08-21","accessedDate":"2026-10-22","tier":"T5","locator":"Operating revenue by business: Government and corporate business RMB 27,631,306 thousand (+43.51%), 'reflecting mainly increase in revenue from server and storage'; carriers' network RMB 31,765,129 thousand"},"note":"T5 allowance, not a documented sale: ZTE's SuperPOD supports GPU, CPU and DPU chips, and nothing in the sources opened shows servers built around a cloud company's own accelerator outside China. Upper end set at US$0.2bn; central at US$0.03bn. Pool: servers for non-GPU (other) accelerators outside mainland China, US$62.0bn over 3Q25-2Q26 (IDC 'other accelerated' servers $72.1bn less China's share, as in the approved card); range US$46-75bn. Share = US$0.03bn / 62bn (range US$0-0.2bn against the pool range)."},"china":{"low":2.967,"central":6.804,"high":13.163,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"LTM to 2026-06-30","periodEnd":"2026-06-30","source":{"url":"https://www1.hkexnews.hk/listedco/listconews/sehk/2026/0821/2026082100843.pdf","title":"ZTE Corporation Interim Report 2026 (results announcement, 21 Aug 2026)","publisher":"ZTE Corporation","publishedDate":"2026-08-21","accessedDate":"2026-10-22","tier":"T5","locator":"Operating revenue by business: Government and corporate business RMB 27,631,306 thousand (+43.51%), 'reflecting mainly increase in revenue from server and storage'; carriers' network RMB 31,765,129 thousand"},"note":"ZTE does not report AI server revenue; the closest segment is Government and corporate business. Anchors (ZTE filings): that segment was RMB 27.631bn in H1 2026, +43.51% (the interim report cited here), and RMB 37.222bn in 2025, +100.49%, 'attributable mainly to the growth in revenue from server and storage' (2025 results announcement, 6 Mar 2026, https://www1.hkexnews.hk/listedco/listconews/sehk/2026/0306/2026030601381.pdf). H1 2025 = 27.631 / 1.4351 = RMB 19.25bn, so the four quarters to June 2026 = 37.222 - 19.25 + 27.631 = RMB 45.6bn. Server and storage is taken as 75% of the segment (65-85%): ZTE said H1 2025 server and storage revenue rose 200% while the segment rose 109.9%, which with the other lines growing 0-30% puts server and storage at about 67-79% of the segment. AI servers are 55% of server and storage (45-65%), the company's H1 2025 figure ('AI服务器营收占比55%', C114, 28 Aug 2025, https://m.c114.com.cn/w127-1295995.html; the base is most probably server and storage revenue). AI servers = RMB 18.8bn (RMB 13.3-25.2bn) at FRED's LTM mean of CNY 6.993 per US$ (https://fred.stlouisfed.org/data/EXCHUS) = US$2.69bn (US$1.91-3.60bn). Location: the segment's growth comes from leading domestic internet companies, banks and power companies, and the overseas business expansion of Chinese enterprises (2025 results; the 2026 interim report says ZTE 'deepened its involvement in the overseas business expansion of Chinese enterprises'), while ZTE's international revenue is 35.7% of the group (mostly carriers), so 85% (70-95%) is taken as sold in mainland China and 15% (5-30%) outside. IDC China's releases do not name ZTE among the top three. China AI servers = US$2.29bn (US$1.34-3.42bn). Pool: AI servers sold in mainland China, about US$33.6bn over 3Q25-2Q26 (9% of IDC's $373.4bn accelerated servers, as in the approved card); range US$26-45bn. Share = US$2.29bn / 33.6bn (range US$1.34-3.42bn against the pool range)."}},"screening":{"gpu-systems":{"levelLow":1,"levelHigh":2},"custom-accelerator-systems":{"levelLow":1,"levelHigh":2},"china":{"levelLow":1,"levelHigh":3}},"checks":{"sourceSupport":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-22","passed":true,"sha256":"a134d12687f831342840aade6813e4c22d1dda354da627425928c166399cc892"},"consistency":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-22","passed":true,"sha256":"a134d12687f831342840aade6813e4c22d1dda354da627425928c166399cc892"}}},"jabil":{"schema":"5lc-index-inputs/1","key":"infrastructure-servers-systems","entityId":"jabil","name":"Jabil","kind":"public","commercialActivityVerified":true,"shares":{"gpu-systems":{"low":0.274,"central":0.829,"high":2.129,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"FY2026 (year to 2026-08-31)","periodEnd":"2026-08-31","source":{"url":"https://www.webull.com/news/15667757115769856","title":"Jabil Reports Q4 2026 Results: Full Earnings Call Transcript (call of 30 Sep 2026)","publisher":"Jabil (transcript published by Webull)","publishedDate":"2026-09-30","accessedDate":"2026-10-22","tier":"T5","locator":"CFO: 'In fiscal 2026, the team delivered approximately $14.4 billion of AI-related revenue, up $5.4 billion year over year'; FY27 AI-related revenue about $22.1bn, up 54%; cloud and data center infrastructure about $17.5bn, up 52%"},"note":"Jabil does not report AI server or rack revenue. Anchors (company statements on the Q4 FY26 call, 30 Sep 2026, transcript on the page cited here; Jabil's own slide deck could not be opened): AI-related revenue in FY2026 (year to 31 Aug 2026) was about US$14.4bn, up US$5.4bn (so about US$9.0bn in FY2025); FY2027 cloud and data center infrastructure revenue is guided to about US$17.5bn, up about 52%, which implies about US$11.5bn in FY2026 (17.5 / 1.52); the segment head said the largest account 'developed from a server and rack program into servers, racks, power, cooling and services'. Group revenue was US$36.0bn (press release, https://s27.q4cdn.com/276975351/files/doc_news/Jabil-Posts-Fourth-Quarter-and-Fiscal-Year-2026-Results-2026.pdf). AI-related revenue also includes power, cooling, photonics, networking racks and components that belong to other sublayers, so only part is complete AI servers and racks sold to data-centre operators. Judgment (T5): 20% (8-40%) of AI-related revenue is complete AI server and rack systems = US$2.9bn (US$1.2-5.8bn); the central is roughly the geometric mean of the ends. The customers are unnamed hyperscalers; 80% of the credited amount (70-95%) is taken as GPU systems = US$2.3bn (US$0.8-5.5bn), in line with the card's own mix of GPU and own-chip systems outside China (74:17). Pool: GPU servers and racks outside mainland China, US$277.8bn over 3Q25-2Q26 (IDC figures as used in the approved card: GPU $301.3bn of $373.4bn accelerated servers, less China at 9% and 70% GPU); range US$257-294bn. Share = US$2.3bn / 277.8bn (range US$0.8-5.5bn against the pool range)."},"custom-accelerator-systems":{"low":0.077,"central":0.929,"high":3.757,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"FY2026 (year to 2026-08-31)","periodEnd":"2026-08-31","source":{"url":"https://www.webull.com/news/15667757115769856","title":"Jabil Reports Q4 2026 Results: Full Earnings Call Transcript (call of 30 Sep 2026)","publisher":"Jabil (transcript published by Webull)","publishedDate":"2026-09-30","accessedDate":"2026-10-22","tier":"T5","locator":"CFO: 'In fiscal 2026, the team delivered approximately $14.4 billion of AI-related revenue, up $5.4 billion year over year'; FY27 AI-related revenue about $22.1bn, up 54%; cloud and data center infrastructure about $17.5bn, up 52%"},"note":"Jabil does not report AI server or rack revenue. Anchors (company statements on the Q4 FY26 call, 30 Sep 2026, transcript on the page cited here; Jabil's own slide deck could not be opened): AI-related revenue in FY2026 (year to 31 Aug 2026) was about US$14.4bn, up US$5.4bn (so about US$9.0bn in FY2025); FY2027 cloud and data center infrastructure revenue is guided to about US$17.5bn, up about 52%, which implies about US$11.5bn in FY2026 (17.5 / 1.52); the segment head said the largest account 'developed from a server and rack program into servers, racks, power, cooling and services'. Group revenue was US$36.0bn (press release, https://s27.q4cdn.com/276975351/files/doc_news/Jabil-Posts-Fourth-Quarter-and-Fiscal-Year-2026-Results-2026.pdf). AI-related revenue also includes power, cooling, photonics, networking racks and components that belong to other sublayers, so only part is complete AI servers and racks sold to data-centre operators. Judgment (T5): 20% (8-40%) of AI-related revenue is complete AI server and rack systems = US$2.9bn (US$1.2-5.8bn); the central is roughly the geometric mean of the ends. 20% of the credited amount (5-30%) is taken as systems around cloud companies' own accelerators = US$0.58bn (US$0.06-1.73bn), by the card's 74:17 mix outside China, shaded down because no ASIC program is named. Pool: servers for non-GPU (other) accelerators outside mainland China, US$62.0bn over 3Q25-2Q26 (IDC 'other accelerated' servers $72.1bn less China's share, as in the approved card); range US$46-75bn. Share = US$0.58bn / 62bn (range US$0.06-1.73bn against the pool range)."},"china":{"low":0,"central":0.089,"high":0.385,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"FY2026 (year to 2026-08-31)","periodEnd":"2026-08-31","source":{"url":"https://www.webull.com/news/15667757115769856","title":"Jabil Reports Q4 2026 Results: Full Earnings Call Transcript (call of 30 Sep 2026)","publisher":"Jabil (transcript published by Webull)","publishedDate":"2026-09-30","accessedDate":"2026-10-22","tier":"T5","locator":"CFO: 'In fiscal 2026, the team delivered approximately $14.4 billion of AI-related revenue, up $5.4 billion year over year'; FY27 AI-related revenue about $22.1bn, up 54%; cloud and data center infrastructure about $17.5bn, up 52%"},"note":"T5 allowance, not a documented sale: the customers described are hyperscale accounts served from Mexico and the US, and nothing in the sources opened shows AI server sales inside mainland China. Upper end set at US$0.1bn; central at US$0.03bn. Pool: AI servers sold in mainland China, about US$33.6bn over 3Q25-2Q26 (9% of IDC's $373.4bn accelerated servers, as in the approved card); range US$26-45bn. Share = US$0.03bn / 33.6bn (range US$0-0.1bn against the pool range)."}},"screening":{"gpu-systems":{"levelLow":1,"levelHigh":2},"custom-accelerator-systems":{"levelLow":1,"levelHigh":2},"china":{"levelLow":1,"levelHigh":2}},"checks":{"sourceSupport":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-22","passed":true,"sha256":"7b172267be9679c76d199f959c4e1d1eddb7701c625cfa8d60086192eb47869f"},"consistency":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-22","passed":true,"sha256":"7b172267be9679c76d199f959c4e1d1eddb7701c625cfa8d60086192eb47869f"}}},"sanmina":{"schema":"5lc-index-inputs/1","key":"infrastructure-servers-systems","entityId":"sanmina","name":"Sanmina (incl. ZT Systems manufacturing)","kind":"public","commercialActivityVerified":true,"shares":{"gpu-systems":{"low":0.195,"central":0.734,"high":1.436,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"LTM to 2026-06-27 (8 months of ZT Systems)","periodEnd":"2026-06-27","source":{"url":"https://s201.q4cdn.com/209924174/files/doc_financials/2026/q3/SANM-Q326-Earnings-Presentation-7_27_26-FINAL.pdf","title":"Sanmina Q3 Fiscal 2026 Financial Results (earnings presentation, 27 Jul 2026)","publisher":"Sanmina Corporation","publishedDate":"2026-07-27","accessedDate":"2026-10-22","tier":"T5","locator":"Slides 6 and 8: 'ZT Systems revenue of $1.1 billion' in Q3 FY26; slide 13: ZT Systems revenue of $4.8-5.0bn for 11 months of FY26, within a $5-6bn annualized run rate; core Sanmina $9.1-9.2bn"},"note":"Sanmina sells rack-scale AI systems through ZT Systems (bought from AMD, closed 27 Oct 2025); core Sanmina's communications, cloud and AI work (about US$1.05bn in Q3 FY26: end-market revenue of US$2.148bn less ZT's US$1.1bn) is taken as contract manufacturing for other brands and is not credited. ZT revenue since closing, from Sanmina's filings and presentation: US$1.1bn for 27 Oct-27 Dec 2025 (Form 10-Q for the quarter ended 27 Dec 2025, Note 13, https://www.sec.gov/Archives/edgar/data/897723/000089772326000015/sanm-20251227.htm); US$1.9bn in Q2 and US$3.0bn for the six months to 28 Mar 2026 (10-Q for the quarter ended 28 Mar 2026, Note 13, https://www.sec.gov/Archives/edgar/data/0000897723/000089772326000025/sanm-20260328.htm); US$1.1bn in Q3 (the Q3 FY26 earnings presentation cited here) = US$4.1bn for the 8 months to 27 Jun 2026. That is the whole LTM, because ZT was not owned before; FY2025 revenue (to 27 Sep 2025, without ZT) was US$8.13bn (8-K earnings release for the year to 27 Sep 2025, https://www.sec.gov/Archives/edgar/data/897723/000089772325000036/sanmina_exx991xsep272025.htm). Note 13 calls ZT 'a manufacturer of AI and general purpose computer infrastructure for hyperscale computing companies'; on the Q2 call the CEO said no Nvidia product shipped and accelerated-compute shipments were 'all based on AMD technology' (MarketBeat summary of the 27 Apr 2026 call, https://finance.yahoo.com/markets/stocks/articles/sanmina-q2-earnings-call-highlights-231349261.html). Judgment (T5): AI accelerated systems are 65% of ZT revenue (40-90%), the rest being general-purpose computing; 85% of that (50-100%) is sold under Sanmina's name to hyperscale customers rather than built for another brand, so credited AI systems = 4.1 x 0.65 x 0.85 = US$2.3bn (US$0.8-3.7bn); 90% of it (70-100%) is AMD GPU systems = US$2.0bn (US$0.6-3.7bn). The earlier screening figure (US$3.5bn of AI systems, assuming a FY25 base of about US$9bn) was not supported by the filings. Pool: GPU servers and racks outside mainland China, US$277.8bn over 3Q25-2Q26 (IDC figures as used in the approved card: GPU $301.3bn of $373.4bn accelerated servers, less China at 9% and 70% GPU); range US$257-294bn. Share = US$2.04bn / 277.8bn (range US$0.57-3.69bn against the pool range)."},"custom-accelerator-systems":{"low":0,"central":0.365,"high":2.407,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"LTM to 2026-06-27 (8 months of ZT Systems)","periodEnd":"2026-06-27","source":{"url":"https://s201.q4cdn.com/209924174/files/doc_financials/2026/q3/SANM-Q326-Earnings-Presentation-7_27_26-FINAL.pdf","title":"Sanmina Q3 Fiscal 2026 Financial Results (earnings presentation, 27 Jul 2026)","publisher":"Sanmina Corporation","publishedDate":"2026-07-27","accessedDate":"2026-10-22","tier":"T5","locator":"Slides 6 and 8: 'ZT Systems revenue of $1.1 billion' in Q3 FY26; slide 13: ZT Systems revenue of $4.8-5.0bn for 11 months of FY26, within a $5-6bn annualized run rate; core Sanmina $9.1-9.2bn"},"note":"Sanmina sells rack-scale AI systems through ZT Systems (bought from AMD, closed 27 Oct 2025); core Sanmina's communications, cloud and AI work (about US$1.05bn in Q3 FY26: end-market revenue of US$2.148bn less ZT's US$1.1bn) is taken as contract manufacturing for other brands and is not credited. ZT revenue since closing, from Sanmina's filings and presentation: US$1.1bn for 27 Oct-27 Dec 2025 (Form 10-Q for the quarter ended 27 Dec 2025, Note 13, https://www.sec.gov/Archives/edgar/data/897723/000089772326000015/sanm-20251227.htm); US$1.9bn in Q2 and US$3.0bn for the six months to 28 Mar 2026 (10-Q for the quarter ended 28 Mar 2026, Note 13, https://www.sec.gov/Archives/edgar/data/0000897723/000089772326000025/sanm-20260328.htm); US$1.1bn in Q3 (the Q3 FY26 earnings presentation cited here) = US$4.1bn for the 8 months to 27 Jun 2026. That is the whole LTM, because ZT was not owned before; FY2025 revenue (to 27 Sep 2025, without ZT) was US$8.13bn (8-K earnings release for the year to 27 Sep 2025, https://www.sec.gov/Archives/edgar/data/897723/000089772325000036/sanmina_exx991xsep272025.htm). Note 13 calls ZT 'a manufacturer of AI and general purpose computer infrastructure for hyperscale computing companies'; on the Q2 call the CEO said no Nvidia product shipped and accelerated-compute shipments were 'all based on AMD technology' (MarketBeat summary of the 27 Apr 2026 call, https://finance.yahoo.com/markets/stocks/articles/sanmina-q2-earnings-call-highlights-231349261.html). Judgment (T5): 10% of the credited AI systems (0-30%) are built around a cloud company's own accelerator; the CEO said Q2 accelerated shipments were all AMD (GPUs), and no ASIC program is named in the sources, so the allowance is small. Credited AI systems US$2.3bn (US$0.8-3.7bn) x 10% = US$0.23bn (0-1.1bn). Pool: servers for non-GPU (other) accelerators outside mainland China, US$62.0bn over 3Q25-2Q26 (IDC 'other accelerated' servers $72.1bn less China's share, as in the approved card); range US$46-75bn. Share = US$0.23bn / 62bn (range US$0-1.11bn against the pool range)."},"china":{"low":0,"central":0.074,"high":0.385,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"LTM to 2026-06-27","periodEnd":"2026-06-27","source":{"url":"https://s201.q4cdn.com/209924174/files/doc_financials/2026/q3/SANM-Q326-Earnings-Presentation-7_27_26-FINAL.pdf","title":"Sanmina Q3 Fiscal 2026 Financial Results (earnings presentation, 27 Jul 2026)","publisher":"Sanmina Corporation","publishedDate":"2026-07-27","accessedDate":"2026-10-22","tier":"T5","locator":"Slides 6 and 8: 'ZT Systems revenue of $1.1 billion' in Q3 FY26; slide 13: ZT Systems revenue of $4.8-5.0bn for 11 months of FY26, within a $5-6bn annualized run rate; core Sanmina $9.1-9.2bn"},"note":"T5 allowance, not a documented sale: ZT builds for hyperscale customers and AMD-based systems; nothing in the sources opened shows sales of AI servers inside mainland China, and US export rules restrict the leading accelerators there. Upper end set at US$0.1bn; central at a quarter of that. Pool: AI servers sold in mainland China, about US$33.6bn over 3Q25-2Q26 (9% of IDC's $373.4bn accelerated servers, as in the approved card); range US$26-45bn. Share = US$0.025bn / 33.6bn (range US$0-0.1bn against the pool range)."}},"screening":{"gpu-systems":{"levelLow":1,"levelHigh":2.5},"custom-accelerator-systems":{"levelLow":1,"levelHigh":3},"china":{"levelLow":1,"levelHigh":2}},"checks":{"sourceSupport":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-22","passed":true,"sha256":"7a526b62c0ef528532615ce85f7cd41c597a3012cbf756a3ceefb3f2a42e0ca3"},"consistency":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-22","passed":true,"sha256":"7a526b62c0ef528532615ce85f7cd41c597a3012cbf756a3ceefb3f2a42e0ca3"}}},"mitac":{"schema":"5lc-index-inputs/1","key":"infrastructure-servers-systems","entityId":"mitac","name":"MiTAC (TYAN)","kind":"public","commercialActivityVerified":true,"shares":{"gpu-systems":{"low":0.34,"central":0.684,"high":1.17,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"Estimate to 2026-03-31","periodEnd":"2026-03-31","source":{"url":"https://www.taipeitimes.com/News/biz/archives/2026/05/29/2003858148","title":"MiTAC Holdings sees sustainable sales momentum","publisher":"Taipei Times","publishedDate":"2026-05-29","accessedDate":"2026-10-22","tier":"T5","locator":"Quarterly revenue NT$31.86bn, 90.1% from MiTAC Computing (servers and storage); orders from hyperscalers, CSPs and neoclouds"},"note":"MiTAC Holdings quarterly revenue NT$31.86bn (most likely Q1 2026), 90.1% from MiTAC Computing (servers and storage) = NT$28.7bn (US$0.9bn) a quarter; server sales were 89.5% of revenue in H1 2025 (+145%). Growing quickly; LTM server revenue about US$3bn (range 1.5-4.5), AI about 65% = $1.9bn (1.0-3.0). Tier T5; only a May 2026 article was opened, no later quarter. Pool: GPU servers and racks outside mainland China, US$277.8bn over 3Q25-2Q26 (IDC figures as used in the approved card: GPU $301.3bn of $373.4bn accelerated servers, less China at 9% and 70% GPU); range US$257-294bn. Share = US$1.9bn / 277.8bn (range US$1-3bn against the pool range)."