Premium: The AI Compute Demand Story Is A Lie

Ed Zitron argues that the narrative of massive AI compute demand is fabricated. The apparent capacity crunch is not driven by organic market demand but by hyperscalers (Google, Amazon, Microsoft) funneling hundreds of billions into OpenAI and Anthropic — companies they themselves incubated and subsidize. Anthropic projects losing $11 billion annually in 2026 and 2027 while planning to spend $86 billion on training costs through 2029. Microsoft, Google, and Amazon have collectively spent over $803 billion in capex, with an estimated 75-85% of AI data center capacity built to serve just two customers. Outside of OpenAI, Anthropic, and the hyperscalers themselves, less than $1 billion in genuine AI compute demand can be identified. The piece concludes that AI demand is a circular financing illusion, not a real market signal.