Aug 31, 2026 · 1h 15m
AI demand tests the limits of compute supply
Gavin Baker: Why AI Demand Is Outrunning Compute Supply
The episode frames AI’s central economic question as a potential infrastructure shortage, with consequences for prices, access, productivity, and industrial investment.
- 1Inference, enterprise adoption, and token consumption may be growing faster than new compute capacity can come online.
- 2The AI ecosystem could remain positive-sum as frontier labs, open-source models, applications, clouds, and chipmakers expand usage together.
- 3Orbital compute and multi-model enterprise systems could reshape infrastructure economics, but both depend on difficult engineering and financing choices.
Don't miss
The most expansive turn comes when Gavin Baker links cheaper Starship launches to orbital compute, asteroid mining, lunar infrastructure, and a Mars settlement.
The brief
Gavin Baker and David George search for evidence that AI demand is weakening, but find a market still accelerating across inference, enterprise adoption, and token use.
The central tension is whether AI is a bubble or an infrastructure shortage: data centers, energy, copper, wafers, and financing are all becoming constraints.
They argue that open source is not free, NVIDIA’s ecosystem remains unusually powerful, and frontier labs, clouds, applications, and chipmakers can benefit together.
The conversation stretches from orbital data centers and asteroid mining to enterprise routers combining open, internal, and frontier models behind one intelligence stack.
The takeaway is less a forecast than a systems view: AI’s upside depends on expanding affordable compute without letting scarcity reserve advanced intelligence for the largest buyers.
Featuring
Jen-Hsun Huang
StarlinkBooks & mentions
Listen to the full episode and explore every guest, topic, and moment on PodLume.