Episode Details
Back to EpisodesTOMORROWX — What Austin Said While the Market Fell
Published 5 hours ago
Description
Marco opens from Austin: an arena of AI bulls during the worst week of the selloff, one week after Jensen Huang's resignation from Nvidia, and the summit's two stories — the build-out bull case, and the open campaign by Baker, Gerstner and Musk to keep the door open for Huang's return.
Tom goes around the horn on the builders: Gerstner's Anthropic math — a $100 billion run rate on 5,000 people, only 26% of compute serving customers, an offering roughly 10 times oversubscribed, and his own warning that he may hedge an overshoot; Baker's rationing argument — a slower build-out means a longer cycle, and retail is now the most powerful force in the market; Johnsen's SpaceX cost curve at low-to-mid $30s per watt against a $50–60 industry, with sub-one-year paybacks and a $40 billion GPU facility; and Musk adding half a gigawatt a month at about 60% of consensus cost.
Victor maps the four competing answers to where AI profits land — Gerstner's model maximalism, Solomon's application layer, Rauch's deployment toll, and Druckenmiller's merger of the maps, in which the labs sell tasks, not tokens — and what each map does to Anthropic versus the software layer.
Marco then takes Druckenmiller's session as the positioning headline: flat equities, short copper, capital preservation, every January pillar flipped, and the bear thesis he respects most — politics slash the financing.
Trent closes with the room against our tape: the desk's capitulation file, the SemiAnalysis cap ex slope cut as Baker's rationing arriving in data, the Compound slowdown math, and the discipline close — the bulls in Austin argued about 2030; the tape decides November.