For twenty-three years, SpaceX answered to no one who could see its books. On Wednesday it filed a four-hundred-page document explaining, in detail, why that was the point.
The filing is an S-1, the prospectus a company must publish before it sells shares to the public. SpaceX chose Nasdaq, the symbol SPCX, and what stands to be the largest initial public offering in history. The headline number is the one Wall Street wanted: $18.7 billion in 2025 revenue, up a third year over year. The number underneath it is the one the company spent two decades not having to print. A loss of $4.9 billion — down from a $791 million profit the year before.
A profitable rocket company became an unprofitable one in a single year. The S-1 explains why, and the explanation is the whole story: in February, SpaceX acquired xAI.
The system was built to disclose nothing
Private was never an accident at SpaceX. It was the design. A company that stays private files no quarterly reports, holds no earnings calls, answers no analyst who wants to know what a Starship test failure cost. For most of the firm's life this was a feature pitched to employees and early backers as freedom — freedom from the short-term tyranny of public markets, freedom to lose money on Mars for as long as Mars took.
The cost of that freedom was paid by anyone who wanted to value the thing. A year ago, the only window into SpaceX was an opaque secondary market where insiders sold slices of stock to one another at prices no outside investor could check against a balance sheet. You bought the narrative — reusable rockets, Starlink subscribers, the trajectory — because the narrative was all there was. The structure rewarded believing.
An IPO inverts that structure. The same act that lets the public buy in forces the company to show everything: every loss, every related-party transaction, every lawsuit it has set money aside to lose. The prospectus is not a pitch. It is a legal obligation to confess. And a company that spent twenty-three years optimized for telling you nothing has, all at once, to tell you everything.
The acquisition that became a line item
What it has to tell you, first, is what xAI cost. Buried in the same filing is the number that turned the empire red: xAI posted a $6.4 billion operating loss on $3.2 billion in revenue in 2025. Run the subtraction. A space business that earned $791 million on its own absorbed an AI business losing roughly eight times that, and the combined entity went four-point-nine billion dollars into the hole.
The acquisition was not framed that way in February. The $250 billion all-stock deal was announced around a vision — data centers launched into orbit, a full stack from rocket to bandwidth to frontier model, AI on demand anywhere on Earth. xAI at the time was burning a billion dollars a month and had just raised at a $230 billion private valuation it could sustain only as long as it never had to publish a margin. Folding it into a profitable rocket company was a way to make that burn somebody else's problem — to give a cash-incinerating AI lab the balance sheet of a company that actually earns money.
The IPO undoes the trick. You cannot acquire your way out of disclosure. The losses you fold into a bigger entity do not disappear when that entity files an S-1; they show up as a line item, audited, in a document the SEC requires you to mean. The orbital data centers the deal was sold on appear in the same filing too — as "unproven technologies" that "may not achieve commercial viability."
The growth story shows up in the risk factors
Grok was supposed to be the part of xAI that justified the price — the consumer wedge, the model riding on X's distribution. The S-1 discloses how that is going. The company set aside $530 million for potential litigation losses, naming Grok's "Spicy" and "Unhinged" modes as presenting "heightened risks." It warned investors that investigations into Grok generating sexually abusive imagery could cost it access to entire markets. The growth story is now a risk factor, in the company's own words, because the SEC obliges it to be.
And the same week the prospectus had to admit all this, the outside record arrived to corroborate it. Federal procurement data showed Grok was used in just 3 of more than 400 publicly identified federal AI use cases in 2025 — behind 234 for ChatGPT, 33 for Gemini, 26 for Claude. The model that was meant to be the empire's frontier is, in the one buyer that matters most for a defense-adjacent company, a rounding error. Private valuations are built on stories about where a product is going. Federal contract logs are built on where it actually went.
Even the founder is in the disclosure. The same filing that gives Musk 85.1% of the voting power — control he keeps no matter how few shares the public ends up holding — also has to list him, by law, as a risk factor: the man whose attention is split across a half-dozen companies, whose conduct the firm cannot fully govern, on whom everything depends.
The whole cohort, walking into the light
SpaceX is not doing this alone, and that is the part that makes it more than one company's bad quarter. The same trading session carried the rest of the AI build-out toward the same exit. Blockchain.com filed confidentially for a US listing. Smart-ring maker Oura filed confidentially too. And the two labs racing SpaceX into the public markets had their own numbers pulled into the daylight: OpenAI generated $5.7 billion in Q1 revenue against an adjusted operating margin of negative 122%, while Anthropic, a billion behind on revenue, edged toward its first profitable quarter.
Back in January, the forecast was that 2026 would be the year of the mega IPO — Anthropic, OpenAI, and SpaceX all taking early steps toward the public markets, a watershed for the AI boom. The watershed has a property nobody emphasized at the time. The same private markets that let ten loss-making AI startups add nearly a trillion dollars in valuation in a single year did so precisely because no one had to print the margins. Going public is the mechanism that ends that. The exit everyone was racing toward is also the disclosure everyone was, for years, racing away from.
This is the reversal. Each of these companies grew up in an environment engineered to convert losses into valuation — where the absence of a public balance sheet was not a gap in the story but the condition that let the story compound. The IPO is the same environment run in reverse. It converts valuation back into losses, on the record, audited, with the company's signature on the page.
For twenty-three years the silence was the asset. The prospectus is the bill for breaking it.
The trick had always been to grow in the dark and let the narrative compound where no one could mark the books. The IPO is the moment that stops — when the document designed to raise capital becomes the document that prices the burn, the litigation, the founder, and the loss the acquisition was supposed to bury. A company that disclosed nothing for two decades disclosed all of it at once, and the first thing the light revealed was that the empire is running at a loss.