},"custom-accelerator-systems":{"low":0,"central":0.202,"high":1.09,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"Estimate to 2026-03-31","periodEnd":"2026-03-31","source":{"url":"https://www.taipeitimes.com/News/biz/archives/2026/05/29/2003858148","title":"MiTAC Holdings sees sustainable sales momentum","publisher":"Taipei Times","publishedDate":"2026-05-29","accessedDate":"2026-10-22","tier":"T5","locator":"Quarterly revenue NT$31.86bn, 90.1% from MiTAC Computing (servers and storage); orders from hyperscalers, CSPs and neoclouds"},"note":"Wide placeholder estimate, tier T5: no system sales built on cloud companies' own accelerators are shown in the sources opened. Upper end set at US$0.5bn; central at a quarter of that. This function is not material for the company. Pool: servers for non-GPU (other) accelerators outside mainland China, US$62.0bn over 3Q25-2Q26 (IDC 'other accelerated' servers $72.1bn less China's share, as in the approved card); range US$46-75bn. Share = US$0.125bn / 62bn (range US$0-0.5bn against the pool range)."},"china":{"low":0,"central":0.223,"high":1.15,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"Estimate to 2026-03-31","periodEnd":"2026-03-31","source":{"url":"https://www.taipeitimes.com/News/biz/archives/2026/05/29/2003858148","title":"MiTAC Holdings sees sustainable sales momentum","publisher":"Taipei Times","publishedDate":"2026-05-29","accessedDate":"2026-10-22","tier":"T5","locator":"Quarterly revenue NT$31.86bn, 90.1% from MiTAC Computing (servers and storage); orders from hyperscalers, CSPs and neoclouds"},"note":"Wide placeholder estimate, tier T5: no sale of AI servers inside mainland China is shown in the sources opened. Upper end set at US$0.3bn; central at a quarter of that. This function is not material for the company. Pool: AI servers sold in mainland China, about US$33.6bn over 3Q25-2Q26 (9% of IDC's $373.4bn accelerated servers, as in the approved card); range US$26-45bn. Share = US$0.075bn / 33.6bn (range US$0-0.3bn against the pool range)."}},"screening":{"gpu-systems":{"levelLow":1,"levelHigh":2.5},"custom-accelerator-systems":{"levelLow":1,"levelHigh":2},"china":{"levelLow":1,"levelHigh":2}},"checks":{"sourceSupport":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-22","passed":true,"sha256":"7c521f6e6ce9ac8209f657aec4b060a531f905807165a9d10be497b14fe0b6f5"},"consistency":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-22","passed":true,"sha256":"7c521f6e6ce9ac8209f657aec4b060a531f905807165a9d10be497b14fe0b6f5"}}},"huakun-zhenyu":{"schema":"5lc-index-inputs/1","key":"infrastructure-servers-systems","entityId":"huakun-zhenyu","name":"Huakun Zhenyu","kind":"private","commercialActivityVerified":true,"shares":{"gpu-systems":{"low":0,"central":0.009,"high":0.039,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"LTM to Jun 2026 (IDC 3Q25-2Q26)","periodEnd":"2026-06-30","source":{"url":"https://apps.digitimes.com/news/a20260707RS401.html?chid=2","title":"DIGITIMES Research: Huakun leads China's non-x86 server segment (13.7% share in 2025)","publisher":"DIGITIMES","publishedDate":"2026-07-13","accessedDate":"2026-10-22","tier":"T5","locator":"Visible text: Huakun 13.7% of China's non-x86 servers in 2025; rest paywalled"},"note":"Wide placeholder estimate, tier T5: sales outside China are not documented in the sources opened. Upper end set at US$0.1bn; central at a quarter of that. This function is not material for the company. Pool: GPU servers and racks outside mainland China, US$277.8bn over 3Q25-2Q26 (IDC figures as used in the approved card: GPU $301.3bn of $373.4bn accelerated servers, less China at 9% and 70% GPU); range US$257-294bn. Share = US$0.025bn / 277.8bn (range US$0-0.1bn against the pool range)."},"custom-accelerator-systems":{"low":0,"central":0.081,"high":0.438,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"LTM to Jun 2026 (IDC 3Q25-2Q26)","periodEnd":"2026-06-30","source":{"url":"https://apps.digitimes.com/news/a20260707RS401.html?chid=2","title":"DIGITIMES Research: Huakun leads China's non-x86 server segment (13.7% share in 2025)","publisher":"DIGITIMES","publishedDate":"2026-07-13","accessedDate":"2026-10-22","tier":"T5","locator":"Visible text: Huakun 13.7% of China's non-x86 servers in 2025; rest paywalled"},"note":"Wide placeholder estimate, tier T5: no system sales built on cloud companies' own accelerators are shown in the sources opened. Upper end set at US$0.2bn; central at a quarter of that. This function is not material for the company. Pool: servers for non-GPU (other) accelerators outside mainland China, US$62.0bn over 3Q25-2Q26 (IDC 'other accelerated' servers $72.1bn less China's share, as in the approved card); range US$46-75bn. Share = US$0.05bn / 62bn (range US$0-0.2bn against the pool range)."},"china":{"low":1.5,"central":4,"high":8,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"Estimate to 2026-06-30 from 2024 evidence","periodEnd":"2025-12-31","source":{"url":"https://apps.digitimes.com/news/a20260707RS401.html?chid=2","title":"DIGITIMES Research: Huakun leads China's non-x86 server segment (13.7% share in 2025)","publisher":"DIGITIMES","publishedDate":"2026-07-13","accessedDate":"2026-10-22","tier":"T5","locator":"Visible text: Huakun 13.7% of China's non-x86 servers in 2025; rest paywalled"},"note":"DIGITIMES Research says Huakun led China's non-x86 server segment in 2025 with 13.7%; that segment includes non-AI servers and is a fraction of all accelerated servers, so its share of the accelerated pool is taken as 4% (1.5-8), tier T5. Market size and other vendors' shares are behind the paywall. Pool: AI servers sold in mainland China, about US$33.6bn over 3Q25-2Q26 (9% of IDC's $373.4bn accelerated servers, as in the approved card); range US$26-45bn."}},"screening":{"gpu-systems":{"levelLow":1,"levelHigh":2},"custom-accelerator-systems":{"levelLow":1,"levelHigh":2},"china":{"levelLow":1,"levelHigh":3}},"checks":{"sourceSupport":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-22","passed":true,"sha256":"8c12e315d157ffa88313c6c42e839d1c23569e34bf04644b3ff8b028fae48ffb"},"consistency":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-22","passed":true,"sha256":"8c12e315d157ffa88313c6c42e839d1c23569e34bf04644b3ff8b028fae48ffb"}}},"flex":{"schema":"5lc-index-inputs/1","key":"infrastructure-servers-systems","entityId":"flex","name":"Flex","kind":"public","commercialActivityVerified":true,"shares":{"gpu-systems":{"low":0.106,"central":0.453,"high":1.183,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"FY2026 (year to 2026-03-31)","periodEnd":"2026-03-31","source":{"url":"https://www.sec.gov/Archives/edgar/data/0000866374/000086637426000012/flex-20260331.htm","title":"Flex Ltd. Form 10-K for fiscal year ended March 31, 2026","publisher":"Flex Ltd.","publishedDate":"2026-05","accessedDate":"2026-10-22","tier":"T5","locator":"Filing shows only the month (auditor's report dated 20 May 2026). Item 7: Cloud and Power Infrastructure (CPI) net sales $6,614m in FY2026 (24% of total), $4,799m in FY2025, up 38% on a 29% rise in Cloud and Cooling and 61% in Power; Item 1: Cloud and Cooling is 'integrated compute systems' plus liquid cooling"},"note":"Flex does not report AI server revenue. Anchors (Form 10-K for the year to 31 Mar 2026, source cited here): CPI net sales were US$6,614m (+38% from US$4,799m), 'primarily related to increased demand in the data center market', made up of Cloud and Cooling (+29%) and Power (+61%); Item 1 describes Cloud and Cooling as 'integrated compute systems' plus liquid cooling, with 'compute and rack-level infrastructure' among its products. Solving 0.29 x C + 0.61 x P = 1.815 with C + P = 4.799 gives FY2025 Cloud and Cooling of about US$3.5bn and Power US$1.3bn, so FY2026 Cloud and Cooling is about US$4.5bn (4.0-5.0, allowing for rounded growth rates) and Power about US$2.1bn. The year to March 2026 trails the LTM to June 2026 by one quarter of strong growth (Q1 FY2027 net sales were US$7.9bn, up 21%, press release of 29 Jul 2026, https://s202.q4cdn.com/732614612/files/doc_earnings/2027/q1/earnings-result/FLEX_PR_27Q1.pdf), so the share is probably a little low. Judgment (T5): compute systems are 60% of Cloud and Cooling (40-80%), the rest being liquid cooling; AI is 65% of compute systems (50-80%), the rest general-purpose; 90% (60-100%) is sold under Flex's name to data-centre operators rather than assembled for another brand: 4.48 x 0.60 x 0.65 x 0.90 = US$1.57bn credited (US$0.48-3.20bn). 80% of it (65-95%) is GPU systems = US$1.26bn. Pool: GPU servers and racks outside mainland China, US$277.8bn over 3Q25-2Q26 (IDC figures as used in the approved card: GPU $301.3bn of $373.4bn accelerated servers, less China at 9% and 70% GPU); range US$257-294bn. Share = US$1.26bn / 277.8bn (range US$0.31-3.04bn against the pool range)."},"custom-accelerator-systems":{"low":0.032,"central":0.507,"high":2.435,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"FY2026 (year to 2026-03-31)","periodEnd":"2026-03-31","source":{"url":"https://www.sec.gov/Archives/edgar/data/0000866374/000086637426000012/flex-20260331.htm","title":"Flex Ltd. Form 10-K for fiscal year ended March 31, 2026","publisher":"Flex Ltd.","publishedDate":"2026-05","accessedDate":"2026-10-22","tier":"T5","locator":"Filing shows only the month (auditor's report dated 20 May 2026). Item 7: Cloud and Power Infrastructure (CPI) net sales $6,614m in FY2026 (24% of total), $4,799m in FY2025, up 38% on a 29% rise in Cloud and Cooling and 61% in Power; Item 1: Cloud and Cooling is 'integrated compute systems' plus liquid cooling"},"note":"Flex does not report AI server revenue. Anchors (Form 10-K for the year to 31 Mar 2026, source cited here): CPI net sales were US$6,614m (+38% from US$4,799m), 'primarily related to increased demand in the data center market', made up of Cloud and Cooling (+29%) and Power (+61%); Item 1 describes Cloud and Cooling as 'integrated compute systems' plus liquid cooling, with 'compute and rack-level infrastructure' among its products. Solving 0.29 x C + 0.61 x P = 1.815 with C + P = 4.799 gives FY2025 Cloud and Cooling of about US$3.5bn and Power US$1.3bn, so FY2026 Cloud and Cooling is about US$4.5bn (4.0-5.0, allowing for rounded growth rates) and Power about US$2.1bn. The year to March 2026 trails the LTM to June 2026 by one quarter of strong growth (Q1 FY2027 net sales were US$7.9bn, up 21%, press release of 29 Jul 2026, https://s202.q4cdn.com/732614612/files/doc_earnings/2027/q1/earnings-result/FLEX_PR_27Q1.pdf), so the share is probably a little low. Judgment (T5): compute systems are 60% of Cloud and Cooling (40-80%), the rest being liquid cooling; AI is 65% of compute systems (50-80%), the rest general-purpose; 90% (60-100%) is sold under Flex's name to data-centre operators rather than assembled for another brand: 4.48 x 0.60 x 0.65 x 0.90 = US$1.57bn credited (US$0.48-3.20bn). 20% of it (5-35%) is taken as systems around cloud companies' own accelerators = US$0.31bn, by the card's 74:17 mix outside China shaded down because no ASIC program is named. Pool: servers for non-GPU (other) accelerators outside mainland China, US$62.0bn over 3Q25-2Q26 (IDC 'other accelerated' servers $72.1bn less China's share, as in the approved card); range US$46-75bn. Share = US$0.31bn / 62bn (range US$0.02-1.12bn against the pool range)."},"china":{"low":0,"central":0.089,"high":0.385,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"FY2026 (year to 2026-03-31)","periodEnd":"2026-03-31","source":{"url":"https://www.sec.gov/Archives/edgar/data/0000866374/000086637426000012/flex-20260331.htm","title":"Flex Ltd. Form 10-K for fiscal year ended March 31, 2026","publisher":"Flex Ltd.","publishedDate":"2026-05","accessedDate":"2026-10-22","tier":"T5","locator":"Filing shows only the month (auditor's report dated 20 May 2026). Item 7: Cloud and Power Infrastructure (CPI) net sales $6,614m in FY2026 (24% of total), $4,799m in FY2025, up 38% on a 29% rise in Cloud and Cooling and 61% in Power; Item 1: Cloud and Cooling is 'integrated compute systems' plus liquid cooling"},"note":"T5 allowance, not a documented sale: Flex's data-centre customers are described as hyperscale and cloud companies (no customer above 10% of sales; ten largest 45%), and nothing in the sources opened shows AI server sales inside mainland China. Upper end set at US$0.1bn; central at US$0.03bn. Pool: AI servers sold in mainland China, about US$33.6bn over 3Q25-2Q26 (9% of IDC's $373.4bn accelerated servers, as in the approved card); range US$26-45bn. Share = US$0.03bn / 33.6bn (range US$0-0.1bn against the pool range)."}},"screening":{"gpu-systems":{"levelLow":1,"levelHigh":2},"custom-accelerator-systems":{"levelLow":1,"levelHigh":2},"china":{"levelLow":1,"levelHigh":2}},"checks":{"sourceSupport":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-22","passed":true,"sha256":"73622bbfb6c4425d7dc6735142b3012ef14ce54f58f5efa03ca1a28f72fe9106"},"consistency":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-22","passed":true,"sha256":"73622bbfb6c4425d7dc6735142b3012ef14ce54f58f5efa03ca1a28f72fe9106"}}},"sugon":{"schema":"5lc-index-inputs/1","key":"infrastructure-servers-systems","entityId":"sugon","name":"Sugon","kind":"public","commercialActivityVerified":true,"shares":{"gpu-systems":{"low":0,"central":0.007,"high":0.012,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"LTM to 2026-06-30","periodEnd":"2026-06-30","source":{"url":"https://stock.10jqka.com.cn/20260826/c679314335.shtml","title":"Sugon (中科曙光) 2026 interim results, 21世纪经济报道 via 同花顺, 26 Aug 2026","publisher":"同花顺 (10jqka) citing 21世纪经济报道","publishedDate":"2026-08-26","accessedDate":"2026-10-22","tier":"T5","locator":"H1 2026 revenue RMB 74.66亿 (+27.62%); IT equipment RMB 66.66亿, 89.3% of revenue; no split between servers, AI servers, storage and super-nodes"},"note":"T5 allowance, not a documented sale: 2% of Sugon's AI server sales (US$0.019bn central, up to US$0.03bn) is allowed for sales outside mainland China; the sources opened show none. Pool: GPU servers and racks outside mainland China, US$277.8bn over 3Q25-2Q26 (IDC figures as used in the approved card: GPU $301.3bn of $373.4bn accelerated servers, less China at 9% and 70% GPU); range US$257-294bn. Share = US$0.019bn / 277.8bn."},"custom-accelerator-systems":{"low":0,"central":0.032,"high":0.217,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"LTM to 2026-06-30","periodEnd":"2026-06-30","source":{"url":"https://stock.10jqka.com.cn/20260826/c679314335.shtml","title":"Sugon (中科曙光) 2026 interim results, 21世纪经济报道 via 同花顺, 26 Aug 2026","publisher":"同花顺 (10jqka) citing 21世纪经济报道","publishedDate":"2026-08-26","accessedDate":"2026-10-22","tier":"T5","locator":"H1 2026 revenue RMB 74.66亿 (+27.62%); IT equipment RMB 66.66亿, 89.3% of revenue; no split between servers, AI servers, storage and super-nodes"},"note":"T5 allowance, not a documented sale: nothing in the sources opened shows Sugon building servers around a cloud company's own accelerator outside China. Upper end set at US$0.1bn; central at US$0.02bn. Pool: servers for non-GPU (other) accelerators outside mainland China, US$62.0bn over 3Q25-2Q26 (IDC 'other accelerated' servers $72.1bn less China's share, as in the approved card); range US$46-75bn. Share = US$0.02bn / 62bn (range US$0-0.1bn against the pool range)."},"china":{"low":1.153,"central":2.779,"high":5.586,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"LTM to 2026-06-30","periodEnd":"2026-06-30","source":{"url":"https://stock.10jqka.com.cn/20260826/c679314335.shtml","title":"Sugon (中科曙光) 2026 interim results, 21世纪经济报道 via 同花顺, 26 Aug 2026","publisher":"同花顺 (10jqka) citing 21世纪经济报道","publishedDate":"2026-08-26","accessedDate":"2026-10-22","tier":"T5","locator":"H1 2026 revenue RMB 74.66亿 (+27.62%); IT equipment RMB 66.66亿, 89.3% of revenue; no split between servers, AI servers, storage and super-nodes"},"note":"Sugon (中科曙光, 603019) does not report AI server revenue. Anchors: H1 2026 revenue RMB 7.466bn, +27.62%, of which IT equipment RMB 6.666bn (89.3%) (the page cited here, reporting the 25 Aug 2026 interim report); 2025 revenue RMB 14.964bn, +13.81% (Hualong Securities note on the Q3 2026 flash results, 19 Oct 2026, https://pdf.dfcfw.com/pdf/H3_AP202610201830710520_1.pdf, from Wind). H1 2025 = 7.466 / 1.2762 = RMB 5.85bn, so the four quarters to June 2026 = 14.964 - 5.85 + 7.466 = RMB 16.58bn, of which IT equipment at the H1 2026 share of 89.3% = RMB 14.81bn. IT equipment mixes AI servers, general-purpose servers, storage, supercomputers and super-nodes (scaleX). Judgment (T5): AI-accelerated systems are 45% of IT equipment (25-70%), between ZTE's 55% of server and storage (company, via C114) and a general-purpose-heavy mix, since the interim report capitalises an 'AI compute cluster project' and launches the scaleX40 super-node. AI servers = RMB 6.66bn (RMB 3.70-10.36bn) at FRED's LTM mean of CNY 6.993 per US$ (https://fred.stlouisfed.org/data/EXCHUS) = US$0.95bn (US$0.53-1.48bn). IDC China's releases never name Sugon among the top three by sales or shipments. Sales are taken as 98% in mainland China (the sources show no overseas AI server sales). Pool: AI servers sold in mainland China, about US$33.6bn over 3Q25-2Q26 (9% of IDC's $373.4bn accelerated servers, as in the approved card); range US$26-45bn. Share = US$0.93bn / 33.6bn (range US$0.52-1.45bn against the pool range)."