Morningstar put a number on SpaceX this week: $780 billion — half what the company reportedly wants from its IPO. The gap isn't a rounding error. It's the story.
The AI era has quietly stopped being a technology story and become a capital-allocation one. The question that matters now isn't whether the models work — they plainly do. It's whether public markets can absorb the trillions in private value the buildout has minted, and at what price.
Ben Thompson made the cleaner version of the argument this morning, reframing Alphabet as “the Google Capital Company” — a firm whose real edge is no longer search but the balance sheet underwriting the data-center buildout. Set that beside the Economist asking, plainly, whether the stockmarket can swallow Anthropic, SpaceX, and OpenAI at once, and the shape resolves. The scarce input isn't talent or compute architecture anymore. It's capital willing to hold positions this large at valuations this thin on present earnings.
The scarce input isn't talent or compute architecture anymore. It's capital willing to hold positions this large at valuations this thin on present earnings.
Even Tyler Cowen's line that America's next great export is intelligence, sold as a service, is a balance-sheet claim in trade-policy clothing. Exports need financing too.
The obvious objection: valuation gaps close upward all the time. Amazon looked absurd for a decade and was right; maybe Morningstar is the one anchored to the wrong multiples, pricing a road-builder on the earnings of a toll booth. Granted. But that concedes the point rather than refuting it — the bull case is a bet on capital's continued appetite, not on any model release.
When the analyst and the founder disagree by a factor of two, the argument isn't really about SpaceX. It's about whether the market can keep saying yes. That's the question 2026 is actually asking.