SpaceX's $75B IPO bets on three moonshots: orbital AI data centers
Reusable rockets, a new US chip foundry, and satellite factories must all succeed.
SpaceX is bringing its $75 billion IPO to market Friday, and despite losses and CEO Elon Musk's controversial behavior, institutional investors are oversubscribing with $10 billion blocks. The company's vision now hinges on three hard-tech moonshots: a fully reusable rocket (Starship), a new American chip foundry to produce custom AI silicon, and a manufacturing sprint to build satellites faster than ever. These underpin a plan to deploy orbital data centers that could provide massive compute for AI workloads. However, valuation analyses from Morningstar ($825B) and NYU's Damodaran ($1.2T) fall far short of SpaceX's own $1.8T estimate, with the gap representing the market's skepticism about delivering space-based AI infrastructure at scale.
The heart of SpaceX's AI strategy is dual: it both sells compute to competitors like Anthropic and Google while developing its own models (via acqui-hires from Cursor and the "Macrohard" project for white-collar agents). Musk argues space data centers solve the compute crunch, but experts say they are a decade away. The company's S-1 frames its largest opportunity in enterprise AI ($22.7T market), yet acting as a neocloud raises questions about where value accrues in the AI stack. Whether SpaceX can simultaneously be a compute provider and model builder remains the biggest gamble for IPO investors.
- SpaceX's $75B IPO is oversubscribed despite $1.8T valuation vs $825B-$1.2T analyst estimates.
- Three moonshots required: reusable Starship, new US chip foundry, and fast satellite production for orbital data centers.
- SpaceX's AI business plans to sell compute to rivals while building its own models, creating a conflict in value capture.
Why It Matters
SpaceX's IPO forces investors to bet on space-based AI compute, a decade-out vision that could reshape cloud and AI infrastructure.