}},"screening":{"gpu-systems":{"levelLow":1,"levelHigh":2},"custom-accelerator-systems":{"levelLow":1,"levelHigh":2},"china":{"levelLow":1,"levelHigh":3}},"checks":{"sourceSupport":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-22","passed":true,"sha256":"00ca81f00f6b8daaf714274c6959e19c4f463853b4340f0bcc7dd3aa5a1fe9b0"},"consistency":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-22","passed":true,"sha256":"00ca81f00f6b8daaf714274c6959e19c4f463853b4340f0bcc7dd3aa5a1fe9b0"}}},"pegatron":{"schema":"5lc-index-inputs/1","key":"infrastructure-servers-systems","entityId":"pegatron","name":"Pegatron","kind":"public","commercialActivityVerified":true,"shares":{"gpu-systems":{"low":0.139,"central":0.358,"high":0.694,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"LTM to 2026-06-30 (3Q25-2Q26)","periodEnd":"2026-06-30","source":{"url":"https://telecomlead.com/electronics/pegatron-q2-2026-revenue-rises-2-7-as-ai-server-boom-drives-84-computing-growth-127327","title":"Pegatron Q2 2026 Revenue Rises 2.7% as AI Server Boom Drives 84% Computing Growth (TelecomLead, 17 Aug 2026)","publisher":"TelecomLead","publishedDate":"2026-08-17","accessedDate":"2026-10-22","tier":"T5","locator":"Q2 2026 revenue NT$274.448bn; Computing segment +84% y/y and +82% q/q 'primarily due to higher Server shipments', 23% of revenue (14% in Q1 2026, 13% in Q2 2025); server revenue targeted at about NT$100bn in 2026"},"note":"Pegatron does not report server revenue by quarter; this is a T5 build from opened statements. Anchors: (1) Pegatron said on 11 Mar 2026 that it aims to grow AI server revenue more than 10-fold in 2026, driven by Nvidia GB300 orders from neocloud and large cloud customers, and that shipments will rise each quarter with the second half stronger (Taipei Times, 12 Mar 2026, https://www.taipeitimes.com/News/biz/archives/2026/03/12/2003853649); the server revenue target is about NT$100bn for 2026 (TelecomLead, the page cited here), which puts 2025 AI server revenue near NT$10bn; (2) the Computing segment was 23% of Q2 2026 revenue of NT$274.4bn = NT$63.1bn against 13% of NT$267.2bn (= 274.448 / 1.027) = NT$34.7bn in Q2 2025, so about NT$28bn of the increase is servers; Q1 2026 Computing was 14% of NT$244.1bn = NT$34.2bn. Server revenue taken as NT$2.5bn in Q3 2025 (1.5-4), NT$4.5bn in Q4 2025 (3-7), NT$5bn in Q1 2026 (2-9) and NT$31bn in Q2 2026 (25-36) = NT$43bn (31.5-56), converted at FRED's quarterly means of NT$ per US$ (https://fred.stlouisfed.org/data/EXTAUS: 29.95, 31.03, 31.62, 31.60) = US$1.37bn (US$1.00-1.78bn). AI servers are 90% of it (80-100%; the company also expects general-purpose server demand to rise). Credit: Pegatron sells to neocloud and cloud customers under its own name, so 85% (60-100%) is credited = US$1.05bn (US$0.48-1.78bn). The platforms named are Nvidia GB300 NVL72, B200 and B300, so 95% (85-100%) of the credited amount is GPU systems = US$0.99bn. Pool: GPU servers and racks outside mainland China, US$277.8bn over 3Q25-2Q26 (IDC figures as used in the approved card: GPU $301.3bn of $373.4bn accelerated servers, less China at 9% and 70% GPU); range US$257-294bn. Share = US$0.99bn / 277.8bn (range US$0.41-1.78bn against the pool range)."},"custom-accelerator-systems":{"low":0,"central":0.084,"high":0.581,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"LTM to 2026-06-30 (3Q25-2Q26)","periodEnd":"2026-06-30","source":{"url":"https://telecomlead.com/electronics/pegatron-q2-2026-revenue-rises-2-7-as-ai-server-boom-drives-84-computing-growth-127327","title":"Pegatron Q2 2026 Revenue Rises 2.7% as AI Server Boom Drives 84% Computing Growth (TelecomLead, 17 Aug 2026)","publisher":"TelecomLead","publishedDate":"2026-08-17","accessedDate":"2026-10-22","tier":"T5","locator":"Q2 2026 revenue NT$274.448bn; Computing segment +84% y/y and +82% q/q 'primarily due to higher Server shipments', 23% of revenue (14% in Q1 2026, 13% in Q2 2025); server revenue targeted at about NT$100bn in 2026"},"note":"Pegatron does not report server revenue by quarter; this is a T5 build from opened statements. Anchors: (1) Pegatron said on 11 Mar 2026 that it aims to grow AI server revenue more than 10-fold in 2026, driven by Nvidia GB300 orders from neocloud and large cloud customers, and that shipments will rise each quarter with the second half stronger (Taipei Times, 12 Mar 2026, https://www.taipeitimes.com/News/biz/archives/2026/03/12/2003853649); the server revenue target is about NT$100bn for 2026 (TelecomLead, the page cited here), which puts 2025 AI server revenue near NT$10bn; (2) the Computing segment was 23% of Q2 2026 revenue of NT$274.4bn = NT$63.1bn against 13% of NT$267.2bn (= 274.448 / 1.027) = NT$34.7bn in Q2 2025, so about NT$28bn of the increase is servers; Q1 2026 Computing was 14% of NT$244.1bn = NT$34.2bn. Server revenue taken as NT$2.5bn in Q3 2025 (1.5-4), NT$4.5bn in Q4 2025 (3-7), NT$5bn in Q1 2026 (2-9) and NT$31bn in Q2 2026 (25-36) = NT$43bn (31.5-56), converted at FRED's quarterly means of NT$ per US$ (https://fred.stlouisfed.org/data/EXTAUS: 29.95, 31.03, 31.62, 31.60) = US$1.37bn (US$1.00-1.78bn). AI servers are 90% of it (80-100%; the company also expects general-purpose server demand to rise). Credit: Pegatron sells to neocloud and cloud customers under its own name, so 85% (60-100%) is credited = US$1.05bn (US$0.48-1.78bn). Co-CEO Gary Cheng said inference demand should raise demand for ASIC-based solutions, but no ASIC server shipment is named, so 5% (0-15%) of the credited amount is allowed for systems around cloud companies' own accelerators = US$0.05bn. Pool: servers for non-GPU (other) accelerators outside mainland China, US$62.0bn over 3Q25-2Q26 (IDC 'other accelerated' servers $72.1bn less China's share, as in the approved card); range US$46-75bn. Share = US$0.05bn / 62bn (range US$0-0.27bn against the pool range)."},"china":{"low":0,"central":0.089,"high":0.385,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"LTM to 2026-06-30 (3Q25-2Q26)","periodEnd":"2026-06-30","source":{"url":"https://telecomlead.com/electronics/pegatron-q2-2026-revenue-rises-2-7-as-ai-server-boom-drives-84-computing-growth-127327","title":"Pegatron Q2 2026 Revenue Rises 2.7% as AI Server Boom Drives 84% Computing Growth (TelecomLead, 17 Aug 2026)","publisher":"TelecomLead","publishedDate":"2026-08-17","accessedDate":"2026-10-22","tier":"T5","locator":"Q2 2026 revenue NT$274.448bn; Computing segment +84% y/y and +82% q/q 'primarily due to higher Server shipments', 23% of revenue (14% in Q1 2026, 13% in Q2 2025); server revenue targeted at about NT$100bn in 2026"},"note":"T5 allowance, not a documented sale: the platforms and customers named for Pegatron's AI servers are Nvidia GB300, B200 and B300 systems for neocloud and cloud customers, which US export rules keep out of mainland China, and Pegatron is 'pursuing additional large hyperscale cloud customers' (TelecomLead, 17 Aug 2026). Upper end set at US$0.1bn; central at US$0.03bn. Pool: AI servers sold in mainland China, about US$33.6bn over 3Q25-2Q26 (9% of IDC's $373.4bn accelerated servers, as in the approved card); range US$26-45bn. Share = US$0.03bn / 33.6bn (range US$0-0.1bn against the pool range)."}},"screening":{"gpu-systems":{"levelLow":1,"levelHigh":2},"custom-accelerator-systems":{"levelLow":1,"levelHigh":2},"china":{"levelLow":1,"levelHigh":2}},"checks":{"sourceSupport":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-22","passed":true,"sha256":"a88c8f8158f2a975bbac07d030011353ed062c664d843c1822cabb4480be3e50"},"consistency":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-22","passed":true,"sha256":"a88c8f8158f2a975bbac07d030011353ed062c664d843c1822cabb4480be3e50"}}},"compal":{"schema":"5lc-index-inputs/1","key":"infrastructure-servers-systems","entityId":"compal","name":"Compal","kind":"public","commercialActivityVerified":true,"shares":{"gpu-systems":{"low":0.183,"central":0.248,"high":0.39,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"LTM to 2026-06-30 (3Q25-2Q26)","periodEnd":"2026-06-30","source":{"url":"https://www.taiwannews.com.tw/en/news/6420917","title":"Taiwan News, 13 Aug 2026 (citing CNA and Anue): Compal second-quarter 2026 results, AI servers about 70% of server revenue","publisher":"Taiwan News","publishedDate":"2026-08-13","accessedDate":"2026-10-22","tier":"T5","locator":"Q2 2026 revenue NT$238.29bn, +18.38% q/q and +32.06% y/y; non-PC about 34% of revenue; 'AI servers made up about 70% of total server revenue'"},"note":"Compal does not report server revenue by quarter; this is a T5 build from opened statements. AI share of server revenue: about 70% in Q2 2026 (Taiwan News, 13 Aug 2026, the page cited here, reporting the results briefing); AI servers expected at about 80% of server revenue in 2026, up from about 20% at the end of the previous quarter (Taipei Times, 23 Jan 2026, https://www.taipeitimes.com/News/biz/archives/2026/01/23/2003851053; the same 80% in Taipei Times, 3 Mar 2026, https://www.taipeitimes.com/News/biz/archives/2026/03/03/2003853144). Server revenue: Q1 2026 NT$10bn, 5% of revenue (Taipei Times, 12 Aug 2026, https://www.taipeitimes.com/News/biz/archives/2026/08/12/2003862319); Q4 2025 about NT$5bn (4-6), because Q1 server revenue was projected to roughly double from the previous quarter (Taipei Times, 3 Mar 2026); Q3 2025 about NT$3bn (2-5), assumed at or below Q4; Q2 2026 about NT$18bn (13-26), from two cross-checks: Q2 non-PC revenue of 34% of NT$238.29bn = NT$81bn against 30% of NT$180.4bn = NT$54bn a year earlier (+52%, driven mainly by servers, Taiwan News), and the company's aim of servers at about 10% of 2026 revenue (Taipei Times, 3 Mar 2026) with Q3 server revenue guided up by a high double-digit percentage. AI server revenue by quarter = server revenue x AI share (Q3 2025 20%, Q4 2025 20%, Q1 2026 75% (70-80%), Q2 2026 70%) converted at FRED's quarterly means of NT$ per US$ (https://fred.stlouisfed.org/data/EXTAUS: 29.95, 31.03, 31.62, 31.60) = US$0.69bn (US$0.54-1.00bn). Compal's shipments are Nvidia HGX B300 and RTX AI platforms, so all of it is placed in GPU systems; it is credited to Compal because the customers named (GMI Cloud, Verda) are neoclouds buying L10 systems. The earlier assumption of US$1.1bn of LTM server revenue at 80% AI was higher than this build because AI was only about 20% of server revenue in the second half of 2025. Pool: GPU servers and racks outside mainland China, US$277.8bn over 3Q25-2Q26 (IDC figures as used in the approved card: GPU $301.3bn of $373.4bn accelerated servers, less China at 9% and 70% GPU); range US$257-294bn. Share = US$0.69bn / 277.8bn (range US$0.54-1.00bn against the pool range)."},"custom-accelerator-systems":{"low":0,"central":0.121,"high":0.656,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"LTM to 2026-06-30","periodEnd":"2026-06-30","source":{"url":"https://www.taipeitimes.com/News/biz/archives/2026/01/23/2003851053","title":"Compal targets revenue of NT$1 trillion (Taipei Times, Meryl Kao, 23 Jan 2026)","publisher":"Taipei Times","publishedDate":"2026-01-23","accessedDate":"2026-10-22","tier":"T5","locator":"'The server business will keep focusing on GPU-based systems', with talks under way with tier-one cloud providers and OEMs about ASIC-based servers"},"note":"T5 allowance, not a documented sale: Compal says its server business keeps focusing on GPU-based systems and is only in talks about ASIC-based servers (Taipei Times, 23 Jan 2026), so no system built around a cloud company's own accelerator is shown as shipped. Upper end set at US$0.3bn; central at a quarter of that. Pool: servers for non-GPU (other) accelerators outside mainland China, US$62.0bn over 3Q25-2Q26 (IDC 'other accelerated' servers $72.1bn less China's share, as in the approved card); range US$46-75bn. Share = US$0.075bn / 62bn (range US$0-0.3bn against the pool range)."},"china":{"low":0,"central":0.074,"high":0.383,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"LTM to 2026-06-30","periodEnd":"2026-06-30","source":{"url":"https://www.taipeitimes.com/News/biz/archives/2026/08/12/2003862319","title":"Taipei Times, 12 August 2026, p. 9 (Meryl Kao): Compal server outlook","publisher":"Taipei Times","publishedDate":"2026-08-12","accessedDate":"2026-10-22","tier":"T5","locator":"AI server shipments go mainly to neocloud customers (GMI Cloud, Verda); Texas plants serve US hyperscalers and enterprises, Taiwan serves neoclouds"},"note":"T5 allowance, not a documented sale: the customers and plants named for Compal's AI servers are outside mainland China, and US export rules restrict Nvidia B300 systems there. Upper end set at US$0.1bn; central at a quarter of that. Pool: AI servers sold in mainland China, about US$33.6bn over 3Q25-2Q26 (9% of IDC's $373.4bn accelerated servers, as in the approved card); range US$26-45bn. Share = US$0.025bn / 33.6bn (range US$0-0.1bn against the pool range)."}},"screening":{"gpu-systems":{"levelLow":1,"levelHigh":2},"custom-accelerator-systems":{"levelLow":1,"levelHigh":2},"china":{"levelLow":1,"levelHigh":2}},"checks":{"sourceSupport":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-22","passed":true,"sha256":"0135aa9ca427a65eb113db407a7d3cd091b21eafcac7668e69e065f7df990fd2"},"consistency":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-22","passed":true,"sha256":"0135aa9ca427a65eb113db407a7d3cd091b21eafcac7668e69e065f7df990fd2"}}},"cerebras":{"schema":"5lc-index-inputs/1","key":"infrastructure-servers-systems","entityId":"cerebras","name":"Cerebras (CS-3 systems)","kind":"public","commercialActivityVerified":true,"shares":{"gpu-systems":{"participates":false,"evidence":{"url":"https://www.sec.gov/Archives/edgar/data/2021728/000162828026035214/cerebras-424b4.htm","title":"Cerebras Systems Inc. final prospectus (Form 424B4), May 13, 2026 (read from the local copy .cache/sec/cerebras)","publisher":"Cerebras Systems Inc.","publishedDate":"2026-05-14","accessedDate":"2026-10-22","tier":"T1","locator":"Business and MD&A: hardware revenue is sales of its own AI systems built on the Wafer-Scale Engine, not GPU servers"}},"custom-accelerator-systems":{"low":0.453,"central":0.577,"high":0.919,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"FY2025 (calendar year 2025)","periodEnd":"2025-12-31","source":{"url":"https://www.sec.gov/Archives/edgar/data/2021728/000162828026035214/cerebras-424b4.htm","title":"Cerebras Systems Inc. final prospectus (Form 424B4), May 13, 2026 (read from the local copy .cache/sec/cerebras)","publisher":"Cerebras Systems Inc.","publishedDate":"2026-05-14","accessedDate":"2026-10-22","tier":"T4","locator":"MD&A, year ended 31 Dec 2025: hardware revenue $358.44M (2024: $211.97M); hardware demand driven by strategic partner MBZUAI"},"note":"Cerebras sells complete systems built on its own wafer-scale accelerator, so it counts in the non-GPU function by analogy with NVIDIA's DGX. FY2025 hardware revenue $358.44M (MD&A results table in the prospectus, read from the local copy .cache/sec/cerebras), against IDC's 2025-26 window; the Q2 2026 10-Q hardware figure (H1 2026 $164.7M vs $140.0M, via the dossier) is not re-opened because investors.cerebras.ai returned 403. Tier T4 (company figure over an IDC pool). Pool: servers for non-GPU (other) accelerators outside mainland China, US$62.0bn over 3Q25-2Q26 (IDC 'other accelerated' servers $72.1bn less China's share, as in the approved card); range US$46-75bn. Share = US$0.358bn / 62bn (range US$0.34-0.42bn against the pool range)."},"china":{"low":0,"central":0.05,"high":0.15,"unit":"percent","basis":"Share of worldwide AI server sales in the function by value, latest four quarters, US dollars, credited to the seller of the finished system","period":"FY2025","periodEnd":"2025-12-31","source":{"url":"https://www.sec.gov/Archives/edgar/data/2021728/000162828026035214/cerebras-424b4.htm","title":"Cerebras Systems Inc. final prospectus (Form 424B4), May 13, 2026 (read from the local copy .cache/sec/cerebras)","publisher":"Cerebras Systems Inc.","publishedDate":"2026-05-14","accessedDate":"2026-10-22","tier":"T5","locator":"MD&A, year ended 31 Dec 2025: hardware revenue $358.44M (2024: $211.97M); hardware demand driven by strategic partner MBZUAI"},"note":"Hardware revenue is concentrated in UAE partners (G42 24% and MBZUAI 62% of FY2025 group revenue per the dossier from the prospectus), leaving at most about $50M for all other customers; no China sales identified. Tier T5. Pool: AI servers sold in mainland China, about US$33.6bn over 3Q25-2Q26 (9% of IDC's $373.4bn accelerated servers, as in the approved card); range US$26-45bn."}},"screening":{"custom-accelerator-systems":{"levelLow":2,"levelHigh":4},"china":{"levelLow":1,"levelHigh":2}},"checks":{"sourceSupport":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-22","passed":true,"sha256":"04f3f74aa02a5e914bd5f35f04b6528980d99fa42c507df047d6c35031d5d155"},"consistency":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-22","passed":true,"sha256":"04f3f74aa02a5e914bd5f35f04b6528980d99fa42c507df047d6c35031d5d155"}}}},"ratings":{"hon-hai":{"a":{"schema":"5lc-index-rating/1","key":"infrastructure-servers-systems","entityId":"hon-hai","rater":"a","model":"claude-opus-5-5","ratedAt":"2026-10-22","functions":{"gpu-systems":{"alternatives":{"level":3,"evidence":[{"url":"https://www.digitimes.com/news/a20251210PD211.html","date":"2025-12-10","support":"DIGITIMES: Nvidia may narrow its approved assemblers and consolidate L10 compute-tray assembly to three ODMs (not finalised; full date shown)."},{"url":"https://focustaiwan.tw/business/202608120019","date":"2026-08-12","support":"Focus Taiwan: Hon Hai says it holds over 40% of the global AI server market and aims for 50% (company claim; full date shown)."}]},"switchingTime":{"level":3,"evidence":[{"url":"https://www.trendforce.com/news/2025/01/14/news-nvidia-gb200-racks-reportedly-overheat-major-clients-cut-orders/","date":"2025-01-14","support":"TrendForce, citing The Information: early GB200 racks overheated, rack designs were revised several times and hyperscalers scaled back orders (full date shown)."},{"url":"https://www.taipeitimes.com/News/biz/archives/2026/05/15/2003857345","date":"2026-05-15","support":"Taipei Times: Quanta aims to double AI-related capacity this year and expects more ASIC server shipments (company statements; full date shown)."}]},"capabilityGap":{"level":3,"evidence":[{"url":"https://focustaiwan.tw/business/202608120019","date":"2026-08-12","support":"Focus Taiwan: Hon Hai says it holds over 40% of the global AI server market and aims for 50% (company claim; full date shown)."},{"url":"https://www.taipeitimes.com/News/biz/archives/2026/05/15/2003857344","date":"2026-05-15","support":"Taipei Times: Hon Hai projects 2026 AI rack shipments to more than double, with capex up over 30% (company guidance; full date shown)."}]}},"custom-accelerator-systems":{"alternatives":{"level":3,"evidence":[{"url":"https://wccftech.com/nvidia-key-supply-chain-partner-foxconn-has-reportedly-won-orders-for-google-tpu/","date":"2025-12-01","support":"Wccftech, citing Taiwan's Economic Daily: Foxconn reportedly received orders for Google TPU compute trays (single-source report; full date shown)."},{"url":"https://www.digitimes.com/news/a20251210PD211.html","date":"2025-12-10","support":"DIGITIMES: Inventec is a key Google TPU server supplier; Quanta, Foxconn, Wiwynn, Wistron and Mitac also pursue ASIC server orders (full date shown)."},{"url":"https://www.taipeitimes.com/News/biz/archives/2026/02/10/2003852050","date":"2026-02-10","support":"Taipei Times: Wiwynn's clients include Google, AWS, Meta and Microsoft; it expects strong ASIC AI server shipments (company statements; full date)."}]},"switchingTime":{"level":3,"evidence":[{"url":"https://www.sec.gov/Archives/edgar/data/1030894/000103089426000011/cls-20251231.htm","date":"2026-02-27","support":"Celestica 10-K (Item 1A): ramping new programs can take several months to over a year before production; date is EDGAR timestamp."},{"url":"https://www.digitimes.com/news/a20260623PD213.html","date":"2026-06-23","support":"DIGITIMES: AWS Trainium 3 servers moving into volume production, with the ramp possibly extending into early 2027 (full date shown)."}]},"capabilityGap":{"level":2,"evidence":[{"url":"https://www.taipeitimes.com/News/biz/archives/2026/02/10/2003852050","date":"2026-02-10","support":"Taipei Times: Wiwynn's clients include Google, AWS, Meta and Microsoft; it expects strong ASIC AI server shipments (company statements; full date)."},{"url":"https://www.taipeitimes.com/News/biz/archives/2026/05/15/2003857345","date":"2026-05-15","support":"Taipei Times: Quanta aims to double AI-related capacity this year and expects more ASIC server shipments (company statements; full date shown)."