← Home
The All-In Podcast · September 17, 2026 · 18m 10s

Brad Gerstner: No AI Bubble, Semis Eat the Nasdaq & AI's Take Off Problem

Brad Gerstner

Brad Gerstner joins the All-In crew to argue there is no AI bubble — the make-or-break question is whether AI revenue can cover the enormous capex the industry is deploying. The discussion centers on what Gerstner frames as AI's take-off problem: the physical build-out involving gigawatts of power, expanded TAM, token growth, and margin expansion, and whether it can keep pace with the technology. The closing segment runs through the risks that could break the thesis, including AI regulation, the nuclear regulatory precedent, power limits, and rising interest rates.

This summary was generated from show notes and public descriptions, not from a full transcript review. Details may contain inaccuracies.

Canon

•
The AI build-out is colliding with hard physical limits — gigawatts of power, land, and grid interconnections — marking the end of the era in which scaling software was cheap and capital-light.

Highlights

•
Can AI Revenue Pay for the CapEx?00:05:07
Gerstner frames the entire AI investment debate as a single question: whether the revenue AI companies can generate will be enough to pay for the massive capital expenditures they are making.
•
The CAC Scan00:01:01
Gerstner applies a customer acquisition cost scan — measuring what it costs to acquire customers against the value they generate — as a way to judge whether AI spending is grounded in real customer economics.
•
AI's Take-Off Problem00:08:53
Gerstner identifies a take-off problem for AI: the gap between what the technology can do and whether the physical build-out of gigawatts, expanded TAM, token growth, and margin expansion can keep up.
•
The Risks to the AI Build-Out00:12:30
Gerstner closes by running through the main risks to the AI build-out: AI regulation, the precedent of nuclear power's stalled rollout, physical power limits, and rising interest rates.
Was this useful?