Here's How The AI Bubble Bursts — With Paul Kedrosky
Summary
Investor Paul Kodrosky argues that the current AI infrastructure spending constitutes a historic bubble, exceeding past industrial buildouts in scale and speed. He contends that rapid hardware depreciation and collapsing token prices make it mathematically unlikely for returns to justify the massive capital expenditure. Kodrosky challenges the notion that AI firms can easily move upmarket to compensate, warning that this economic mismatch could lead to a significant market unraveling.
Topics discussed
Introduction: Is the AI investment bubble unraveling?
Historical context: AI spending vs. past infrastructure booms
The rush to invest without natural stop points
Why the commercial real estate analogy fails for AI
GPU depreciation, failure rates, and token deflation
Model convergence and the Jevons paradox argument
Can AI labs move upmarket to justify CapEx?
Why frontier models won't replace vertical software
The 'AGI call option' and irrational exuberance
Sovereign wealth funds and the mechanics of the bubble
Sponsors and break
Triggers for the bubble burst and the role of harnesses
Systemic risk and investor exposure to AI debt
China's approach and the future of token economics
Listen ad-free on Castria