}]}},"china":{"alternatives":{"level":1,"evidence":[{"url":"https://www.idc.com/?p=104194","date":"2025-10-08","support":"IDC China H1 2025 accelerated servers: Inspur, H3C, Lenovo top three by revenue near 50%; by units Inspur, H3C, Nettrix about 43% (page date)."},{"url":"https://www.cls.cn/detail/1678572","date":"2024-05-16","support":"Cailian Press: China Mobile's 2024-25 AI computing procurement split among seven makers, e.g. Henan Kunlun 21.05%, Huakun Zhenyu 17.54%, Baode 15.79% (full date)."}]},"switchingTime":{"level":2,"evidence":[{"url":"https://www.cls.cn/detail/1678572","date":"2024-05-16","support":"Cailian Press: China Mobile's 2024-25 AI computing procurement split among seven makers, e.g. Henan Kunlun 21.05%, Huakun Zhenyu 17.54%, Baode 15.79% (full date)."}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://www.idc.com/?p=104194","date":"2025-10-08","support":"IDC China H1 2025 accelerated servers: Inspur, H3C, Lenovo top three by revenue near 50%; by units Inspur, H3C, Nettrix about 43% (page date)."}]}}}},"b":{"schema":"5lc-index-rating/1","key":"infrastructure-servers-systems","entityId":"hon-hai","rater":"b","model":"claude-sonnet-5-5","ratedAt":"2026-10-22","functions":{"gpu-systems":{"alternatives":{"level":1,"evidence":[{"url":"https://investor.nvidia.com/news/press-release-details/2025/NVIDIA-Blackwell-Ultra-AI-Factory-Platform-Paves-Way-for-Age-of-AI-Reasoning/","date":"2025-03-18","support":"NVIDIA names 16 system makers (Dell, HPE, Lenovo, Supermicro, Cisco, ASUS, Foxconn, GIGABYTE, Inventec, Pegatron, QCT, Wistron, Wiwynn and others) expected to deliver Blackwell Ultra servers from 2H 2025, so the same GPU platform has many qualified sellers. (full date shown)"},{"url":"https://finance.technews.tw/2026/09/10/morgan-stanley-is-bullish-on-nvidias-rack-server-shipments-expecting-them-to-double-this-year/","date":"2026-09-10","support":"Morgan Stanley (via TechNews): 2026 NVL72 rack supply Foxconn 39%, Quanta 26%, Wistron 22%, others 13% (7% in 2025). (full date shown)"},{"url":"https://s204.q4cdn.com/707617056/files/doc_financials/2026/q4/V2/FQ426-SMCI-Prepared-Remarks.pdf","date":"2026-08-11","support":"Supermicro says it is shipping volume SKUs across its GB300 NVL72 and HGX B300 lines (company statement), so branded makers also supply racks. (full date shown)"}]},"switchingTime":{"level":2,"evidence":[{"url":"https://finance.technews.tw/2026/09/10/morgan-stanley-is-bullish-on-nvidias-rack-server-shipments-expecting-them-to-double-this-year/","date":"2026-09-10","support":"Makers outside the three big ODMs rise from 7% to 13% of NVL72 supply in a year and industry output reached about 8,400 racks in Aug 2026 (+27% on July), so a share of Foxconn's size can be re-spread in months (inference). (full date shown)"},{"url":"https://www.taipeitimes.com/News/biz/archives/2026/03/17/2003853939","date":"2026-03-17","support":"Hon Hai says AI rack shipments could roughly double in 2026 and capex rises 30% for new capacity, indicating that about a year is needed to add capacity of this size (inference). (full date shown)"}]},"capabilityGap":{"level":2,"evidence":[{"url":"https://finance.technews.tw/2026/09/10/morgan-stanley-is-bullish-on-nvidias-rack-server-shipments-expecting-them-to-double-this-year/","date":"2026-09-10","support":"Aug 2026 racks: Foxconn about 2,600, Quanta 2,500-2,550, Wistron 2,300-2,400. Full-year 2026 forecasts are 31,700 against 18,700 and 17,900, so the best alternative needs roughly 70-77% more output to match Foxconn. (full date shown)"},{"url":"https://www.taipeitimes.com/News/biz/archives/2026/03/17/2003853939","date":"2026-03-17","support":"Company says it accounts for more than 40% of the global AI server market (company claim). (full date shown)"}]}},"custom-accelerator-systems":{"alternatives":{"level":3,"evidence":[{"url":"https://www.digitimes.com/news/a20251210PD211.html","date":"2025-12-10","support":"DIGITIMES lists Quanta, Foxconn, Wiwynn, Wistron, Inventec and Mitac as ODMs pursuing ASIC server orders and names Inventec as a Google TPU server supplier; in the sources I opened, one to three builders are confirmed per cloud program (my count). (full date shown)"},{"url":"https://iconnect007.com/article/148613/celestica-announces-q4-fy-2025-financial-results/148610/pcb","date":"2026-01-29","support":"Celestica says it is a preferred manufacturing partner for Google's TPU systems and is adding US capacity due in 2027 (company claim). (full date shown)"},{"url":"https://www.igorslab.de/en/foxconn-reaches-for-the-heart-of-the-ki-infrastructure-major-order-for-googles-tpu-server-reveals-new-dynamics-in-the-server-market/","date":"2025-12-02","support":"Report (Taiwan Economic Daily via Igor's Lab) that Foxconn will supply compute trays at 1:1 with Google TPU racks; the page names no other TPU builder. (full date shown)"}]},"switchingTime":{"level":3,"evidence":[{"url":"https://www.sdxcentral.com/news/meta-places-order-for-its-next-gen-asic-powered-ai-servers-partners-with-broadcom-quanta/","date":"2025-08-04","support":"Meta's ASIC rack program: designs still being finalised in 2H 2025, trial production in Q4 2025, deployment from end-2025; reported racks run above 180 kW with custom cabinets and water cooling (supply-chain reports), roughly a year from design to volume. (full date shown)"},{"url":"https://www.sec.gov/Archives/edgar/data/1030894/000103089426000011/cls-20251231.htm","date":"2026-02-27","support":"Celestica's 10-K says a technology transition in one hyperscaler's AI/ML compute program cut its Enterprise end market in 2025: program changes take quarters and disrupt the builder. (full date shown)"}]},"capabilityGap":{"level":2,"evidence":[{"url":"https://www.taipeitimes.com/News/biz/archives/2026/03/17/2003853939","date":"2026-03-17","support":"About 20% of Hon Hai servers use AI ASICs and 80% GPUs (company statement), with rack shipments set to roughly double in 2026. (full date shown)"},{"url":"https://www.theregister.com/2025/12/07/trainium3_all_nvidia_nvl72_mold/","date":"2025-12-07","support":"AWS's Trainium3 racks are MGX-style like Nvidia's NVL72 and Trainium4 is to fit Nvidia's MGX chassis, so other rack builders can follow, but the largest ASIC volume is not matched by any one alternative (inference). (full date shown)"}]}},"china":{"alternatives":{"level":1,"evidence":[{"url":"https://www.chinabaogao.com/jingzheng/202510/767092.html","date":"2025-10-14","support":"IDC China 1H25: Inspur, H3C and Lenovo top accelerated-server sales (about 50% together); by shipments Inspur, H3C and Nettrix (Ningchang), about 43%: roughly half of the market sits with other domestic makers. (full date shown)"},{"url":"https://www.trendforce.com/news/2026/07/09/news-china-reportedly-to-allow-nvidia-h200-imports-but-approvals-may-be-capped-below-200k-chips-less-than-half-requested/","date":"2026-07-09","support":"Foxconn was named with Lenovo as an approved H200 distributor for China (TrendForce, citing Reuters); no Foxconn China AI-server revenue was found, so evidence is thin. (full date shown)"}]},"switchingTime":{"level":1,"evidence":[{"url":"https://www.chinabaogao.com/jingzheng/202510/767092.html","date":"2025-10-14","support":"The IDC China top three do not include Hon Hai; the top three hold about half of the market, so Hon Hai's small volume is easily re-spread (inference). (full date shown)"}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://technode.com/2026/10/14/nvidias-china-ai-chip-share-falls-to-34-as-huawei-reaches-28/","date":"2026-10-14","support":"China's AI accelerator supply is the constraint (Nvidia shipments down 44% in 1H26); other server makers have the same access, so no alternative is behind. (full date shown)"}]}}}},"reconciled":{"schema":"5lc-index-rating/1","key":"infrastructure-servers-systems","entityId":"hon-hai","rater":"reconciled","model":"claude-opus-5-5","ratedAt":"2026-10-22","functions":{"gpu-systems":{"alternatives":{"level":2,"evidence":[{"url":"https://finance.technews.tw/2026/09/10/morgan-stanley-is-bullish-on-nvidias-rack-server-shipments-expecting-them-to-double-this-year/","date":"2026-09-10","support":"Morgan Stanley via TechNews: 2026 NVL72 rack supply Foxconn 39%, Quanta 26%, Wistron 22%, others 13% (up from 7%); several makers ship racks at volume (full date shown)."},{"url":"https://www.dell.com/en-us/dt/corporate/newsroom/announcements/detailpage.press-releases~usa~2026~09~dell-technologies-delivers-second-quarter-fiscal-2027-financial-results.htm","date":"2026-09-01","support":"Dell says Q2 FY27 AI-optimized server revenue was $16.4bn and raises its FY27 outlook to $74.0bn: a branded seller at comparable scale (company statement; full date shown)."}]},"switchingTime":{"level":3,"evidence":[{"url":"https://www.taipeitimes.com/News/biz/archives/2026/05/15/2003857345","date":"2026-05-15","support":"Taipei Times: Quanta 'aims to double its AI-related capacity this year', with NT$30bn capex for US and Thailand capacity, so large capacity additions take about a year (company statement; full date shown)."},{"url":"https://www.sec.gov/Archives/edgar/data/1030894/000103089426000011/cls-20251231.htm","date":"2026-02-27","support":"Celestica 10-K, Item 1A: 'Ramping new programs can range from several months to over a year before production starts' (date is the EDGAR filing timestamp)."},{"url":"https://finance.technews.tw/2026/09/10/morgan-stanley-is-bullish-on-nvidias-rack-server-shipments-expecting-them-to-double-this-year/","date":"2026-09-10","support":"Foxconn holds 39% of 2026 NVL72 racks, so the other makers would need about 64% more output to absorb it (5LayerCake arithmetic: 39/61; full date shown)."}]},"capabilityGap":{"level":2,"evidence":[{"url":"https://finance.technews.tw/2026/09/10/morgan-stanley-is-bullish-on-nvidias-rack-server-shipments-expecting-them-to-double-this-year/","date":"2026-09-10","support":"August 2026 racks: Foxconn about 2,600, Quanta 2,500-2,550 (+22% on July), Wistron 2,300-2,400; the best alternative is near Foxconn's monthly output (full date shown)."},{"url":"https://www.dell.com/en-us/dt/corporate/newsroom/announcements/detailpage.press-releases~usa~2026~09~dell-technologies-delivers-second-quarter-fiscal-2027-financial-results.htm","date":"2026-09-01","support":"Dell reports a record $95bn AI backlog and $16.4bn quarterly AI server revenue, showing large existing GPU-system capacity outside Foxconn (company statement; full date shown)."}]}}}}},"quanta-computer":{"a":{"schema":"5lc-index-rating/1","key":"infrastructure-servers-systems","entityId":"quanta-computer","rater":"a","model":"claude-opus-5-5","ratedAt":"2026-10-22","functions":{"gpu-systems":{"alternatives":{"level":2,"evidence":[{"url":"https://www.sec.gov/Archives/edgar/data/0001375365/000137536526000022/smci-20260630.htm","date":"2026-08-31","support":"Supermicro's 10-K (Item 1, Competition) names Cisco, Dell, HPE, Lenovo and ODMs Foxconn, Quanta, Wiwynn as rivals; date is the EDGAR filing timestamp."},{"url":"https://nvidianews.nvidia.com/news/nvidia-blackwell-ultra-ai-factory-platform-paves-way-for-age-of-ai-reasoning","date":"2025-03-18","support":"NVIDIA names Cisco, Dell, HPE, Lenovo, Supermicro plus Aivres, ASRock Rack, ASUS, Eviden, Foxconn, GIGABYTE, Inventec, Pegatron, QCT, Wistron, Wiwynn as Blackwell Ultra system makers (full date shown)."}]},"switchingTime":{"level":2,"evidence":[{"url":"https://www.digitimes.com/news/a20241112PD202/ai-server-nvidia-zt-systems-supermicro-amd-acquisition.html","date":"2024-11-12","support":"DIGITIMES: experts say moving AI server orders between makers is a complex process that cannot be done overnight (full date shown)."},{"url":"https://www.taipeitimes.com/News/biz/archives/2026/05/15/2003857344","date":"2026-05-15","support":"Taipei Times: Hon Hai projects 2026 AI rack shipments to more than double, with capex up over 30% (company guidance; full date shown)."}]},"capabilityGap":{"level":2,"evidence":[{"url":"https://focustaiwan.tw/business/202608120019","date":"2026-08-12","support":"Focus Taiwan: Hon Hai says it holds over 40% of the global AI server market and aims for 50% (company claim; full date shown)."}]}},"custom-accelerator-systems":{"alternatives":{"level":3,"evidence":[{"url":"https://www.digitimes.com/news/a20251210PD211.html","date":"2025-12-10","support":"DIGITIMES: Inventec is a key Google TPU server supplier; Quanta, Foxconn, Wiwynn, Wistron and Mitac also pursue ASIC server orders (full date shown)."},{"url":"https://www.taipeitimes.com/News/biz/archives/2026/02/10/2003852050","date":"2026-02-10","support":"Taipei Times: Wiwynn's clients include Google, AWS, Meta and Microsoft; it expects strong ASIC AI server shipments (company statements; full date)."}]},"switchingTime":{"level":3,"evidence":[{"url":"https://www.sec.gov/Archives/edgar/data/1030894/000103089426000011/cls-20251231.htm","date":"2026-02-27","support":"Celestica 10-K (Item 1A): ramping new programs can take several months to over a year before production; date is EDGAR timestamp."},{"url":"https://www.digitimes.com/news/a20260623PD213.html","date":"2026-06-23","support":"DIGITIMES: AWS Trainium 3 servers moving into volume production, with the ramp possibly extending into early 2027 (full date shown)."}]},"capabilityGap":{"level":2,"evidence":[{"url":"https://wccftech.com/nvidia-key-supply-chain-partner-foxconn-has-reportedly-won-orders-for-google-tpu/","date":"2025-12-01","support":"Wccftech, citing Taiwan's Economic Daily: Foxconn reportedly received orders for Google TPU compute trays (single-source report; full date shown)."},{"url":"https://www.taipeitimes.com/News/biz/archives/2026/02/10/2003852050","date":"2026-02-10","support":"Taipei Times: Wiwynn's clients include Google, AWS, Meta and Microsoft; it expects strong ASIC AI server shipments (company statements; full date)."}]}}}},"b":{"schema":"5lc-index-rating/1","key":"infrastructure-servers-systems","entityId":"quanta-computer","rater":"b","model":"claude-sonnet-5-5","ratedAt":"2026-10-22","functions":{"gpu-systems":{"alternatives":{"level":1,"evidence":[{"url":"https://investor.nvidia.com/news/press-release-details/2025/NVIDIA-Blackwell-Ultra-AI-Factory-Platform-Paves-Way-for-Age-of-AI-Reasoning/","date":"2025-03-18","support":"QCT is one of 16 named makers (Dell, HPE, Lenovo, Supermicro, ASUS, Foxconn, GIGABYTE, Inventec, Pegatron, Wistron, Wiwynn and others) expected to deliver Blackwell Ultra servers. (full date shown)"},{"url":"https://www.sec.gov/Archives/edgar/data/0001375365/000137536526000022/smci-20260630.htm","date":"2026-08-31","support":"Supermicro's 10-K lists Quanta among ODMs it competes with, alongside Foxconn and Wiwynn and the branded vendors Cisco, Dell, HPE and Lenovo (Item 1). (full date shown)"}]},"switchingTime":{"level":2,"evidence":[{"url":"https://finance.technews.tw/2026/09/10/morgan-stanley-is-bullish-on-nvidias-rack-server-shipments-expecting-them-to-double-this-year/","date":"2026-09-10","support":"Morgan Stanley puts Quanta at 23% of 2025 and 26% of 2026 NVL72 rack supply, about 18,700 racks in 2026 with upside; re-spreading that volume takes capacity additions over several quarters (inference). (full date shown)"},{"url":"https://www.taipeitimes.com/News/biz/archives/2025/11/14/2003847162","date":"2025-11-14","support":"Quanta says AI server order visibility runs through 2027 and it expects server production capacity to double, so capacity is committed rather than idle. (full date shown)"}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://finance.technews.tw/2026/09/10/morgan-stanley-is-bullish-on-nvidias-rack-server-shipments-expecting-them-to-double-this-year/","date":"2026-09-10","support":"Foxconn is larger (39% of 2026 racks) and Wistron near parity in Aug 2026 (about 2,350 against Quanta's 2,500-2,550), so the best alternative is not smaller. (full date shown)"}]}},"custom-accelerator-systems":{"alternatives":{"level":3,"evidence":[{"url":"https://www.sdxcentral.com/news/meta-places-order-for-its-next-gen-asic-powered-ai-servers-partners-with-broadcom-quanta/","date":"2025-08-04","support":"Economic Daily News (via SDxCentral) reports Quanta handles final assembly of Meta's next-generation ASIC servers, up to 6,000 racks, with Broadcom the chip supplier and SynMing custom cabinets; no second assembler is named. (full date shown)"},{"url":"https://www.digitimes.com/news/a20251210PD211.html","date":"2025-12-10","support":"DIGITIMES lists Quanta, Foxconn, Wiwynn, Wistron, Inventec and Mitac as ODMs pursuing ASIC server orders and names Inventec as a Google TPU server supplier. (full date shown)"}]},"switchingTime":{"level":3,"evidence":[{"url":"https://www.sdxcentral.com/news/meta-places-order-for-its-next-gen-asic-powered-ai-servers-partners-with-broadcom-quanta/","date":"2025-08-04","support":"Meta was still finalising two to three ASIC server designs in 2H 2025 with trial production in Q4 2025 and deployment from end-2025; racks above 180 kW need custom cabinets and water cooling (supply-chain reports). (full date shown)"},{"url":"https://www.taipeitimes.com/News/biz/archives/2025/11/14/2003847162","date":"2025-11-14","support":"Quanta expects ASIC-based AI server shipments to double in 2026 and says it is working with customers on assembly, so replacement means a new customer-specific build and ramp. (full date shown)"}]},"capabilityGap":{"level":2,"evidence":[{"url":"https://www.sdxcentral.com/news/meta-places-order-for-its-next-gen-asic-powered-ai-servers-partners-with-broadcom-quanta/","date":"2025-08-04","support":"Highly customised cabinets, sidecars and water cooling are reported for the Meta program (supply-chain reports); the source names no other assembler for it, so alternatives would start without that program's history (inference). (full date shown)"},{"url":"https://www.theregister.com/2025/12/07/trainium3_all_nvidia_nvl72_mold/","date":"2025-12-07","support":"Trainium3 racks use an MGX-style design shared with Nvidia's racks, which narrows the gap for builders who already make NVL72 racks (The Register). (full date shown)"}]}}}}},"dell":{"a":{"schema":"5lc-index-rating/1","key":"infrastructure-servers-systems","entityId":"dell","rater":"a","model":"claude-opus-5-5","ratedAt":"2026-10-22","functions":{"gpu-systems":{"alternatives":{"level":2,"evidence":[{"url":"https://www.sec.gov/Archives/edgar/data/0001375365/000137536526000022/smci-20260630.htm","date":"2026-08-31","support":"Supermicro's 10-K (Item 1, Competition) names Cisco, Dell, HPE, Lenovo and ODMs Foxconn, Quanta, Wiwynn as rivals; date is the EDGAR filing timestamp."},{"url":"https://www.sec.gov/Archives/edgar/data/1571996/000157199626000008/dell-20260130.htm","date":"2026-03-16","support":"Dell's 10-K (Item 1A) says large AI deals are subject to intense competition and pricing pressure; date is the EDGAR filing timestamp."}]},"switchingTime":{"level":2,"evidence":[{"url":"https://www.digitimes.com/news/a20241112PD202/ai-server-nvidia-zt-systems-supermicro-amd-acquisition.html","date":"2024-11-12","support":"DIGITIMES: experts say moving AI server orders between makers is a complex process that cannot be done overnight (full date shown)."