Premium: What If…We’re In An AI Bubble? (Part 3)

Source: Where’s Your Ed At Date: 2026-05-29 URL: https://www.wheresyoured.at/premium-what-if-were-in-an-ai-bubble-part-3/

Summary

Ed Zitron’s third installment of the AI bubble series marshals financial data to argue the current infrastructure build-out is mathematically incoherent. Key claims: NVIDIA’s trillion-dollar GPU sales projections require $435B in annual compute demand with no visible buyer base at that scale; OpenAI and Anthropic have committed $1.1T in compute through 2030 while projecting $358B in combined annual revenue by 2029 against an $852B cash burn for OpenAI alone; and enterprise AI spend is stalling because buyers cannot draw a line between AI expenditure and measurable business returns.

Implications

• Direct signal for the AI bubble/economics thread: this is the clearest public accounting of the demand–supply mismatch in foundation model infrastructure.
• The enterprise adoption plateau (Uber example, stalled WAU at 905M) is relevant to capital markets/IPO framing—subscriber growth deceleration undermines the revenue multiples baked into current valuations.
• Structurally, the piece argues that the bubble is sustained by circular compute commitments rather than genuine end-user demand—a dynamic that affects how seriously to weight any vendor’s stated pipeline.