Cancelled
Said on the Dwarkesh Podcast (with John Collison), published Feb 5 2026. Distinct from the capacity target in the June 2026 SpaceX talk (1 GW/yr of space AI compute by end 2027, id 232-adjacent #228): this is a COST-CROSSOVER claim — that orbit becomes the cheapest/most economically compelling place to run AI compute — not a capacity number. Deadline uses the outer 36-month bound (Feb 2029) per the range-selection rule.
On the Dwarkesh Podcast published February 5, 2026, Musk said 'mark my words' that within 36 months — probably closer to 30 — the most economically compelling place to put AI will be space, as Earth's power-generation and cooling limits push large-scale compute into orbit where solar power is continuous. He argued it would then get 'ridiculously better' to be in space. The falsifiable test is whether, by roughly early 2029, orbit is genuinely the cheapest place to deploy AI compute at scale.
Cancelled
Partial
Delivered · LateComment on predictions and back the project. We email you a one-tap sign-in link — no password to remember.
By signing in you agree to keep it civil and lawful. We never sell your email.
Discussion