},{"url":"https://futurumgroup.com/insights/dell-launches-gb300-nvl72-offering-for-enterprise-demand/","date":"2025-07-10","support":"Futurum analysis: Dell was first to deliver GB300 NVL72 (to CoreWeave) while other major OEMs had not yet shipped GB300 (full date shown)."}]},"capabilityGap":{"level":2,"evidence":[{"url":"https://www.dell.com/en-us/dt/corporate/newsroom/announcements/detailpage.press-releases~usa~2026~09~dell-technologies-delivers-second-quarter-fiscal-2027-financial-results.htm","date":"2026-09-01","support":"Dell says it booked $60.9B of AI server orders in the quarter and exited with a record $95B AI backlog (full date shown)."},{"url":"https://www.idc.com/resource-center/press-releases/2q2026-wwserver-tracker/","date":"2026-09-10","support":"IDC 2Q26 server revenue: Dell $22.2B, Supermicro $10.2B, Lenovo $8.4B, HPE $5.9B, IEIT $4.0B, ODM Direct $89.7B (vendor table; full date shown)."}]}}}},"b":{"schema":"5lc-index-rating/1","key":"infrastructure-servers-systems","entityId":"dell","rater":"b","model":"claude-sonnet-5-5","ratedAt":"2026-10-22","functions":{"gpu-systems":{"alternatives":{"level":1,"evidence":[{"url":"https://investor.nvidia.com/news/press-release-details/2025/NVIDIA-Blackwell-Ultra-AI-Factory-Platform-Paves-Way-for-Age-of-AI-Reasoning/","date":"2025-03-18","support":"NVIDIA names 16 system makers (Dell, HPE, Lenovo, Supermicro, Cisco, ASUS, Foxconn, GIGABYTE, Inventec, Pegatron, QCT, Wistron, Wiwynn and others) expected to deliver Blackwell Ultra servers from 2H 2025, so the same GPU platform has many qualified sellers. (full date shown)"},{"url":"https://iconnect007.com/article/151595/worldwide-server-market-revenue-hits-record-1663-billion-in-q2-2026/","date":"2026-09-15","support":"IDC 2Q26 server revenue shares: Dell 13.4%, Supermicro 6.1%, Lenovo 5.1%, HPE 3.5%, IEIT 2.4%, ODM Direct vendors 53.9%; branded OEMs gaining share from ODM Direct (reprint of the IDC release). (full date shown)"}]},"switchingTime":{"level":2,"evidence":[{"url":"https://www.businesstoday.com.tw/article/category/183008/post/202411190033/","date":"2024-11-19","support":"CoreWeave's earlier H-series orders went largely to Supermicro and Gigabyte, but its first GB200 NVL72 racks came from Dell, shipped Nov 2024: a large buyer re-sourced within a product generation. (full date shown)"},{"url":"https://finance.technews.tw/2026/09/10/morgan-stanley-is-bullish-on-nvidias-rack-server-shipments-expecting-them-to-double-this-year/","date":"2026-09-10","support":"Morgan Stanley: NVL72 racks about 26,500 in 2025 and 80,000-85,000 expected in 2026; makers outside the big three go from 7% to 13% share. Annual NVL72 output is forecast to roughly triple in a year, so replacing Dell's volume would take months rather than years (inference; no direct measurement found). (full date shown)"}]},"capabilityGap":{"level":2,"evidence":[{"url":"https://www.dell.com/en-us/dt/corporate/newsroom/announcements/detailpage.press-releases~usa~2026~09~dell-technologies-delivers-second-quarter-fiscal-2027-financial-results.htm","date":"2026-09-01","support":"Dell reports AI-optimized server revenue of $16.4 billion for the quarter (+100%) and a $95 billion backlog; no other branded maker ships that volume. (full date shown)"},{"url":"https://s204.q4cdn.com/707617056/files/doc_financials/2026/q4/V2/FQ426-SMCI-Prepared-Remarks.pdf","date":"2026-08-11","support":"Supermicro's Q4 FY26 revenue was $11.1 billion with AI solutions about 60% (roughly $6.7 billion), well under Dell's AI volume. (full date shown)"},{"url":"https://investors.hpe.com/~/media/Files/H/HP-Enterprise-IR/documents/q3-2026/q3-2026-transcript.pdf","date":"2026-09-02","support":"HPE's AI systems revenue was almost $1.6 billion in its quarter, so the next branded makers fall short of Dell's scale; the large ODMs narrow the gap, hence small rather than material. (full date shown)"}]}}}}},"wistron":{"a":{"schema":"5lc-index-rating/1","key":"infrastructure-servers-systems","entityId":"wistron","rater":"a","model":"claude-opus-5-5","ratedAt":"2026-10-22","functions":{"gpu-systems":{"alternatives":{"level":2,"evidence":[{"url":"https://nvidianews.nvidia.com/news/nvidia-blackwell-ultra-ai-factory-platform-paves-way-for-age-of-ai-reasoning","date":"2025-03-18","support":"NVIDIA names Cisco, Dell, HPE, Lenovo, Supermicro plus Aivres, ASRock Rack, ASUS, Eviden, Foxconn, GIGABYTE, Inventec, Pegatron, QCT, Wistron, Wiwynn as Blackwell Ultra system makers (full date shown)."},{"url":"https://www.sec.gov/Archives/edgar/data/0001375365/000137536526000022/smci-20260630.htm","date":"2026-08-31","support":"Supermicro's 10-K (Item 1, Competition) names Cisco, Dell, HPE, Lenovo and ODMs Foxconn, Quanta, Wiwynn as rivals; date is the EDGAR filing timestamp."}]},"switchingTime":{"level":2,"evidence":[{"url":"https://www.digitimes.com/news/a20241112PD202/ai-server-nvidia-zt-systems-supermicro-amd-acquisition.html","date":"2024-11-12","support":"DIGITIMES: experts say moving AI server orders between makers is a complex process that cannot be done overnight (full date shown)."},{"url":"https://www.taipeitimes.com/News/biz/archives/2026/05/15/2003857344","date":"2026-05-15","support":"Taipei Times: Hon Hai projects 2026 AI rack shipments to more than double, with capex up over 30% (company guidance; full date shown)."}]},"capabilityGap":{"level":2,"evidence":[{"url":"https://focustaiwan.tw/business/202608120019","date":"2026-08-12","support":"Focus Taiwan: Hon Hai says it holds over 40% of the global AI server market and aims for 50% (company claim; full date shown)."},{"url":"https://www.taipeitimes.com/News/biz/archives/2026/05/15/2003857345","date":"2026-05-15","support":"Taipei Times: Quanta aims to double AI-related capacity this year and expects more ASIC server shipments (company statements; full date shown)."}]}},"custom-accelerator-systems":{"alternatives":{"level":3,"evidence":[{"url":"https://www.digitimes.com/news/a20251210PD211.html","date":"2025-12-10","support":"DIGITIMES: Inventec is a key Google TPU server supplier; Quanta, Foxconn, Wiwynn, Wistron and Mitac also pursue ASIC server orders (full date shown)."},{"url":"https://wccftech.com/nvidia-key-supply-chain-partner-foxconn-has-reportedly-won-orders-for-google-tpu/","date":"2025-12-01","support":"Wccftech, citing Taiwan's Economic Daily: Foxconn reportedly received orders for Google TPU compute trays (single-source report; full date shown)."}]},"switchingTime":{"level":3,"evidence":[{"url":"https://www.sec.gov/Archives/edgar/data/1030894/000103089426000011/cls-20251231.htm","date":"2026-02-27","support":"Celestica 10-K (Item 1A): ramping new programs can take several months to over a year before production; date is EDGAR timestamp."},{"url":"https://www.digitimes.com/news/a20260623PD213.html","date":"2026-06-23","support":"DIGITIMES: AWS Trainium 3 servers moving into volume production, with the ramp possibly extending into early 2027 (full date shown)."}]},"capabilityGap":{"level":2,"evidence":[{"url":"https://wccftech.com/nvidia-key-supply-chain-partner-foxconn-has-reportedly-won-orders-for-google-tpu/","date":"2025-12-01","support":"Wccftech, citing Taiwan's Economic Daily: Foxconn reportedly received orders for Google TPU compute trays (single-source report; full date shown)."},{"url":"https://www.taipeitimes.com/News/biz/archives/2026/05/15/2003857345","date":"2026-05-15","support":"Taipei Times: Quanta aims to double AI-related capacity this year and expects more ASIC server shipments (company statements; full date shown)."}]}}}},"b":{"schema":"5lc-index-rating/1","key":"infrastructure-servers-systems","entityId":"wistron","rater":"b","model":"claude-sonnet-5-5","ratedAt":"2026-10-22","functions":{"gpu-systems":{"alternatives":{"level":1,"evidence":[{"url":"https://investor.nvidia.com/news/press-release-details/2025/NVIDIA-Blackwell-Ultra-AI-Factory-Platform-Paves-Way-for-Age-of-AI-Reasoning/","date":"2025-03-18","support":"Wistron and Wiwynn are separately named among 16 makers expected to deliver Blackwell Ultra servers, with Dell, HPE, Lenovo, Supermicro, Foxconn, QCT and others. (full date shown)"},{"url":"https://finance.technews.tw/2026/09/10/morgan-stanley-is-bullish-on-nvidias-rack-server-shipments-expecting-them-to-double-this-year/","date":"2026-09-10","support":"Morgan Stanley: Wistron (with Wiwynn) holds 22% of NVL72 rack supply in both 2025 and 2026, against Foxconn 39%, Quanta 26% and others 13% in 2026. (full date shown)"}]},"switchingTime":{"level":2,"evidence":[{"url":"https://finance.technews.tw/2026/09/10/morgan-stanley-is-bullish-on-nvidias-rack-server-shipments-expecting-them-to-double-this-year/","date":"2026-09-10","support":"About 17,900 racks forecast for 2026 and Aug 2026 output of 2,300-2,400 rack equivalents; replacing this volume needs other makers to add capacity over several quarters (inference). (full date shown)"},{"url":"https://www.businesstoday.com.tw/article/category/183008/post/202411190033/","date":"2024-11-19","support":"Wistron is also an important contract manufacturer for Dell and Supermicro (not credited here), so its capacity is already in demand and is not idle. (full date shown)"}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://finance.technews.tw/2026/09/10/morgan-stanley-is-bullish-on-nvidias-rack-server-shipments-expecting-them-to-double-this-year/","date":"2026-09-10","support":"Foxconn (39%) and Quanta (26%) are larger or equal on 2026 rack output, so the best alternative does not fall short. (full date shown)"}]}},"custom-accelerator-systems":{"alternatives":{"level":3,"evidence":[{"url":"https://money.udn.com/money/story/5607/9189980","date":"2025-12-08","support":"Morgan Stanley's supply-chain survey (via UDN) names Wiwynn as the rack-assembly partner for AWS Trainium3, with Gold Circuit as main board supplier; no other rack assembler is named. (full date shown)"},{"url":"https://www.digitimes.com/news/a20251210PD211.html","date":"2025-12-10","support":"DIGITIMES lists Quanta, Foxconn, Wiwynn, Wistron, Inventec and Mitac as ODMs pursuing ASIC server orders. (full date shown)"}]},"switchingTime":{"level":3,"evidence":[{"url":"https://money.udn.com/money/story/5607/9189980","date":"2025-12-08","support":"Trainium3 boards ramp in Q2 2026 and racks in Q3 2026 (Morgan Stanley); Morgan Stanley puts Wiwynn's 2026 AWS-related revenue near NT$463 billion, so the program is large and mid-ramp. (full date shown)"},{"url":"https://www.theregister.com/2025/12/07/trainium3_all_nvidia_nvl72_mold/","date":"2025-12-07","support":"AWS revealed the Trainium3 UltraServer rack in Dec 2025; the design is MGX-style, helping others to follow in time, but not a drop-in replacement for an existing supplier's ramp. (full date shown)"}]},"capabilityGap":{"level":2,"evidence":[{"url":"https://money.udn.com/money/story/5607/9189980","date":"2025-12-08","support":"Morgan Stanley names Wiwynn alone for Trainium3 rack assembly and estimates 1.2-1.8 million Trainium3 chips shipped through it in 2026; no other assembler is named for this program. (full date shown)"},{"url":"https://www.theregister.com/2025/12/07/trainium3_all_nvidia_nvl72_mold/","date":"2025-12-07","support":"Racks follow an MGX-style layout like Nvidia's, so other rack builders have the base skills; the gap is program experience and volume, not technology (inference). (full date shown)"}]}}}}},"super-micro":{"a":{"schema":"5lc-index-rating/1","key":"infrastructure-servers-systems","entityId":"super-micro","rater":"a","model":"claude-opus-5-5","ratedAt":"2026-10-22","functions":{"gpu-systems":{"alternatives":{"level":1,"evidence":[{"url":"https://www.sec.gov/Archives/edgar/data/0001375365/000137536526000022/smci-20260630.htm","date":"2026-08-31","support":"Supermicro's 10-K (Item 1, Competition) names Cisco, Dell, HPE, Lenovo and ODMs Foxconn, Quanta, Wiwynn as rivals; date is the EDGAR filing timestamp."},{"url":"https://nvidianews.nvidia.com/news/nvidia-blackwell-ultra-ai-factory-platform-paves-way-for-age-of-ai-reasoning","date":"2025-03-18","support":"NVIDIA names Cisco, Dell, HPE, Lenovo, Supermicro plus Aivres, ASRock Rack, ASUS, Eviden, Foxconn, GIGABYTE, Inventec, Pegatron, QCT, Wistron, Wiwynn as Blackwell Ultra system makers (full date shown)."}]},"switchingTime":{"level":2,"evidence":[{"url":"https://www.digitimes.com/news/a20241112PD202/ai-server-nvidia-zt-systems-supermicro-amd-acquisition.html","date":"2024-11-12","support":"DIGITIMES: experts say moving AI server orders between makers is a complex process that cannot be done overnight (full date shown)."},{"url":"https://www.sec.gov/Archives/edgar/data/0001375365/000137536526000022/smci-20260630.htm","date":"2026-08-31","support":"Supermicro says it sells on purchase orders that customers may cancel or defer on short notice (Item 1A); no long-term lock-in; date is EDGAR timestamp."}]},"capabilityGap":{"level":2,"evidence":[{"url":"https://www.idc.com/resource-center/press-releases/2q2026-wwserver-tracker/","date":"2026-09-10","support":"IDC 2Q26 server revenue: Dell $22.2B, Supermicro $10.2B, Lenovo $8.4B, HPE $5.9B, IEIT $4.0B, ODM Direct $89.7B (vendor table; full date shown)."},{"url":"https://www.dell.com/en-us/dt/corporate/newsroom/announcements/detailpage.press-releases~usa~2026~09~dell-technologies-delivers-second-quarter-fiscal-2027-financial-results.htm","date":"2026-09-01","support":"Dell says it booked $60.9B of AI server orders in the quarter and exited with a record $95B AI backlog (full date shown)."}]}}}},"b":{"schema":"5lc-index-rating/1","key":"infrastructure-servers-systems","entityId":"super-micro","rater":"b","model":"claude-sonnet-5-5","ratedAt":"2026-10-22","functions":{"gpu-systems":{"alternatives":{"level":1,"evidence":[{"url":"https://www.sec.gov/Archives/edgar/data/0001375365/000137536526000022/smci-20260630.htm","date":"2026-08-31","support":"Supermicro's own 10-K names Cisco, Dell, HPE and Lenovo plus ODMs Foxconn, Quanta and Wiwynn as principal competitors (Item 1, Competition, p.7). (full date shown)"},{"url":"https://investor.nvidia.com/news/press-release-details/2025/NVIDIA-Blackwell-Ultra-AI-Factory-Platform-Paves-Way-for-Age-of-AI-Reasoning/","date":"2025-03-18","support":"NVIDIA names 16 system makers (Dell, HPE, Lenovo, Supermicro, Cisco, ASUS, Foxconn, GIGABYTE, Inventec, Pegatron, QCT, Wistron, Wiwynn and others) expected to deliver Blackwell Ultra servers from 2H 2025, so the same GPU platform has many qualified sellers. (full date shown)"}]},"switchingTime":{"level":2,"evidence":[{"url":"https://s204.q4cdn.com/707617056/files/doc_financials/2026/q4/V2/FQ426-SMCI-Prepared-Remarks.pdf","date":"2026-08-11","support":"Company says FY26 had 9 customers above $1 billion and one data-centre/CSP customer at 28% of revenue; buyers of this size can move whole programs, but replacing a business with FY26 revenue of $39.1 billion (+78%) takes quarters (inference). (full date shown)"},{"url":"https://www.businesstoday.com.tw/article/category/183008/post/202411190033/","date":"2024-11-19","support":"CoreWeave's earlier H-series orders went largely to Supermicro and Gigabyte; its first GB200 NVL72 racks went to Dell, showing buyers re-source between generations. (full date shown)"}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://s204.q4cdn.com/707617056/files/doc_financials/2026/q4/V2/FQ426-SMCI-Prepared-Remarks.pdf","date":"2026-08-11","support":"Company says its capacity is on track to exceed 6,000 racks a month, including over 3,000 direct liquid-cooled racks (company claim). (full date shown)"},{"url":"https://investors.coreweave.com/news/news-details/2024/CoreWeave-and-Dell-Technologies-Extend-Relationship-to-Deliver-AI-at-Scale/default.aspx","date":"2024-12-09","support":"Dell says it was first to ship GB200 NVL72 racks (liquid-cooled IR7000 racks) to CoreWeave, so liquid-cooled rack-scale supply is not unique to Supermicro (company release). (full date shown)"}]}}}}},"inspur":{"a":{"schema":"5lc-index-rating/1","key":"infrastructure-servers-systems","entityId":"inspur","rater":"a","model":"claude-opus-5-5","ratedAt":"2026-10-22","functions":{"gpu-systems":{"alternatives":{"level":1,"evidence":[{"url":"https://nvidianews.nvidia.com/news/nvidia-blackwell-ultra-ai-factory-platform-paves-way-for-age-of-ai-reasoning","date":"2025-03-18","support":"NVIDIA names Cisco, Dell, HPE, Lenovo, Supermicro plus Aivres, ASRock Rack, ASUS, Eviden, Foxconn, GIGABYTE, Inventec, Pegatron, QCT, Wistron, Wiwynn as Blackwell Ultra system makers (full date shown)."},{"url":"https://www.sec.gov/Archives/edgar/data/0001375365/000137536526000022/smci-20260630.htm","date":"2026-08-31","support":"Supermicro's 10-K (Item 1, Competition) names Cisco, Dell, HPE, Lenovo and ODMs Foxconn, Quanta, Wiwynn as rivals; date is the EDGAR filing timestamp."}]},"switchingTime":{"level":2,"evidence":[{"url":"https://www.digitimes.com/news/a20241112PD202/ai-server-nvidia-zt-systems-supermicro-amd-acquisition.html","date":"2024-11-12","support":"DIGITIMES: experts say moving AI server orders between makers is a complex process that cannot be done overnight (full date shown)."}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://www.idc.com/resource-center/press-releases/2q2026-wwserver-tracker/","date":"2026-09-10","support":"IDC 2Q26: IEIT Systems server revenue $4.0B, down 8.1% while the market grew 52%; larger rivals Dell, Supermicro, Lenovo, HPE (vendor table)."}]}},"china":{"alternatives":{"level":2,"evidence":[{"url":"https://www.idc.com/?p=104194","date":"2025-10-08","support":"IDC China H1 2025 accelerated servers: Inspur, H3C, Lenovo top three by revenue near 50%; by units Inspur, H3C, Nettrix about 43% (page date)."},{"url":"https://www.idc.com/?p=104362","date":"2025-03-30","support":"IDC China 2024 accelerated servers: Inspur, Nettrix, H3C top three by revenue over 50%; by units Inspur, Nettrix, Huawei near 55% (page date)."}]},"switchingTime":{"level":2,"evidence":[{"url":"https://www.cls.cn/detail/1678572","date":"2024-05-16","support":"Cailian Press: China Mobile's 2024-25 AI computing procurement split among seven makers, e.g. Henan Kunlun 21.05%, Huakun Zhenyu 17.54%, Baode 15.79% (full date)."},{"url":"https://m.c114.com.cn/w127-1295268.html","date":"2025-08-18","support":"C114: ZTE named winning candidate for packages 1-3 of China Mobile's 2025 AI inference equipment tender; volume reallocated by tender (full date shown)."}]},"capabilityGap":{"level":2,"evidence":[{"url":"https://www.idc.com/?p=104194","date":"2025-10-08","support":"IDC China H1 2025 accelerated servers: Inspur, H3C, Lenovo top three by revenue near 50%; by units Inspur, H3C, Nettrix about 43% (page date)."},{"url":"https://www.idc.com/?p=104362","date":"2025-03-30","support":"IDC China 2024 accelerated servers: Inspur, Nettrix, H3C top three by revenue over 50%; by units Inspur, Nettrix, Huawei near 55% (page date)."