Google DeepMind director of AGI economics Alex Imas and Epoch AI's Phil Trammell discuss what remain...
The AMW Read
Incremental conceptual framing from DeepMind's economist, no new empirical data; significance is segment-level because it potentiall reshapes investor priors on pricing power across the foundation-model stack.
Google DeepMind director of AGI economics Alex Imas and Epoch AI's Phil Trammell discuss what remains scarce after AI advances in an interview with Dwarkesh Patel.
The conversation tackles a foundational structural force in the AI market: as model capabilities compress the cost of cognitive labor near zero, which inputs retain scarcity value—compute, data, human judgment, or regulatory permission? The exchange updates the capital-cycle arc and the context-engineering moat debate, suggesting that even in a world of abundant AI, bottlenecks around trusted inference infrastructure and high-stakes decision-making may persist.
Trammell's historical lens on commodity-pricing dynamics for compute and Imas's behavioral-economics perspective on human-AI delegation create a useful analytical framework for investors tracking which parts of the AI stack will command pricing power as foundation-model margins compress. The interview implicitly reinforces the thesis that distribution moats and regulatory charters, not raw model quality, may determine long-term winner capture.
