AI CapEx boom risks structural overinvestment
The bear case argued for the AI infrastructure boom is that massive, duplicative capital expenditure driven by game-theoretic pressures will lead to severe overcapacity and a subsequent collapse in profitability.
The argument
The guest compared the current AI build-out to historical technology manias like the 1840s railway mania and the late-1990s telecom boom, where companies overinvested due to a 'prisoner's dilemma' dynamic. He argued that demand is being vastly overestimated, and that tech firms are already extending GPU depreciation schedules to temporarily mask the impact on reported profits.
The thesis, stress-tested
✓ What validates it
- ✓A sharp curtailment in CapEx guidance by major hyperscalers
- ✓Massive write-downs or accelerated depreciation of GPU assets on tech earnings reports
- ✓A decline in reported cloud revenue growth relative to capital expenditure
▸ Risks discussed
- ▸AI technology achieves a rapid breakthrough in agentic reasoning or AGI
- ▸Enterprise demand for LLMs grows faster than historical technology adoption curves
- ▸A single player establishes an early, highly profitable monopoly that justifies the CapEx
Hear it yourself
"In addition, he's the author of The Price of Time, The Real Story of Interest, and most recently helped our former boss, Jeremy Grantham, write his autobiography. Ed, welcome to the show. I see you, Kai. What you what you did mention is that you you came to GMO as my analyst back in 2008. So you get back we get back to the financial crisis together, don't we? Yeah. And and and the other piece being that, to add to your many accomplishments, helping advise me on my economics thesis on credit cycles and bubbles, at Harvard. And if if I'm allowed to blow our trumpet together, do you remember"
AFFILIATE LINK · ZORTIX MAY EARN A COMMISSION · NEVER A RECOMMENDATION TO TRADE