}]}}}},"b":{"schema":"5lc-index-rating/1","key":"infrastructure-servers-systems","entityId":"inspur","rater":"b","model":"claude-sonnet-5-5","ratedAt":"2026-10-22","functions":{"gpu-systems":{"alternatives":{"level":1,"evidence":[{"url":"https://iconnect007.com/article/151595/worldwide-server-market-revenue-hits-record-1663-billion-in-q2-2026/","date":"2026-09-15","support":"IDC 2Q26: IEIT Systems has 2.4% of worldwide server revenue (down 8.1% year on year) against Dell, Supermicro, Lenovo, HPE and ODM Direct. (full date shown)"},{"url":"https://investor.nvidia.com/news/press-release-details/2025/NVIDIA-Blackwell-Ultra-AI-Factory-Platform-Paves-Way-for-Age-of-AI-Reasoning/","date":"2025-03-18","support":"At least 15 other makers are named for the same NVIDIA Blackwell Ultra platforms. (full date shown)"}]},"switchingTime":{"level":1,"evidence":[{"url":"https://iconnect007.com/article/151595/worldwide-server-market-revenue-hits-record-1663-billion-in-q2-2026/","date":"2026-09-15","support":"IEIT's server revenue fell 8.1% year on year while Dell, Supermicro, Lenovo and HPE grew 46-165%, so others are carrying world growth and can re-spread IEIT's overseas AI volume (inference). (full date shown)"}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://investor.nvidia.com/news/press-release-details/2025/NVIDIA-Blackwell-Ultra-AI-Factory-Platform-Paves-Way-for-Age-of-AI-Reasoning/","date":"2025-03-18","support":"The same GPU platforms come from many Western and Taiwanese makers; I found no capability only IEIT has in GPU servers outside China. (full date shown)"}]}},"china":{"alternatives":{"level":1,"evidence":[{"url":"https://www.chinabaogao.com/jingzheng/202510/767092.html","date":"2025-10-14","support":"IDC China 1H25: Inspur, H3C and Lenovo top accelerated-server sales (about 50% together); by shipments Inspur, H3C and Nettrix (Ningchang), about 43%: roughly half of the market sits with other domestic makers. (full date shown)"},{"url":"https://eu.36kr.com/en/p/3913165545837700","date":"2026-07-27","support":"Super-node systems now come from Huawei, Sugon, H3C, Inspur, ZTE and Alibaba (Digital Intelligence Frontline via 36Kr). (full date shown)"}]},"switchingTime":{"level":2,"evidence":[{"url":"https://eu.36kr.com/en/p/3930314422533512","date":"2026-08-10","support":"Customers test each chip with their own models and compare cost; Inspur says it cut its chip-to-computing-power cycle to 6-8 months (company claim), so a replacement vendor needs months to adapt (36Kr). (full date shown)"},{"url":"https://www.chinabaogao.com/jingzheng/202510/767092.html","date":"2025-10-14","support":"Inspur ranks first by both sales and shipments in IDC China 1H25, so its volume is the largest to re-spread. (full date shown)"}]},"capabilityGap":{"level":2,"evidence":[{"url":"https://www.chinabaogao.com/jingzheng/202510/767092.html","date":"2025-10-14","support":"H3C is second on both measures and Lenovo third by sales; the best alternative is smaller than Inspur but already serves the same buyers. (full date shown)"},{"url":"https://technode.com/2026/10/14/nvidias-china-ai-chip-share-falls-to-34-as-huawei-reaches-28/","date":"2026-10-14","support":"Nvidia 34% and Huawei 28% of China AI accelerator shipments in 1H26: the binding limit is chip supply to all server makers, not Inspur's integration work. (full date shown)"}]}}}}},"huawei":{"a":{"schema":"5lc-index-rating/1","key":"infrastructure-servers-systems","entityId":"huawei","rater":"a","model":"claude-opus-5-5","ratedAt":"2026-10-22","functions":{"china":{"alternatives":{"level":2,"evidence":[{"url":"https://www.cls.cn/detail/1678572","date":"2024-05-16","support":"Cailian Press: China Mobile's 2024-25 AI computing procurement split among seven makers, e.g. Henan Kunlun 21.05%, Huakun Zhenyu 17.54%, Baode 15.79% (full date)."},{"url":"https://www.ithome.com/0/895/328.htm","date":"2025-11-06","support":"IT Home: Sugon launched scaleX640, a single-rack 640-card supernode supporting multiple accelerator brands (full date shown)."}]},"switchingTime":{"level":2,"evidence":[{"url":"https://www.cls.cn/detail/1678572","date":"2024-05-16","support":"Cailian Press: China Mobile's 2024-25 AI computing procurement split among seven makers, e.g. Henan Kunlun 21.05%, Huakun Zhenyu 17.54%, Baode 15.79% (full date)."}]},"capabilityGap":{"level":2,"evidence":[{"url":"https://kyodonewsprwire.jp/release/202509235711","date":"2025-09-23","support":"Huawei Cloud says CloudMatrix supernodes will grow from 384 to 8,192 cards, offered as cloud compute services (company statement; full date shown)."},{"url":"https://www.ithome.com/0/895/328.htm","date":"2025-11-06","support":"IT Home: Sugon launched scaleX640, a single-rack 640-card supernode supporting multiple accelerator brands (full date shown)."}]}}}},"b":{"schema":"5lc-index-rating/1","key":"infrastructure-servers-systems","entityId":"huawei","rater":"b","model":"claude-sonnet-5-5","ratedAt":"2026-10-22","functions":{"china":{"alternatives":{"level":3,"evidence":[{"url":"https://eu.36kr.com/en/p/3913165545837700","date":"2026-07-27","support":"Huawei shows a 1,024-accelerator Ascend 950 node and has over 750 commercial 384-accelerator nodes; Sugon has a 640-accelerator cabinet, Inspur 64, H3C up to 512, Alibaba 64-accelerator units, ZTE a node with no specs (36Kr). (full date shown)"},{"url":"https://www.theregister.com/2025/07/29/huawei_rackscale_boogeyman/","date":"2025-07-29","support":"The Register (Jul 2025): Huawei has little if any competition in rack-scale systems for Chinese model developers; newer entrants since then are mostly smaller scale-up domains. (full date shown)"}]},"switchingTime":{"level":3,"evidence":[{"url":"https://eu.36kr.com/en/p/3930314422533512","date":"2026-08-10","support":"Customers test each chip with their own models and compare cost; Inspur says chip adaptation now takes 6-8 months, so moving a Huawei Ascend estate to another stack means re-porting software at scale (inference, 1-2 years). (full date shown)"},{"url":"https://www.crnasia.com/news/2026/artificial-intelligence/huawei-outlines-ascend-960-roadmap-and-million-processor-ai","date":"undated","support":"Huawei says over 1,000 Atlas 900 A3 SuperPoDs are deployed, Atlas 950 is in commercial use and a 256,000-card Atlas 950 SuperCluster is being deployed (company claim). (page shows no publication date)"}]},"capabilityGap":{"level":4,"evidence":[{"url":"https://technode.com/2026/10/14/nvidias-china-ai-chip-share-falls-to-34-as-huawei-reaches-28/","date":"2026-10-14","support":"IDC via Reuters: Huawei shipped 572,000 AI accelerators in 1H26 (28% share) against T-Head 7%, Cambricon 5%, Kunlunxin 4% and Hygon 3%; Nvidia shipments fell 44% under export limits. (full date shown)"},{"url":"https://www.trendforce.com/news/2026/07/09/news-china-reportedly-to-allow-nvidia-h200-imports-but-approvals-may-be-capped-below-200k-chips-less-than-half-requested/","date":"2026-07-09","support":"Beijing is expected to approve fewer than 200,000 H200 chips, and inference buyers are directed to domestic processors including Huawei's, so the Nvidia alternative is capped. (full date shown)"}]}}}},"reconciled":{"schema":"5lc-index-rating/1","key":"infrastructure-servers-systems","entityId":"huawei","rater":"reconciled","model":"claude-opus-5-5","ratedAt":"2026-10-22","functions":{"china":{"alternatives":{"level":3,"evidence":[{"url":"https://sinolytics.de/global-business-news/blog/technology/waic-china-ai-supernodes/","date":"2026-08-10","support":"Sinolytics: 'a limited number of vendors have achieved scaled commercial deployment', naming Huawei, Sugon and Kunlunxin; many others are prototypes (full date shown)."},{"url":"https://eu.36kr.com/en/p/3913165545837700","date":"2026-07-27","support":"36Kr: Sugon (640-card node, 100,000-card cluster) and Baidu Tianchi are in commercial use; Inspur, H3C, ZTE and others are at launch or design stage (full date shown)."},{"url":"https://www.cls.cn/detail/1678572","date":"2024-05-16","support":"Cailian Press: China Mobile's AI server tender split among seven Ascend-ecosystem builders, so standard Ascend servers have many makers, credited to them (full date shown)."}]},"switchingTime":{"level":3,"evidence":[{"url":"https://eu.36kr.com/en/p/3930314422533512","date":"2026-08-10","support":"36Kr: swapping accelerators means redoing boards, power, cooling and firmware, then drivers, compilers and operator libraries; ported models often 'only can run' (section 01; full date shown)."},{"url":"https://m.c114.com.cn/w126-1314095.html","date":"2026-07-17","support":"C114: Huawei says its 384-card supernode has over 750 commercial deployments and is China's only supernode in scaled commercial use (company claim; full date shown)."}]},"capabilityGap":{"level":3,"evidence":[{"url":"https://eu.36kr.com/en/p/3913165545837700","date":"2026-07-27","support":"36Kr: only Huawei publishes a commercial node count (750+ of 384 accelerators, roughly 288,000 cards); Sugon reports one 100,000-card cluster (5LayerCake arithmetic; full date shown)."},{"url":"https://sinolytics.de/global-business-news/blog/technology/waic-china-ai-supernodes/","date":"2026-08-10","support":"Sinolytics groups Sugon and Kunlunxin with Huawei among scaled deployments, so alternatives exist at scale but behind Huawei's base (full date shown)."}]}}}}},"lenovo":{"a":{"schema":"5lc-index-rating/1","key":"infrastructure-servers-systems","entityId":"lenovo","rater":"a","model":"claude-opus-5-5","ratedAt":"2026-10-22","functions":{"gpu-systems":{"alternatives":{"level":1,"evidence":[{"url":"https://nvidianews.nvidia.com/news/nvidia-blackwell-ultra-ai-factory-platform-paves-way-for-age-of-ai-reasoning","date":"2025-03-18","support":"NVIDIA names Cisco, Dell, HPE, Lenovo, Supermicro plus Aivres, ASRock Rack, ASUS, Eviden, Foxconn, GIGABYTE, Inventec, Pegatron, QCT, Wistron, Wiwynn as Blackwell Ultra system makers (full date shown)."},{"url":"https://www.sec.gov/Archives/edgar/data/0001375365/000137536526000022/smci-20260630.htm","date":"2026-08-31","support":"Supermicro's 10-K (Item 1, Competition) names Cisco, Dell, HPE, Lenovo and ODMs Foxconn, Quanta, Wiwynn as rivals; date is the EDGAR filing timestamp."}]},"switchingTime":{"level":2,"evidence":[{"url":"https://www.digitimes.com/news/a20241112PD202/ai-server-nvidia-zt-systems-supermicro-amd-acquisition.html","date":"2024-11-12","support":"DIGITIMES: experts say moving AI server orders between makers is a complex process that cannot be done overnight (full date shown)."}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://www.idc.com/resource-center/press-releases/2q2026-wwserver-tracker/","date":"2026-09-10","support":"IDC 2Q26 server revenue: Dell $22.2B, Supermicro $10.2B, Lenovo $8.4B, HPE $5.9B, IEIT $4.0B, ODM Direct $89.7B (vendor table; full date shown)."},{"url":"https://news.lenovo.com/?p=34443","date":"2026-05-22","support":"Lenovo says annual server capacity passed 70,000 racks, including over 11,000 direct liquid-cooled AI racks (full date shown)."}]}},"china":{"alternatives":{"level":1,"evidence":[{"url":"https://www.idc.com/?p=104194","date":"2025-10-08","support":"IDC China H1 2025 accelerated servers: Inspur, H3C, Lenovo top three by revenue near 50%; by units Inspur, H3C, Nettrix about 43% (page date)."},{"url":"https://www.idc.com/?p=104362","date":"2025-03-30","support":"IDC China 2024 accelerated servers: Inspur, Nettrix, H3C top three by revenue over 50%; by units Inspur, Nettrix, Huawei near 55% (page date)."}]},"switchingTime":{"level":2,"evidence":[{"url":"https://www.cls.cn/detail/1678572","date":"2024-05-16","support":"Cailian Press: China Mobile's 2024-25 AI computing procurement split among seven makers, e.g. Henan Kunlun 21.05%, Huakun Zhenyu 17.54%, Baode 15.79% (full date)."}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://www.idc.com/?p=104194","date":"2025-10-08","support":"IDC China H1 2025 accelerated servers: Inspur, H3C, Lenovo top three by revenue near 50%; by units Inspur, H3C, Nettrix about 43% (page date)."}]}}}},"b":{"schema":"5lc-index-rating/1","key":"infrastructure-servers-systems","entityId":"lenovo","rater":"b","model":"claude-sonnet-5-5","ratedAt":"2026-10-22","functions":{"gpu-systems":{"alternatives":{"level":1,"evidence":[{"url":"https://investor.nvidia.com/news/press-release-details/2025/NVIDIA-Blackwell-Ultra-AI-Factory-Platform-Paves-Way-for-Age-of-AI-Reasoning/","date":"2025-03-18","support":"NVIDIA names 16 system makers (Dell, HPE, Lenovo, Supermicro, Cisco, ASUS, Foxconn, GIGABYTE, Inventec, Pegatron, QCT, Wistron, Wiwynn and others) expected to deliver Blackwell Ultra servers from 2H 2025, so the same GPU platform has many qualified sellers. (full date shown)"},{"url":"https://iconnect007.com/article/151595/worldwide-server-market-revenue-hits-record-1663-billion-in-q2-2026/","date":"2026-09-15","support":"IDC 2Q26: Lenovo holds 5.1% of server revenue against Dell 13.4%, Supermicro 6.1%, HPE 3.5% and ODM Direct 53.9%. (full date shown)"}]},"switchingTime":{"level":1,"evidence":[{"url":"https://doc.irasia.com/listco/hk/lenovo/interim/2027/int1q.pdf","date":"2026-08-13","support":"Lenovo gives no AI server revenue in dollars: ISG revenue was $8.5 billion in the June-2026 quarter (+98%) across all servers and storage, with an AI server pipeline of $54.0 billion. (full date shown)"},{"url":"https://www.dell.com/en-us/dt/corporate/newsroom/announcements/detailpage.press-releases~usa~2026~09~dell-technologies-delivers-second-quarter-fiscal-2027-financial-results.htm","date":"2026-09-01","support":"Dell alone reported $16.4 billion of AI-optimized server revenue in its quarter, so other sellers can take Lenovo's volume within a few months (inference). (full date shown)"}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://doc.irasia.com/listco/hk/lenovo/interim/2027/int1q.pdf","date":"2026-08-13","support":"Lenovo describes an 'ODM+' model for cloud customers and GB300 deployments, the same NVIDIA platforms its rivals sell (Q1 FY26/27 announcement). (full date shown)"},{"url":"https://investor.nvidia.com/news/press-release-details/2025/NVIDIA-Blackwell-Ultra-AI-Factory-Platform-Paves-Way-for-Age-of-AI-Reasoning/","date":"2025-03-18","support":"Lenovo is one of 16 named Blackwell Ultra server makers, so no alternative falls short on the platform. (full date shown)"}]}},"china":{"alternatives":{"level":1,"evidence":[{"url":"https://www.chinabaogao.com/jingzheng/202510/767092.html","date":"2025-10-14","support":"IDC China 1H25: Inspur, H3C and Lenovo top accelerated-server sales (about 50% together); by shipments Inspur, H3C and Nettrix (Ningchang), about 43%: roughly half of the market sits with other domestic makers. (full date shown)"},{"url":"https://www.trendforce.com/news/2026/07/09/news-china-reportedly-to-allow-nvidia-h200-imports-but-approvals-may-be-capped-below-200k-chips-less-than-half-requested/","date":"2026-07-09","support":"Lenovo and Foxconn were cleared to handle Nvidia H200 sales to China; chips, not server makers, are the scarce item. (full date shown)"}]},"switchingTime":{"level":1,"evidence":[{"url":"https://www.chinabaogao.com/jingzheng/202510/767092.html","date":"2025-10-14","support":"Lenovo is third by sales in IDC China 1H25 but not in the top three by shipments, which Nettrix holds; the top three together hold only about half of the market. (full date shown)"},{"url":"https://doc.irasia.com/listco/hk/lenovo/interim/2027/int1q.pdf","date":"2026-08-13","support":"Lenovo says its group revenue in China grew 25% and its China Enterprise & SMB business had record first-quarter revenue; it gives no China AI server dollars. (full date shown)"}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://www.trendforce.com/news/2026/07/09/news-china-reportedly-to-allow-nvidia-h200-imports-but-approvals-may-be-capped-below-200k-chips-less-than-half-requested/","date":"2026-07-09","support":"Beijing is expected to approve fewer than 200,000 H200 chips; for inference buyers are told to prefer domestic processors: every server maker faces the same chip limit, so no alternative falls short. (full date shown)"}]}}}}},"gigabyte":{"a":{"schema":"5lc-index-rating/1","key":"infrastructure-servers-systems","entityId":"gigabyte","rater":"a","model":"claude-opus-5-5","ratedAt":"2026-10-22","functions":{"gpu-systems":{"alternatives":{"level":1,"evidence":[{"url":"https://nvidianews.nvidia.com/news/nvidia-blackwell-ultra-ai-factory-platform-paves-way-for-age-of-ai-reasoning","date":"2025-03-18","support":"NVIDIA names Cisco, Dell, HPE, Lenovo, Supermicro plus Aivres, ASRock Rack, ASUS, Eviden, Foxconn, GIGABYTE, Inventec, Pegatron, QCT, Wistron, Wiwynn as Blackwell Ultra system makers (full date shown)."},{"url":"https://www.sec.gov/Archives/edgar/data/0001375365/000137536526000022/smci-20260630.htm","date":"2026-08-31","support":"Supermicro's 10-K (Item 1, Competition) names Cisco, Dell, HPE, Lenovo and ODMs Foxconn, Quanta, Wiwynn as rivals; date is the EDGAR filing timestamp."}]},"switchingTime":{"level":2,"evidence":[{"url":"https://www.digitimes.com/news/a20241112PD202/ai-server-nvidia-zt-systems-supermicro-amd-acquisition.html","date":"2024-11-12","support":"DIGITIMES: experts say moving AI server orders between makers is a complex process that cannot be done overnight (full date shown)."}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://www.idc.com/resource-center/press-releases/2q2026-wwserver-tracker/","date":"2026-09-10","support":"IDC 2Q26 server revenue: Dell $22.2B, Supermicro $10.2B, Lenovo $8.4B, HPE $5.9B, IEIT $4.0B, ODM Direct $89.7B (vendor table; full date shown)."}]}}}},"b":{"schema":"5lc-index-rating/1","key":"infrastructure-servers-systems","entityId":"gigabyte","rater":"b","model":"claude-sonnet-5-5","ratedAt":"2026-10-22","functions":{"gpu-systems":{"alternatives":{"level":1,"evidence":[{"url":"https://investor.nvidia.com/news/press-release-details/2025/NVIDIA-Blackwell-Ultra-AI-Factory-Platform-Paves-Way-for-Age-of-AI-Reasoning/","date":"2025-03-18","support":"GIGABYTE is one of 16 named Blackwell Ultra server makers; the platform is sold by Dell, HPE, Lenovo, Supermicro, Foxconn and others. (full date shown)"},{"url":"https://www.businesstoday.com.tw/article/category/183008/post/202411190033/","date":"2024-11-19","support":"CoreWeave's earlier H-series orders went largely to Supermicro and Gigabyte; its GB200 NVL72 order went to Dell, so buyers do have substitutes. (full date shown)"}]},"switchingTime":{"level":1,"evidence":[{"url":"https://www.taipeitimes.com/News/biz/archives/2026/08/18/2003862657","date":"2026-08-18","support":"H1 2026 revenue was NT$249.27 billion with Q2 gross margin 9.62%; the volume is small against the large ODMs and branded vendors, so others can absorb it in weeks to a few months (inference). (full date shown)"}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://www.taipeitimes.com/News/biz/archives/2026/08/18/2003862657","date":"2026-08-18","support":"Gigabyte is adjusting its mix between GB-series and HGX servers, i.e. selling the same NVIDIA platforms as peers; nothing indicates a unique capability. (full date shown)"}]}}}}},"asus":{"a":{"schema":"5lc-index-rating/1","key":"infrastructure-servers-systems","entityId":"asus","rater":"a","model":"claude-opus-5-5","ratedAt":"2026-10-22","functions":{"gpu-systems":{"alternatives":{"level":1,"evidence":[{"url":"https://nvidianews.nvidia.com/news/nvidia-blackwell-ultra-ai-factory-platform-paves-way-for-age-of-ai-reasoning","date":"2025-03-18","support":"NVIDIA names Cisco, Dell, HPE, Lenovo, Supermicro plus Aivres, ASRock Rack, ASUS, Eviden, Foxconn, GIGABYTE, Inventec, Pegatron, QCT, Wistron, Wiwynn as Blackwell Ultra system makers (full date shown)."