Monologue: AI's Compute Demand Story Is A Lie

The story of massive AI demand is a lie. That is the central argument of this monologue, and it is supported by a detailed financial autopsy of the biggest names in tech. The host, Ed Zitron, argues that the nearly trillion-dollar AI infrastructure buildout is not being driven by genuine market demand, but by a circular, self-licking ice cream cone where the same small group of companies—OpenAI, Anthropic, and the hyperscalers—are effectively funding each other to justify their own existence.

The Numbers Don't Lie: A Trillion Dollars for Pocket Change

The episode’s core evidence is a breakdown of the AI revenue reported by Microsoft and Amazon. Zitron argues these numbers are not just bad; they are catastrophic when compared to the capital expenditures (capex) required to generate them. Microsoft reported a $37 billion AI revenue run rate. However, Zitron’s own reporting estimates that OpenAI alone spent over $18 billion on Microsoft Azure for inference in 2025, with billions more on training. He calculates that OpenAI represents roughly 70% of Microsoft’s AI revenue, meaning the rest of the world’s AI demand is a rounding error. The remaining $7 billion from 20 million Copilot subscribers is, in his view, a pittance.