SpaceX just crossed a line that says a lot about where the money is right now: it made more revenue selling AI compute than selling rides to space. According to The Verge AI, SpaceX’s AI revenue jumped more than three times to $2.6 billion year over year, driven mostly by deals to rent computing power to other AI firms. For a company built on rockets, that’s a striking reordering of priorities.
What stands out here is the gap between the segments. The Verge AI reports that SpaceX’s space division pulled in $962 million this quarter. Its AI arm nearly tripled that. And in the IPO documents behind June’s record-breaking public offering, the company said most of its future value would come from AI, not from rockets or even Starlink.
What actually happened
- AI revenue hit $2.6 billion, up more than 3x from a year earlier.
- That growth came mostly from compute deals, including one with Anthropic in May and one with Google in June.
- The AI division still lost $1.5 billion this quarter, slightly less than the same period last year.
- Capital expenditures ballooned to $18.37 billion as SpaceX races to build compute capacity.
- Overall net loss narrowed to $143 million.
Elon Musk framed it plainly on the investor call: “We’re building AI compute capacity at scale faster than anyone else, we believe, and we’re significantly improving our AI models.”
Why this matters
This is significant because it puts SpaceX in direct competition with neoclouds like CoreWeave, companies whose entire business is renting GPUs to the AI industry. The story here isn’t that SpaceX invented a great AI model. Its Grok model, as The Verge AI notes, has fallen well behind rivals and drew serious backlash for generating non-consensual explicit images. The story is that SpaceX had the data center capacity, the capital, and the appetite to become an AI landlord.
The path to this quarter is telling. SpaceX originally built compute for its own AI ambitions. When Grok couldn’t keep pace, the company rented that capacity out to others instead. That pivot, from failed model builder to compute supplier, is the same move several hardware-rich companies are making as demand for AI infrastructure outstrips supply.
The bets behind the numbers
SpaceX is also trying to buy its way into a real AI product. It agreed to acquire Cursor, which would give it an enterprise coding tool. The deal hasn’t closed yet. Musk said they’re “close to that” but didn’t want to “jump the gun” on regulatory approval.
Then there’s the sci-fi line item: Musk has floated building data centers in space, and pitched an addressable market larger than US GDP. Grand plans, grand price tags.
Space spending rose too. Development costs in the space division climbed $389 million from a year ago, with Starship as the main driver. Starship matters beyond rockets, because it’s meant to launch heavier Starlink satellites, and Starlink is the only profitable part of the company at $4.2 billion in revenue. SpaceX said it launched 20 of the new satellites, though full deployment of 60 at once still looks some distance off.
What to watch next
Investors gave a mixed verdict. SpaceX beat analyst estimates, according to Bloomberg via The Verge AI, but shares slipped after hours following an early pop. That reaction captures the tension in these results: revenue is growing fast, losses are still deep, and capex is enormous.
A few things worth tracking:
- Whether the Cursor acquisition clears regulators and gives SpaceX a real product, not just raw compute.
- How its compute deals with Anthropic and Google scale, and whether more AI labs sign on.
- Whether that $18 billion capex pace can keep narrowing losses instead of widening them.
The takeaway is simple. One of the most famous space companies on Earth just told the market its biggest bet is AI infrastructure. When a rocket maker’s compute business outearns its rockets, that’s a signal about how much capital the AI buildout is pulling in right now. More detail is available in the original report from The Verge AI.