},{"url":"https://www.sec.gov/Archives/edgar/data/0001375365/000137536526000022/smci-20260630.htm","date":"2026-08-31","support":"Supermicro's 10-K (Item 1, Competition) names Cisco, Dell, HPE, Lenovo and ODMs Foxconn, Quanta, Wiwynn as rivals; date is the EDGAR filing timestamp."}]},"switchingTime":{"level":2,"evidence":[{"url":"https://www.digitimes.com/news/a20241112PD202/ai-server-nvidia-zt-systems-supermicro-amd-acquisition.html","date":"2024-11-12","support":"DIGITIMES: experts say moving AI server orders between makers is a complex process that cannot be done overnight (full date shown)."}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://www.idc.com/resource-center/press-releases/2q2026-wwserver-tracker/","date":"2026-09-10","support":"IDC 2Q26 server revenue: Dell $22.2B, Supermicro $10.2B, Lenovo $8.4B, HPE $5.9B, IEIT $4.0B, ODM Direct $89.7B (vendor table; full date shown)."}]}}}},"b":{"schema":"5lc-index-rating/1","key":"infrastructure-servers-systems","entityId":"asus","rater":"b","model":"claude-sonnet-5-5","ratedAt":"2026-10-22","functions":{"gpu-systems":{"alternatives":{"level":1,"evidence":[{"url":"https://investor.nvidia.com/news/press-release-details/2025/NVIDIA-Blackwell-Ultra-AI-Factory-Platform-Paves-Way-for-Age-of-AI-Reasoning/","date":"2025-03-18","support":"ASUS is one of 16 named Blackwell Ultra server makers; Dell, HPE, Lenovo, Supermicro, Foxconn, GIGABYTE and others sell the same platform. (full date shown)"},{"url":"https://iconnect007.com/article/151595/worldwide-server-market-revenue-hits-record-1663-billion-in-q2-2026/","date":"2026-09-15","support":"IDC 2Q26 shows many sellers above ASUS in size (Dell 13.4%, Supermicro 6.1%, Lenovo 5.1%, HPE 3.5%, ODM Direct 53.9%). (full date shown)"}]},"switchingTime":{"level":1,"evidence":[{"url":"https://www.taiwannews.com.tw/news/6420610","date":"2026-08-13","support":"ASUS expects NT$180-200 billion of 2026 server revenue, with AI servers about 90% of its server business; this is small next to Dell's $16.4 billion a quarter, so others can absorb it quickly (inference). (full date shown)"}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://www.taiwannews.com.tw/news/6420610","date":"2026-08-13","support":"ASUS's lineup is NVIDIA HGX systems and GB300 NVL72 racks, with Rubin and a liquid-cooled rack shown at Computex: the same roadmap as peers. (full date shown)"}]}}}}},"h3c":{"a":{"schema":"5lc-index-rating/1","key":"infrastructure-servers-systems","entityId":"h3c","rater":"a","model":"claude-opus-5-5","ratedAt":"2026-10-22","functions":{"china":{"alternatives":{"level":1,"evidence":[{"url":"https://www.idc.com/?p=104194","date":"2025-10-08","support":"IDC China H1 2025 accelerated servers: Inspur, H3C, Lenovo top three by revenue near 50%; by units Inspur, H3C, Nettrix about 43% (page date)."},{"url":"https://www.idc.com/?p=104362","date":"2025-03-30","support":"IDC China 2024 accelerated servers: Inspur, Nettrix, H3C top three by revenue over 50%; by units Inspur, Nettrix, Huawei near 55% (page date)."}]},"switchingTime":{"level":2,"evidence":[{"url":"https://www.cls.cn/detail/1678572","date":"2024-05-16","support":"Cailian Press: China Mobile's 2024-25 AI computing procurement split among seven makers, e.g. Henan Kunlun 21.05%, Huakun Zhenyu 17.54%, Baode 15.79% (full date)."},{"url":"https://m.c114.com.cn/w127-1295268.html","date":"2025-08-18","support":"C114: ZTE named winning candidate for packages 1-3 of China Mobile's 2025 AI inference equipment tender; volume reallocated by tender (full date shown)."}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://www.idc.com/?p=104194","date":"2025-10-08","support":"IDC China H1 2025 accelerated servers: Inspur, H3C, Lenovo top three by revenue near 50%; by units Inspur, H3C, Nettrix about 43% (page date)."}]}}}},"b":{"schema":"5lc-index-rating/1","key":"infrastructure-servers-systems","entityId":"h3c","rater":"b","model":"claude-sonnet-5-5","ratedAt":"2026-10-22","functions":{"china":{"alternatives":{"level":1,"evidence":[{"url":"https://www.chinabaogao.com/jingzheng/202510/767092.html","date":"2025-10-14","support":"IDC China 1H25: Inspur, H3C and Lenovo top accelerated-server sales (about 50% together); by shipments Inspur, H3C and Nettrix (Ningchang), about 43%: roughly half of the market sits with other domestic makers. (full date shown)"},{"url":"https://eu.36kr.com/en/p/3913165545837700","date":"2026-07-27","support":"H3C offers super-node designs up to 512 accelerators; Huawei, Sugon, Inspur, ZTE and Alibaba offer competing nodes (36Kr). (full date shown)"}]},"switchingTime":{"level":2,"evidence":[{"url":"https://eu.36kr.com/en/p/3930314422533512","date":"2026-08-10","support":"Customers test each chip with their own models; chip adaptation cycles are months (Inspur says 6-8 months); H3C's UniPoD S80000 scales to 1,024 cards (36Kr). (full date shown)"},{"url":"https://m.gelonghui.com/news/5306718","date":"2026","support":"H3C H1 2026 revenue RMB 49.98 billion (+37.3%); IDC Q1 2026 puts its x86 server share at 14.9%, second (Ping An Securities note). (year only: page shows 09-04 with no year; its subject, the 2026 half-year report, fixes 2026)"}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://www.chinabaogao.com/jingzheng/202510/767092.html","date":"2025-10-14","support":"Inspur ranks first on both sales and shipments and H3C second on both, so H3C's best alternative (Inspur) is not smaller. (full date shown)"}]}}}}},"xfusion":{"a":{"schema":"5lc-index-rating/1","key":"infrastructure-servers-systems","entityId":"xfusion","rater":"a","model":"claude-opus-5-5","ratedAt":"2026-10-22","functions":{"china":{"alternatives":{"level":1,"evidence":[{"url":"https://www.idc.com/?p=104194","date":"2025-10-08","support":"IDC China H1 2025 accelerated servers: Inspur, H3C, Lenovo top three by revenue near 50%; by units Inspur, H3C, Nettrix about 43% (page date)."},{"url":"https://www.cls.cn/detail/1678572","date":"2024-05-16","support":"Cailian Press: China Mobile's 2024-25 AI computing procurement split among seven makers, e.g. Henan Kunlun 21.05%, Huakun Zhenyu 17.54%, Baode 15.79% (full date)."}]},"switchingTime":{"level":2,"evidence":[{"url":"https://www.cls.cn/detail/1678572","date":"2024-05-16","support":"Cailian Press: China Mobile's 2024-25 AI computing procurement split among seven makers, e.g. Henan Kunlun 21.05%, Huakun Zhenyu 17.54%, Baode 15.79% (full date)."}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://www.idc.com/?p=104194","date":"2025-10-08","support":"IDC China H1 2025 accelerated servers: Inspur, H3C, Lenovo top three by revenue near 50%; by units Inspur, H3C, Nettrix about 43% (page date)."}]}}}},"b":{"schema":"5lc-index-rating/1","key":"infrastructure-servers-systems","entityId":"xfusion","rater":"b","model":"claude-sonnet-5-5","ratedAt":"2026-10-22","functions":{"china":{"alternatives":{"level":1,"evidence":[{"url":"https://www.chinabaogao.com/jingzheng/202510/767092.html","date":"2025-10-14","support":"IDC China 1H25: Inspur, H3C and Lenovo top accelerated-server sales (about 50% together); by shipments Inspur, H3C and Nettrix (Ningchang), about 43%: roughly half of the market sits with other domestic makers. (full date shown)"},{"url":"https://www.yicai.com/news/102154509.html","date":"2024-06-18","support":"Huakun Zhenyu and xFusion are described as the top two whole-machine vendors for Huawei's Kunpeng and Ascend lines, with Digital China another early partner (Yicai, 2024). (full date shown)"}]},"switchingTime":{"level":2,"evidence":[{"url":"https://www.digitimes.com/news/a20260521VL224.html","date":"2026-05-21","support":"DIGITIMES reports xFusion's 2025 revenue at about CNY 58.2 billion (about US$8.6 billion, all products); this scale suggests its AI volume may exceed the 5% placeholder, so replacement takes several quarters. (full date shown)"},{"url":"https://eu.36kr.com/en/p/3930314422533512","date":"2026-08-10","support":"Customers test each chip with their own models; chip adaptation takes months (Inspur says 6-8 months) (36Kr). (full date shown)"}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://www.chinabaogao.com/jingzheng/202510/767092.html","date":"2025-10-14","support":"IDC China's top three by sales (Inspur, H3C, Lenovo) and by shipments (Inspur, H3C, Nettrix) do not include xFusion, so on IDC's AI measure larger alternatives exist. (full date shown)"}]}}}}},"hpe":{"a":{"schema":"5lc-index-rating/1","key":"infrastructure-servers-systems","entityId":"hpe","rater":"a","model":"claude-opus-5-5","ratedAt":"2026-10-22","functions":{"gpu-systems":{"alternatives":{"level":1,"evidence":[{"url":"https://nvidianews.nvidia.com/news/nvidia-blackwell-ultra-ai-factory-platform-paves-way-for-age-of-ai-reasoning","date":"2025-03-18","support":"NVIDIA names Cisco, Dell, HPE, Lenovo, Supermicro plus Aivres, ASRock Rack, ASUS, Eviden, Foxconn, GIGABYTE, Inventec, Pegatron, QCT, Wistron, Wiwynn as Blackwell Ultra system makers (full date shown)."},{"url":"https://www.sec.gov/Archives/edgar/data/0001375365/000137536526000022/smci-20260630.htm","date":"2026-08-31","support":"Supermicro's 10-K (Item 1, Competition) names Cisco, Dell, HPE, Lenovo and ODMs Foxconn, Quanta, Wiwynn as rivals; date is the EDGAR filing timestamp."}]},"switchingTime":{"level":2,"evidence":[{"url":"https://www.digitimes.com/news/a20241112PD202/ai-server-nvidia-zt-systems-supermicro-amd-acquisition.html","date":"2024-11-12","support":"DIGITIMES: experts say moving AI server orders between makers is a complex process that cannot be done overnight (full date shown)."}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://www.idc.com/resource-center/press-releases/2q2026-wwserver-tracker/","date":"2026-09-10","support":"IDC 2Q26 server revenue: Dell $22.2B, Supermicro $10.2B, Lenovo $8.4B, HPE $5.9B, IEIT $4.0B, ODM Direct $89.7B (vendor table; full date shown)."}]}}}},"b":{"schema":"5lc-index-rating/1","key":"infrastructure-servers-systems","entityId":"hpe","rater":"b","model":"claude-sonnet-5-5","ratedAt":"2026-10-22","functions":{"gpu-systems":{"alternatives":{"level":1,"evidence":[{"url":"https://investor.nvidia.com/news/press-release-details/2025/NVIDIA-Blackwell-Ultra-AI-Factory-Platform-Paves-Way-for-Age-of-AI-Reasoning/","date":"2025-03-18","support":"NVIDIA names 16 system makers (Dell, HPE, Lenovo, Supermicro, Cisco, ASUS, Foxconn, GIGABYTE, Inventec, Pegatron, QCT, Wistron, Wiwynn and others) expected to deliver Blackwell Ultra servers from 2H 2025, so the same GPU platform has many qualified sellers. (full date shown)"},{"url":"https://www.sec.gov/Archives/edgar/data/0001375365/000137536526000022/smci-20260630.htm","date":"2026-08-31","support":"Supermicro's 10-K names Cisco, Dell, HPE and Lenovo plus ODMs Foxconn, Quanta and Wiwynn as principal competitors (Item 1, Competition). (full date shown)"}]},"switchingTime":{"level":1,"evidence":[{"url":"https://investors.hpe.com/~/media/Files/H/HP-Enterprise-IR/documents/q3-2026/q3-2026-transcript.pdf","date":"2026-09-02","support":"HPE says AI systems revenue was almost $1.6 billion in the quarter and that supply constraints limit what it can fulfil (CFO/CEO remarks, Q3 FY26 call). (full date shown)"},{"url":"https://www.dell.com/en-us/dt/corporate/newsroom/announcements/detailpage.press-releases~usa~2026~09~dell-technologies-delivers-second-quarter-fiscal-2027-financial-results.htm","date":"2026-09-01","support":"Dell's AI-optimized server revenue was $16.4 billion in its latest quarter, roughly ten times HPE's AI systems revenue, so other makers can absorb HPE's volume quickly (inference). (full date shown)"}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://investor.nvidia.com/news/press-release-details/2025/NVIDIA-Blackwell-Ultra-AI-Factory-Platform-Paves-Way-for-Age-of-AI-Reasoning/","date":"2025-03-18","support":"The same Blackwell Ultra platforms are offered by 15 other named makers; I found no HPE-only capability in GPU servers sold outside China. (full date shown)"},{"url":"https://investors.hpe.com/~/media/Files/H/HP-Enterprise-IR/documents/q3-2026/q3-2026-transcript.pdf","date":"2026-09-02","support":"HPE's cited advantages are demand from enterprises, neo-clouds and sovereign buyers and financing; these are customer-relationship advantages, not a capability others lack (reading of the call). (full date shown)"}]}}}}},"nvidia":{"a":{"schema":"5lc-index-rating/1","key":"infrastructure-servers-systems","entityId":"nvidia","rater":"a","model":"claude-opus-5-5","ratedAt":"2026-10-22","functions":{"gpu-systems":{"alternatives":{"level":1,"evidence":[{"url":"https://nvidianews.nvidia.com/news/nvidia-blackwell-ultra-ai-factory-platform-paves-way-for-age-of-ai-reasoning","date":"2025-03-18","support":"NVIDIA sells DGX GB300 on the GB300 NVL72 rack design, while the same release lists 16 partner makers of Blackwell Ultra systems (full date shown)."}]},"switchingTime":{"level":2,"evidence":[{"url":"https://www.digitimes.com/news/a20241112PD202/ai-server-nvidia-zt-systems-supermicro-amd-acquisition.html","date":"2024-11-12","support":"DIGITIMES: experts say moving AI server orders between makers is a complex process that cannot be done overnight (full date shown)."}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://www.idc.com/resource-center/press-releases/2q2026-wwserver-tracker/","date":"2026-09-10","support":"IDC 2Q26 server revenue: Dell $22.2B, Supermicro $10.2B, Lenovo $8.4B, HPE $5.9B, IEIT $4.0B, ODM Direct $89.7B (vendor table; full date shown)."},{"url":"https://nvidianews.nvidia.com/news/nvidia-blackwell-ultra-ai-factory-platform-paves-way-for-age-of-ai-reasoning","date":"2025-03-18","support":"NVIDIA sells DGX GB300 on the GB300 NVL72 rack design, while the same release lists 16 partner makers of Blackwell Ultra systems (full date shown)."}]}}}},"b":{"schema":"5lc-index-rating/1","key":"infrastructure-servers-systems","entityId":"nvidia","rater":"b","model":"claude-sonnet-5-5","ratedAt":"2026-10-22","functions":{"gpu-systems":{"alternatives":{"level":1,"evidence":[{"url":"https://www.nvidia.com/en-us/data-center/dgx-platform/","date":"undated","support":"NVIDIA says DGX is available through a wide variety of partners and options; HGX and MGX are listed as separate platforms for partner-built systems. (page shows no publication date)"},{"url":"https://investor.nvidia.com/news/press-release-details/2025/NVIDIA-Blackwell-Ultra-AI-Factory-Platform-Paves-Way-for-Age-of-AI-Reasoning/","date":"2025-03-18","support":"16 named makers are expected to deliver servers based on the same Blackwell Ultra GPUs as DGX. (full date shown)"}]},"switchingTime":{"level":1,"evidence":[{"url":"https://www.nvidia.com/en-us/data-center/dgx-platform/","date":"undated","support":"Named DGX customers (Lockheed Martin, Sony, BMW, Shell) can buy HGX or MGX systems from partners; moving a DGX order to a partner is an order change of weeks to months (inference). (page shows no publication date)"}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://investor.nvidia.com/news/press-release-details/2025/NVIDIA-Blackwell-Ultra-AI-Factory-Platform-Paves-Way-for-Age-of-AI-Reasoning/","date":"2025-03-18","support":"Partners ship systems built on the same Blackwell Ultra GPUs, so a customer losing DGX gets the same silicon from others with no capability shortfall. (full date shown)"}]}}}}},"celestica":{"a":{"schema":"5lc-index-rating/1","key":"infrastructure-servers-systems","entityId":"celestica","rater":"a","model":"claude-opus-5-5","ratedAt":"2026-10-22","functions":{"custom-accelerator-systems":{"alternatives":{"level":2,"evidence":[{"url":"https://www.sec.gov/Archives/edgar/data/1030894/000103089426000011/cls-20251231.htm","date":"2026-02-27","support":"Celestica 10-K names Hon Hai, Flex, Jabil, Sanmina, Quanta, Wiwynn among rivals and says customers shift business and second-source; EDGAR timestamp."},{"url":"https://www.digitimes.com/news/a20251210PD211.html","date":"2025-12-10","support":"DIGITIMES: Inventec is a key Google TPU server supplier; Quanta, Foxconn, Wiwynn, Wistron and Mitac also pursue ASIC server orders (full date shown)."}]},"switchingTime":{"level":2,"evidence":[{"url":"https://www.sec.gov/Archives/edgar/data/1030894/000103089426000011/cls-20251231.htm","date":"2026-02-27","support":"Celestica 10-K (Item 1A): ramping new programs can take several months to over a year before production; date is EDGAR timestamp."},{"url":"https://www.sec.gov/Archives/edgar/data/1030894/000103089426000011/cls-20251231.htm","date":"2026-02-27","support":"Celestica 10-K MD&A: Enterprise revenue fell 19% in 2025 on a technology transition in one hyperscaler's AI/ML compute program; EDGAR timestamp."}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://focustaiwan.tw/business/202608120019","date":"2026-08-12","support":"Focus Taiwan: Hon Hai says it holds over 40% of the global AI server market and aims for 50% (company claim; full date shown)."},{"url":"https://www.taipeitimes.com/News/biz/archives/2026/02/10/2003852050","date":"2026-02-10","support":"Taipei Times: Wiwynn's clients include Google, AWS, Meta and Microsoft; it expects strong ASIC AI server shipments (company statements; full date)."}]}}}},"b":{"schema":"5lc-index-rating/1","key":"infrastructure-servers-systems","entityId":"celestica","rater":"b","model":"claude-sonnet-5-5","ratedAt":"2026-10-22","functions":{"custom-accelerator-systems":{"alternatives":{"level":3,"evidence":[{"url":"https://iconnect007.com/article/148613/celestica-announces-q4-fy-2025-financial-results/148610/pcb","date":"2026-01-29","support":"Company says it is a preferred manufacturing partner for Google's data-centre hardware and TPU systems and will add US capacity completing in 2027 (company claim). (full date shown)"},{"url":"https://www.digitimes.com/news/a20251210PD211.html","date":"2025-12-10","support":"DIGITIMES names Inventec as a Google TPU server supplier and lists Quanta, Foxconn, Wiwynn, Wistron and Mitac as ODMs pursuing ASIC server orders. (full date shown)"}]},"switchingTime":{"level":2,"evidence":[{"url":"https://www.sec.gov/Archives/edgar/data/1030894/000103089426000011/cls-20251231.htm","date":"2026-02-27","support":"10-K: top 10 customers were 79% of 2025 revenue and a technology transition in one hyperscaler AI/ML compute program cut Enterprise revenue; program changes take quarters, not years. (full date shown)"},{"url":"https://www.igorslab.de/en/foxconn-reaches-for-the-heart-of-the-ki-infrastructure-major-order-for-googles-tpu-server-reveals-new-dynamics-in-the-server-market/","date":"2025-12-02","support":"Foxconn is reported (unconfirmed) to be adding TPU compute-tray supply in Dec 2025, evidence that Google adds builders within months. (full date shown)"}]},"capabilityGap":{"level":2,"evidence":[{"url":"https://www.igorslab.de/en/foxconn-reaches-for-the-heart-of-the-ki-infrastructure-major-order-for-googles-tpu-server-reveals-new-dynamics-in-the-server-market/","date":"2025-12-02","support":"Foxconn is reported to produce over 1,000 AI racks a week with over 2,000 planned by end-2026, a larger builder than Celestica, so the best alternative has no capacity shortfall but lacks Celestica's TPU history (inference). (full date shown)"},{"url":"https://iconnect007.com/article/148613/celestica-announces-q4-fy-2025-financial-results/148610/pcb","date":"2026-01-29","support":"Celestica plans about $1 billion of 2026 capex for new capacity (Texas, Thailand, Mexico), showing the TPU build depends on capacity not yet in place. (full date shown)"}]}}}}},"inventec":{"a":{"schema":"5lc-index-rating/1","key":"infrastructure-servers-systems","entityId":"inventec","rater":"a","model":"claude-opus-5-5","ratedAt":"2026-10-22","functions":{"custom-accelerator-systems":{"alternatives":{"level":2,"evidence":[{"url":"https://www.digitimes.com/news/a20251210PD211.html","date":"2025-12-10","support":"DIGITIMES: Inventec is a key Google TPU server supplier; Quanta, Foxconn, Wiwynn, Wistron and Mitac also pursue ASIC server orders (full date shown)."