Ed Zitron: OpenAI and Anthropic Prop Up Cloud AI Revenue

The core claim in Ed Zitron's latest Where's Your Ed At essay is that the AI boom's headline numbers only make sense if you avoid asking who is actually paying for the compute. Zitron argues that "the most profitable part of the AI value chain depends on the least profitable part continuing to grow revenue or raise capital", specifically that OpenAI and Anthropic, both of which "lose tens of billions of dollars a year", are the load-bearing customers propping up Microsoft, Google, Amazon and by extension Nvidia. Three of our tracked AI experts shared the piece.

The concentration numbers he assembles are the striking part. Zitron writes that "70% or more of the AI revenues of Microsoft, Google, and Amazon were from either OpenAI or Anthropic", that "OpenAI alone contributed to 70% or more of Microsoft's AI revenues for the year", and that on the projections he cites, "48% of Google Cloud's 2027 revenues will come from OpenAI and Anthropic". The two labs, he notes, "raised a combined $217 billion in the first half of 2026" and "they'll have to raise $150 billion each leading up to or in 2027" just to keep the compute bills paid.

On the supply side, Zitron notes that hyperscaler PP&E "grew by $754.5 billion" since the start of 2022, while Microsoft's non-OpenAI AI revenue of "$10.33 billion in annual revenue is a catastrophic failure" against that spend. He estimates that "for the 190GW of capacity in planning, we need somewhere between $1.62 trillion and $2.92 trillion" in annual demand, and warns of "$250 billion this year and $400 billion next year" being issued in the bond markets on top of "the expected $1.65 trillion in off-balance sheet obligations". His summary line: "If Microsoft doesn't have the demand, nobody has the demand."

He is equally skeptical of "the supposed $500 billion infrastructure deal between NVIDIA and The Avengers of Private Credit", pointing to Jensen Huang's own comment that "It was never a commitment", and to a disclosed "residual-value support mechanism for up to 25% of an opportunity" from Nvidia to grease GPU sales.

What the piece does not do is put a date on when this breaks, or name which lender or public shareholder eats the loss first if OpenAI or Anthropic can't close their next rounds; it is a concentration argument, not a schedule. If you sit on the capex or forecasting side of any of these companies, the useful read is not that the bubble ends tomorrow, but that the revenue base is far narrower than the slide deck suggests.