OpenAI alone accounts for 70% or more of Microsoft's AI revenues in FY2026, contributing $24.1 billion against Microsoft's total capital expenditure of $261.3 billion since 2022. Wells Fargo, Barclays, and UBS analyst notes corroborate this concentration: UBS estimates that Anthropic and OpenAI's combined compute spend will equal 48% of all Google Cloud revenues next year. Bloomberg confirmed the Microsoft figure 24 hours after this newsletter first published it. Two companies, both unprofitable, both losing tens of billions annually, are the load-bearing walls of the cloud industry's AI growth story.
The math is the story. OpenAI plans to spend $50 billion on compute in 2026. Assume Anthropic matches it. That is $100 billion. To fund that level of spend, the two companies raised a combined $217 billion in the first half of 2026 alone. Projecting forward, sustaining the GPU demand Jensen Huang is banking on, roughly $360 to $435 billion in annual compute revenue, requires Anthropic and OpenAI to spend $200 billion or more in 2027, which in turn requires $250 to $300 billion in fresh capital. Apollo chief economist Torsten Slok framed the structural risk plainly: the most profitable part of the AI value chain depends entirely on the least profitable part continuing to grow revenue or raise capital.
Read the original for the analyst note citations, the neocloud dependency chain, and the Friday follow-up piece calculating exactly how much capital is required to keep this structure standing through 2027. If you have information about Anthropic, OpenAI, or any other company in this supply chain, contact the author on Signal at ezitron.76.
[READ ORIGINAL →]