},{"url":"https://www.sec.gov/Archives/edgar/data/1030894/000103089426000011/cls-20251231.htm","date":"2026-02-27","support":"Celestica 10-K names Hon Hai, Flex, Jabil, Sanmina, Quanta, Wiwynn among rivals and says customers shift business and second-source; EDGAR timestamp."}]},"switchingTime":{"level":2,"evidence":[{"url":"https://www.sec.gov/Archives/edgar/data/1030894/000103089426000011/cls-20251231.htm","date":"2026-02-27","support":"Celestica 10-K (Item 1A): ramping new programs can take several months to over a year before production; date is EDGAR timestamp."}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://wccftech.com/nvidia-key-supply-chain-partner-foxconn-has-reportedly-won-orders-for-google-tpu/","date":"2025-12-01","support":"Wccftech, citing Taiwan's Economic Daily: Foxconn reportedly received orders for Google TPU compute trays (single-source report; full date shown)."},{"url":"https://www.taipeitimes.com/News/biz/archives/2026/05/15/2003857345","date":"2026-05-15","support":"Taipei Times: Quanta aims to double AI-related capacity this year and expects more ASIC server shipments (company statements; full date shown)."}]}}}},"b":{"schema":"5lc-index-rating/1","key":"infrastructure-servers-systems","entityId":"inventec","rater":"b","model":"claude-sonnet-5-5","ratedAt":"2026-10-22","functions":{"custom-accelerator-systems":{"alternatives":{"level":3,"evidence":[{"url":"https://www.digitimes.com/news/a20251210PD211.html","date":"2025-12-10","support":"DIGITIMES calls Inventec a supplier for Google's TPU servers (the company does not comment on customers) and lists Quanta, Foxconn, Wiwynn, Wistron and Mitac as ODMs pursuing ASIC orders. (full date shown)"},{"url":"https://iconnect007.com/article/148613/celestica-announces-q4-fy-2025-financial-results/148610/pcb","date":"2026-01-29","support":"Celestica says it is a preferred manufacturing partner for Google's TPU systems (company claim), so at least one other builder serves the same program. (full date shown)"}]},"switchingTime":{"level":2,"evidence":[{"url":"https://www.igorslab.de/en/foxconn-reaches-for-the-heart-of-the-ki-infrastructure-major-order-for-googles-tpu-server-reveals-new-dynamics-in-the-server-market/","date":"2025-12-02","support":"Foxconn was reported in Dec 2025 as newly supplying compute trays alongside Google TPU racks, evidence that Google adds builders within months; the report is unconfirmed. (full date shown)"},{"url":"https://www.taipeitimes.com/News/biz/archives/2025/11/12/2003847039","date":"2025-11-12","support":"AI servers are 40-50% of Inventec's server revenue and it builds L6 and L10 servers for Dell, HP and Lenovo; I found no ASIC volume figure for it. (full date shown)"}]},"capabilityGap":{"level":2,"evidence":[{"url":"https://www.digitimes.com/news/a20251210PD211.html","date":"2025-12-10","support":"DIGITIMES Research expects Google to ship 3.325 million TPUs in 2026 (AWS 1.47 million, Meta 600,000, Microsoft 330,000); Inventec expects double-digit growth in ASIC servers, and other ODMs are competing for TPU allocation. (full date shown)"},{"url":"https://www.igorslab.de/en/foxconn-reaches-for-the-heart-of-the-ki-infrastructure-major-order-for-googles-tpu-server-reveals-new-dynamics-in-the-server-market/","date":"2025-12-02","support":"Foxconn is reported to produce over 1,000 AI racks a week and plan over 2,000 by end-2026, so larger builders could absorb Inventec's TPU work with a small shortfall. (full date shown)"}]}}}}},"nettrix":{"a":{"schema":"5lc-index-rating/1","key":"infrastructure-servers-systems","entityId":"nettrix","rater":"a","model":"claude-opus-5-5","ratedAt":"2026-10-22","functions":{"china":{"alternatives":{"level":1,"evidence":[{"url":"https://www.idc.com/?p=104194","date":"2025-10-08","support":"IDC China H1 2025 accelerated servers: Inspur, H3C, Lenovo top three by revenue near 50%; by units Inspur, H3C, Nettrix about 43% (page date)."},{"url":"https://www.idc.com/?p=104362","date":"2025-03-30","support":"IDC China 2024 accelerated servers: Inspur, Nettrix, H3C top three by revenue over 50%; by units Inspur, Nettrix, Huawei near 55% (page date)."}]},"switchingTime":{"level":2,"evidence":[{"url":"https://www.cls.cn/detail/1678572","date":"2024-05-16","support":"Cailian Press: China Mobile's 2024-25 AI computing procurement split among seven makers, e.g. Henan Kunlun 21.05%, Huakun Zhenyu 17.54%, Baode 15.79% (full date)."}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://www.idc.com/?p=104194","date":"2025-10-08","support":"IDC China H1 2025 accelerated servers: Inspur, H3C, Lenovo top three by revenue near 50%; by units Inspur, H3C, Nettrix about 43% (page date)."}]}}}},"b":{"schema":"5lc-index-rating/1","key":"infrastructure-servers-systems","entityId":"nettrix","rater":"b","model":"claude-sonnet-5-5","ratedAt":"2026-10-22","functions":{"china":{"alternatives":{"level":1,"evidence":[{"url":"https://www.chinabaogao.com/jingzheng/202510/767092.html","date":"2025-10-14","support":"IDC China 1H25: Inspur, H3C and Lenovo top accelerated-server sales (about 50% together); by shipments Inspur, H3C and Nettrix (Ningchang), about 43%: roughly half of the market sits with other domestic makers. (full date shown)"}]},"switchingTime":{"level":2,"evidence":[{"url":"https://www.chinabaogao.com/jingzheng/202510/767092.html","date":"2025-10-14","support":"Nettrix (Ningchang) is third by shipments in IDC China 1H25, so its volume is large enough that re-spreading it takes a few quarters (inference). (full date shown)"}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://www.chinabaogao.com/jingzheng/202510/767092.html","date":"2025-10-14","support":"Inspur and H3C rank first and second on both sales and shipments, so Nettrix's best alternative is not smaller. (full date shown)"}]}}}}},"zte":{"a":{"schema":"5lc-index-rating/1","key":"infrastructure-servers-systems","entityId":"zte","rater":"a","model":"claude-opus-5-5","ratedAt":"2026-10-22","functions":{"china":{"alternatives":{"level":1,"evidence":[{"url":"https://www.idc.com/?p=104194","date":"2025-10-08","support":"IDC China H1 2025 accelerated servers: Inspur, H3C, Lenovo top three by revenue near 50%; by units Inspur, H3C, Nettrix about 43% (page date)."},{"url":"https://m.c114.com.cn/w127-1295268.html","date":"2025-08-18","support":"C114: ZTE named winning candidate for packages 1-3 of China Mobile's 2025 AI inference equipment tender; volume reallocated by tender (full date shown)."}]},"switchingTime":{"level":2,"evidence":[{"url":"https://m.c114.com.cn/w127-1295268.html","date":"2025-08-18","support":"C114: ZTE named winning candidate for packages 1-3 of China Mobile's 2025 AI inference equipment tender; volume reallocated by tender (full date shown)."},{"url":"https://www.cls.cn/detail/1678572","date":"2024-05-16","support":"Cailian Press: China Mobile's 2024-25 AI computing procurement split among seven makers, e.g. Henan Kunlun 21.05%, Huakun Zhenyu 17.54%, Baode 15.79% (full date)."}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://www.idc.com/?p=104194","date":"2025-10-08","support":"IDC China H1 2025 accelerated servers: Inspur, H3C, Lenovo top three by revenue near 50%; by units Inspur, H3C, Nettrix about 43% (page date)."}]}}}},"b":{"schema":"5lc-index-rating/1","key":"infrastructure-servers-systems","entityId":"zte","rater":"b","model":"claude-sonnet-5-5","ratedAt":"2026-10-22","functions":{"china":{"alternatives":{"level":1,"evidence":[{"url":"https://www.chinabaogao.com/jingzheng/202510/767092.html","date":"2025-10-14","support":"IDC China 1H25: Inspur, H3C and Lenovo top accelerated-server sales (about 50% together); by shipments Inspur, H3C and Nettrix (Ningchang), about 43%: roughly half of the market sits with other domestic makers. (full date shown)"},{"url":"https://www2.yicaiglobal.com/news/chinas-computing-power-shifts-from-standalone-chips-to-full-system-ecosystems","date":"2026-07-19","support":"ZTE showed an OEX super-node at WAIC 2026 alongside Huawei, T-Head and MetaX nodes; no specs or customers were given, so evidence on ZTE is thin. (full date shown)"}]},"switchingTime":{"level":1,"evidence":[{"url":"https://www.chinabaogao.com/jingzheng/202510/767092.html","date":"2025-10-14","support":"ZTE is not among the IDC China top three on sales or shipments; its volume is small next to Inspur and H3C and easily re-spread (inference). (full date shown)"}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://eu.36kr.com/en/p/3913165545837700","date":"2026-07-27","support":"ZTE is named only as launching super-node offerings in 2025 with no specs or customers; larger vendors offer comparable nodes, so no alternative falls short (thin evidence). (full date shown)"}]}}}}},"jabil":{},"sanmina":{"a":{"schema":"5lc-index-rating/1","key":"infrastructure-servers-systems","entityId":"sanmina","rater":"a","model":"claude-opus-5-5","ratedAt":"2026-10-22","functions":{"custom-accelerator-systems":{"alternatives":{"level":2,"evidence":[{"url":"https://www.sec.gov/Archives/edgar/data/1030894/000103089426000011/cls-20251231.htm","date":"2026-02-27","support":"Celestica 10-K names Hon Hai, Flex, Jabil, Sanmina, Quanta, Wiwynn among rivals and says customers shift business and second-source; EDGAR timestamp."},{"url":"https://www.digitimes.com/news/a20251210PD211.html","date":"2025-12-10","support":"DIGITIMES: Inventec is a key Google TPU server supplier; Quanta, Foxconn, Wiwynn, Wistron and Mitac also pursue ASIC server orders (full date shown)."}]},"switchingTime":{"level":2,"evidence":[{"url":"https://www.sec.gov/Archives/edgar/data/1030894/000103089426000011/cls-20251231.htm","date":"2026-02-27","support":"Celestica 10-K (Item 1A): ramping new programs can take several months to over a year before production; date is EDGAR timestamp."},{"url":"https://www.sec.gov/Archives/edgar/data/0000897723/000089772326000037/sanm-20260627.htm","date":"2026-07-27","support":"Sanmina 10-Q: ZT Systems, bought 27 Oct 2025, makes AI and general-purpose infrastructure for hyperscalers; date is EDGAR filing timestamp."}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://focustaiwan.tw/business/202608120019","date":"2026-08-12","support":"Focus Taiwan: Hon Hai says it holds over 40% of the global AI server market and aims for 50% (company claim; full date shown)."},{"url":"https://www.taipeitimes.com/News/biz/archives/2026/02/10/2003852050","date":"2026-02-10","support":"Taipei Times: Wiwynn's clients include Google, AWS, Meta and Microsoft; it expects strong ASIC AI server shipments (company statements; full date)."}]}}}},"b":{"schema":"5lc-index-rating/1","key":"infrastructure-servers-systems","entityId":"sanmina","rater":"b","model":"claude-sonnet-5-5","ratedAt":"2026-10-22","functions":{"custom-accelerator-systems":{"alternatives":{"level":3,"evidence":[{"url":"https://www.emsnow.com/?p=51735","date":"2025-10-28","support":"Sanmina says the ZT purchase deepens its cloud and AI hyperscaler relationships, and Sanmina is AMD's US NPI manufacturing partner of choice for rack and cluster-scale AI systems (company statement); customers are not named, so evidence is thin. (full date shown)"},{"url":"https://www.digitimes.com/news/a20251210PD211.html","date":"2025-12-10","support":"DIGITIMES lists Quanta, Foxconn, Wiwynn, Wistron, Inventec and Mitac as ODMs pursuing ASIC server orders; the confirmed builders per hyperscaler program number one to three. (full date shown)"}]},"switchingTime":{"level":2,"evidence":[{"url":"https://s201.q4cdn.com/209924174/files/doc_financials/2026/q3/SANM-Q326-Earnings-Release-7_27_26-FINAL.pdf","date":"2026-07-27","support":"Sanmina reports more customer orders in both core Sanmina and ZT in the June-2026 quarter and expects growth to ramp through FY2027 into FY2028; it discloses no custom-accelerator revenue, so I judge a few quarters to re-spread it (inference). (full date shown)"}]},"capabilityGap":{"level":2,"evidence":[{"url":"https://www.emsnow.com/?p=51735","date":"2025-10-28","support":"ZT's New Jersey and Texas sites offer advanced liquid cooling and Sanmina cites full systems integration at scale (company statement); I found no volume figure to compare with Taiwanese ODMs. (full date shown)"}]}}}}},"mitac":{},"huakun-zhenyu":{"a":{"schema":"5lc-index-rating/1","key":"infrastructure-servers-systems","entityId":"huakun-zhenyu","rater":"a","model":"claude-opus-5-5","ratedAt":"2026-10-22","functions":{"china":{"alternatives":{"level":1,"evidence":[{"url":"https://www.cls.cn/detail/1678572","date":"2024-05-16","support":"Cailian Press: China Mobile's 2024-25 AI computing procurement split among seven makers, e.g. Henan Kunlun 21.05%, Huakun Zhenyu 17.54%, Baode 15.79% (full date)."},{"url":"https://www.idc.com/?p=104194","date":"2025-10-08","support":"IDC China H1 2025 accelerated servers: Inspur, H3C, Lenovo top three by revenue near 50%; by units Inspur, H3C, Nettrix about 43% (page date)."}]},"switchingTime":{"level":2,"evidence":[{"url":"https://www.cls.cn/detail/1678572","date":"2024-05-16","support":"Cailian Press: China Mobile's 2024-25 AI computing procurement split among seven makers, e.g. Henan Kunlun 21.05%, Huakun Zhenyu 17.54%, Baode 15.79% (full date)."}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://www.cls.cn/detail/1678572","date":"2024-05-16","support":"Cailian Press: China Mobile's 2024-25 AI computing procurement split among seven makers, e.g. Henan Kunlun 21.05%, Huakun Zhenyu 17.54%, Baode 15.79% (full date)."}]}}}},"b":{"schema":"5lc-index-rating/1","key":"infrastructure-servers-systems","entityId":"huakun-zhenyu","rater":"b","model":"claude-sonnet-5-5","ratedAt":"2026-10-22","functions":{"china":{"alternatives":{"level":2,"evidence":[{"url":"https://www.yicai.com/news/102154509.html","date":"2024-06-18","support":"xFusion and Huakun Zhenyu are the top two whole-machine vendors for Huawei's Kunpeng and Ascend lines and Digital China is another of its first AI server partners; Talkweb is another ecosystem partner (Yicai, 2024). (full date shown)"},{"url":"https://apps.digitimes.com/news/a20260707RS401.html?chid=2","date":"2026-07-13","support":"DIGITIMES Research says Huakun works under a JDM-like model with Huawei, which supplies core components and reference designs while Huakun handles system design, manufacturing and delivery. (full date shown)"}]},"switchingTime":{"level":1,"evidence":[{"url":"https://apps.digitimes.com/news/a20260707RS401.html?chid=2","date":"2026-07-13","support":"Huawei supplies the production standards, CPUs, accelerators, software stacks and reference designs, so another partner or Huawei itself can build the same systems quickly (inference). (full date shown)"}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://www.yicai.com/news/102154509.html","date":"2024-06-18","support":"xFusion ships comparable Huawei-based whole machines, so Huakun's best alternative is not behind. (full date shown)"}]}}}}},"flex":{},"sugon":{"a":{"schema":"5lc-index-rating/1","key":"infrastructure-servers-systems","entityId":"sugon","rater":"a","model":"claude-opus-5-5","ratedAt":"2026-10-22","functions":{"china":{"alternatives":{"level":1,"evidence":[{"url":"https://www.idc.com/?p=104194","date":"2025-10-08","support":"IDC China H1 2025 accelerated servers: Inspur, H3C, Lenovo top three by revenue near 50%; by units Inspur, H3C, Nettrix about 43% (page date)."},{"url":"https://www.cls.cn/detail/1678572","date":"2024-05-16","support":"Cailian Press: China Mobile's 2024-25 AI computing procurement split among seven makers, e.g. Henan Kunlun 21.05%, Huakun Zhenyu 17.54%, Baode 15.79% (full date)."}]},"switchingTime":{"level":2,"evidence":[{"url":"https://www.cls.cn/detail/1678572","date":"2024-05-16","support":"Cailian Press: China Mobile's 2024-25 AI computing procurement split among seven makers, e.g. Henan Kunlun 21.05%, Huakun Zhenyu 17.54%, Baode 15.79% (full date)."}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://www.idc.com/?p=104194","date":"2025-10-08","support":"IDC China H1 2025 accelerated servers: Inspur, H3C, Lenovo top three by revenue near 50%; by units Inspur, H3C, Nettrix about 43% (page date)."},{"url":"https://www.ithome.com/0/895/328.htm","date":"2025-11-06","support":"IT Home: Sugon launched scaleX640, a single-rack 640-card supernode supporting multiple accelerator brands (full date shown)."}]}}}},"b":{"schema":"5lc-index-rating/1","key":"infrastructure-servers-systems","entityId":"sugon","rater":"b","model":"claude-sonnet-5-5","ratedAt":"2026-10-22","functions":{"china":{"alternatives":{"level":1,"evidence":[{"url":"https://www.chinabaogao.com/jingzheng/202510/767092.html","date":"2025-10-14","support":"IDC China 1H25: Inspur, H3C and Lenovo top accelerated-server sales (about 50% together); by shipments Inspur, H3C and Nettrix (Ningchang), about 43%: roughly half of the market sits with other domestic makers. (full date shown)"},{"url":"https://eu.36kr.com/en/p/3913165545837700","date":"2026-07-27","support":"Sugon's 640-accelerator immersion-cooled cabinet competes with Huawei's 1,024-card node, H3C's 512-card design, Inspur and ZTE nodes (36Kr). (full date shown)"}]},"switchingTime":{"level":2,"evidence":[{"url":"https://eu.36kr.com/en/p/3913165545837700","date":"2026-07-27","support":"Sugon's 100,000-accelerator cluster at the Zhengzhou National Supercomputing Internet node runs at about 90% capacity; immersion-cooled cabinet designs would take months to a year to re-engineer elsewhere (inference). (full date shown)"}]},"capabilityGap":{"level":2,"evidence":[{"url":"https://eu.36kr.com/en/p/3913165545837700","date":"2026-07-27","support":"Sugon's 640 accelerators per cabinet (company claim) exceed rivals' 32-512 per cabinet on display; Huawei's 1,024-card node uses 20 cabinets, so alternatives fall slightly short at cabinet scale. (full date shown)"}]}}}}},"pegatron":{},"compal":{},"cerebras":{}},"validation":[]}