Why smarter AI models could drive up compute prices 10x
Summary
Dwarkesh Patel analyzes the divergence between AI labs' 10x revenue growth and only 3x compute scaling. He argues that rising compute prices and increasing inference margins are bridging this gap, favoring efficient models via the Alkin-Allen effect. This dynamic may price out low-value applications and concentrate power among top labs, as compute supply remains constrained by hardware bottlenecks.
Topics discussed
Anthropic's revenue growth vs compute scaling trends
Rising margins, compute prices, and inference share
Why compute prices must rise to bridge the gap
Monetizing compute and the lump of labor fallacy
The Alkin-Allen effect and pricing out AI apps
Compute scarcity limits and the Simon-Ehrlich bet
Sponsor: Mercury Command for expense automation
Future compute costs and economies of scale
Listen ad-free on Castria