ReadingResearchRadarInvestment framework
Sign in / Sign up
← Back to reading
TopicDisconfirmAI EditorialUpdated 2026-11-16 21:32

AI产业链毛利瀑布

2 steps1 pieces3 Sources

Theory Ventures合伙人Tomasz Tunguz发布AI产业链毛利瀑布模型,将电力、数据中心、Neocloud、推理及应用五层统一折算为“每年每瓦美元”($/w/yr)。在假设应用层毛利率为40%的前提下,应用层(40%)、Neocloud(28%,折旧前)与推理层(24%)合计捕获全链条92%的毛利润,而发电厂与数据中心仅占8%。

该模型揭示算力成本正急剧攀升:CoreWeave CEO Michael Intrator证实新长期合同价格已升至$20-$25/w/yr,远超其当前$9/w/yr的混合车队均价;Anthropic S-1/A文件显示短期突发算力高达$13/GPU-hour。这引发核心争议:若应用无法通过高ROI向终端转嫁成本,上游涨价将挤压中下游利润。此外,Falk Gottlob指出,计入GPU巨额折旧后,Neocloud实际净毛利可能仅剩$0-$2/w/yr,且不同应用定价权差异巨大(如Canva坚持75%毛利 vs Gamma近乎零毛利),表明40%仅为一种情景假设而非普遍现实。

How it got here

  1. 2026-11-11

    Tomasz Tunguz发布《Who Captures the Most Margin in AI?》,提出基于$/w/yr的AI价值链毛利分布模型。

    tomtunguz.com ↗
  2. 2026-11-12

    Falk Gottlob发文质疑Tunguz模型,指出计入GPU折旧后Neocloud毛利大幅缩水,且应用层毛利因ROI差异呈区间分布而非固定值。

    falkster.com ↗

What is still unsettled

  • 随着新算力合约价格翻倍至$20-$25/w/yr,缺乏高ROI验证的应用层能否成功将成本转嫁给终端用户?
  • 在计入GPU硬件折旧(约$6-$8/w/yr)后,Neocloud层的实际可持续盈利能力是否被严重高估?
  • 垂直整合(如CoreWeave收购W&B)能否有效抵消单点算力波动,还是会导致新的垄断性成本结构?

Sources

  • tomtunguz.comTomasz Tunguz: Who Captures the Most Margin in AI?↗
Sign in / Sign up
ReadingResearchRadarInvestment framework
falkster.comFalk Gottlob: Forty Percent Is a Scenario↗
  • thenextgentechinsider.comThe Next Gen Tech Insider: Analysis Reveals AI Software and Neoclouds Capture 92 Percent of Gross Profit↗
  • Coverage in this topic

    • Theory Ventures Partner Tomasz Tunguz: The AI Industry Chain Gross Margin Waterfall — Applications Capture 40%, Neocloud Post-Depreciation Margin Is Only $0 to $2 per Watt-Year2026-11-12News

      Theory Ventures partner Tomasz Tunguz normalizes the five layers of the AI value chain into dollars per watt-year ($/w/yr), constructing a gross margin waterfall based on CoreWeave’s Q3 2026 financial report and Anthropic’s S-1/A filing. The analysis shows that applications capture 40% of the total pool, while Neoclouds show a pre-depreciation margin of 28% but only $0 to $2 per watt-year after depreciation. Inference vendors capture 24%. Three key implications emerge: rising compute prices test the ability to pass on pricing power; differing application ROI determines differences in pricing power; and vertical integration is encroaching on adjacent layer margins. The author acknowledges that the model relies on approximate assumptions and contains significant error.