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

Video summary

Brad Gerstner says AI is no bubble, but Anthropic revenue and 2026 compute buildout decide whether markets reach liftoff.

Brad Gerstner argues the AI rally is earnings-driven rather than a 2000-style bubble: Nvidia trades at roughly 14 times next year’s fully taxed GAAP earnings, while semiconductors account for 70% of Nasdaq returns. The key test is whether AI lab revenue can fund hyperscaler capex. Anthropic’s monthly revenue reportedly surged from $2 billion in January to $11 billion in March, with the top three labs—Anthropic, OpenAI, and SpaceX—near a $100 billion run rate and needing to reach at least $180 billion by year-end. Gerstner sees enormous demand, including 47 quadrillion tokens and 40x Codeex user growth, but doubts SemiAnalysis’s forecast of 43 gigawatts of new compute next year, estimating closer to 25 gigawatts because of permits, grid delays, labor shortages, and sold-out equipment. Higher rates and AI regulation remain major risks, so he recommends staying medium-sized and mentally flexible.

Chapters

1. 0:00Welcome Brad Gerstner: Five Companies and 70 Million Child Investors
2. 1:01Trump Accounts and CAC Scans: 70 Million Child Owners and $100 Heart Screening
3. 5:07Can AI Revenue Pay for CapEx?: Anthropic Reaches $11 Billion Monthly Revenue
4. 8:53The Build Out Issue: 43 Gigawatts, $1.2 Trillion TAM, and 100-Basis-Point Margin Expansion
5. 12:30AI Risks: Nuclear Regulation, 25 Gigawatts of Realistic Capacity, and Rising Rates