The $18.7B Company Going Public Without Making Money

SpaceX is about to debut as the biggest IPO in history — and it lost $4.9B last year.

SpaceX filed for Nasdaq today under the symbol SPCX. It's the largest IPO in history, bar none. And somehow this barely registered in a week dominated by Google's Gemini refresh and Nvidia's record earnings. That's because this market doesn't care about rockets anymore — it cares about compute.

The $18.7B Question

Here's what's strange: SpaceX brought in $18.7B in 2025, up 33% year-over-year. That's not a startup number — that's a Fortune 500 number. And it burned through $4.9B anyway. Contrast that with 2024, when it turned an $791M profit. The delta is $5.7B in shifted economics, going largely into capital expenditures that jumped from $11.2B to $20.7B.

Twenty point seven billion dollars in CapEx in a single year. For a rocket company. Read that again.

$20.7B in CapEx. For a company that makes rockets. Let that sink in.

That number isn't about Starship production or Starlink satellites — it's about building out infrastructure to sell compute at scale. And the filings hint at who's buying: Anthropic, paying SpaceX $1.25B/month under a compute deal that runs through May 2029, now expanding to include Colossus 2 capacity.

Follow the Money

This is the hidden thread: SpaceX isn't just a rocket company anymore. It positions itself as the physical backbone of AI inference at scale. Every Nvidia chip needs a home, and SpaceX is building those homes.

The S-1 makes clear what's happening. Revenue is growing 33% YoY, but the profit swing from +$791M to -$4.9B tells you exactly where that money went. They're not hiding losses — they're explaining an investment thesis that's entirely backwards-looking relative to the AI market's appetite for compute.

By the time SPCX starts trading on June 12 (per sources), this will either be the most expensive infrastructure play ever built, or the clearest signal that compute demand has outpaced even the most aggressive building schedules. Neither outcome is boring.

The market will decide whether $20.7B in CapEx is visionary or reckless. Given what Nvidia just reported — $81.6B quarterly revenue, Q2 guidance at $91B — the bet on more infrastructure doesn't look crazy. It looks inevitable.

Data via TEXXR