Here is a fact that should not be true: the world's most valuable space company does not make its money in space.
SpaceX just filed its first quarterly report as a public company, and the numbers say something the mythology never prepared anyone for. Of the $7.81 billion it earned in the second quarter, 55 percent came from selling internet subscriptions. The rockets lost money.
Starlink reportedly brought in $4.3 billion, up to roughly 12 million subscribers, and it was the only part of the company that turned a profit. The launch business — the thing that landed boosters and rewrote the economics of orbit — posted an operating loss.
So set the frame plainly. Revenue of $7.81 billion, a jump of 92 percent year over year, and a net loss of $541 million. Impressive top line, red bottom line. The interesting story is what happens between those two numbers.
SpaceX now reports in three segments: space, connectivity, and AI. Read them together and the company's real shape appears.
Connectivity — Starlink — earned $1.656 billion in operating profit in the quarter. That is the entire profit engine. Everything else is a cost center.
The space segment generated $962 million in revenue and lost $542 million. The AI segment reportedly generated $2.56 billion in revenue and lost $1.26 billion. Combined, space and AI bled $1.8 billion.
Put it in one sentence: Starlink's $1.656 billion profit is quietly bankrolling a combined $1.8 billion loss across rockets and artificial intelligence. The connectivity segment alone supports the other two.
Starlink's $1.656 billion profit is quietly bankrolling a combined $1.8 billion loss across rockets and artificial intelligence. That is not a diversified conglomerate.
That is not a diversified conglomerate. It is a utility with two very expensive hobbies attached.
Now look at where the money is actually going, because capital expenditure tells you what a company believes about its future far more honestly than its press releases.
Total capex in the quarter was $18.369 billion. The AI segment alone consumed $15.828 billion of it — 86 percent of everything SpaceX spent.
The year-over-year growth rates make the priority unmistakable. Space and connectivity capex reportedly grew a little over 20 percent. AI capex grew 2,013 percent. That is not a typo; it is a strategy.
Most of that went to the Colossus II compute cluster and server procurement — the hardware backbone of a company that in February merged with Musk's xAI and now talks openly about putting data centers in orbit.
So the flywheel, stated bluntly: Starlink generates cash, and that cash is being poured into an AI buildout of unprecedented scale, subsidized by broadband subscribers who signed up for internet.
Here is where the "launch business" story gets even thinner. SpaceX reportedly ran 38 launches in the quarter. Twenty-eight of them were internal Starlink deployments.
SpaceX is its own largest launch customer. A large share of what looks like "launch revenue" is the company billing itself — internal settlement, not external sales.
Strip that out and the real external launch revenue was about $648 million, from just 10 customer launches. The rocket business the entire brand was built on is, in cash terms, a modest external operation propping up a segment that still loses half a billion dollars a quarter.
Those losses come mostly from research and development — $1.076 billion in the quarter, reportedly up 55 percent year over year, primarily on Starship V3. That is a spending line, not a triumph. On the earnings call Musk claimed the heat shield problem is effectively solved, which is the kind of statement a public company now has to be judged against later.
None of this means SpaceX is fragile. It reportedly ended the quarter with $100 billion in cash, after $85.7 billion in IPO net proceeds and a $25 billion bond sale. Even burning at the current rate, it can fund these losses for years. That runway is the fault tolerance the market has granted it.
But cash cushion is not the same as accountability, and this is where the story turns from accounting to power.
SpaceX reportedly went public at a fixed price of $135 a share — no bookbuilding, no price discovery, just a number the company set and dared the market to accept. It implies a valuation near $1.8 trillion.
After the IPO, the governance barely changed. Musk reportedly retains more than 85 percent of the voting power, behind high thresholds for shareholder proposals and CEO protections designed so that going public dilutes his control almost not at all.
Translation: public investors bought equity, not influence. They can fund the vision. They cannot redirect it.
And the vision is explicitly a long one. The prospectus talks about reshaping industries on Earth and building billion-dollar markets on the Moon and Mars, while conceding the company loses money and will not pay dividends for the foreseeable future — partly because of a Starship program that has reportedly absorbed something like $15 billion.
Then there is the supply overhang. Rather than a single 180-day lockup, SpaceX built an unusual nine-wave unlock schedule. The first wave reportedly expired days after the earnings report, freeing more than double the current tradable float in one go.
More sellers than buyers is a simple equation. The stock has already fallen — down roughly 6 to 7 percent after the earnings debut, trading beneath its $135 IPO price after a losing streak — and the mechanics of the lockup could push it lower regardless of how the business performs.
The clearest sign that nobody actually knows what this company is worth: eighteen fair-value estimates compiled by the Simply Wall St community span from about $0.50 to roughly $238 per share.
That is not a valuation range. That is a coin flip stretched across three orders of magnitude. When credible estimates reportedly differ by a factor of 476, the market is not pricing a business. It is pricing a belief.
And here is what the belief actually is. Buy SpaceX today and you are wagering that connectivity revenue collected now will fund orbital AI compute and a Mars settlement decades out — a chain where each link is unproven and the man holding 85 percent of the votes is the only one who gets to decide whether to keep pulling.
The Q2 numbers were, on the surface, good: revenue beat expectations, the net loss narrowed from over a billion a year ago. Management even argued that each dollar spent on AI compute pays for itself within a year, pointing to billions in new cloud deals signed in early Q3.
Maybe that's true. But the segment math underneath tells the durable story, and it does not flatter the mythology.
SpaceX is a broadband utility with a rocket program and an AI moonshot bolted on, funded by subscribers, valued on faith, and steered by one person who has made sure it stays that way.
The rockets are the romance. Starlink is the business. And the shareholders, for all their money, are just along for the ride.