SpaceX is a rocket company in the way Amazon is a bookstore. In its third-quarter filing, artificial intelligence accounted for 52% of income, and the executives spent as much time talking about power and cooling as about launch vehicles. The rockets are still there. They just no longer get top billing.
The numbers, briefly
SpaceX reported $14.6 billion in third-quarter revenue on Tuesday, according to its regulatory filing. Growth came from every segment, but the loudest contribution came from a business line now called "Super Intelligence."
That segment brought in $7.5 billion. It bundles three things: the recently acquired Cursor coding tool, the Grok bot, and a fast-growing neocloud business that now has 1.7 GW of compute. The neocloud is the part the company treats as most important.
The AI business made $431 million in profit. It would have made more, but $1.2 billion in one-time stock-based compensation tied to the Cursor acquisition got in the way. Buying a company is cheap compared with paying everyone who comes with it.
The headline figure for the whole company:
| Metric | Figure |
|---|---|
| Q3 revenue | $14.6 billion |
| Super Intelligence revenue | $7.5 billion |
| Super Intelligence profit | $431 million |
| Q3 net income | $4.4 billion |
| Q3 net income, same quarter last year | $1 billion loss |
| Nine-month net result | $413 million loss |
| Q3 capital expenditures | $19.6 billion |
| Starlink consumer subscribers | 13.6 million |
A profit with an asterisk
The swing from a $1 billion loss a year ago to $4.4 billion of net income is real. It also deserves a careful read.
The top line was lifted by roughly $1 billion tied to the recovery in Bitcoin prices, plus unspecified related-party income. That second item might be attributable to a SpaceX investment in another company, though the filing as described does not say so definitively. A quarter that leans on Bitcoin's mood and an unlabelled related-party line is not quite the same as a quarter of pure operating strength.
The year-to-date picture is more sober. Across the first nine months, SpaceX still shows a net loss of $413 million, so it has some catching up to do if it wants a profitable full year.
The report also says the results were more profitable than investors expected. It gives no consensus figures, so how large the beat was cannot be judged from the material available.
The bill for all that compute
SpaceX spent $19.6 billion on capital expenditures in the quarter. The money is going into a second data center project, Colossus II, and into Starship, the heavy-lift rocket the company sees as essential for deploying a new generation of Starlink satellites and for building orbital data centers.
The next quarter looks heavier. CFO Bret Johnsen said SpaceX would spend $35 billion to $40 billion on new capital investments in the fourth quarter. That is nearly double what a very busy third quarter required.
Musk on the real bottleneck
Elon Musk said he is spending four days a week in Memphis, Tennessee, where Colossus II is being brought online. That is a lot of calendar for a chief executive who also runs rockets.
His argument is that chips are not the main constraint on deploying super intelligence. Power, cooling and infrastructure are. He said he has tasked a team of SpaceX engineers with solving that, and compared the situation to "the New York Yankees playing in Little League." By his account, it is not fair, but it lets SpaceX deploy power and cooling infrastructure far faster than anyone else.
That is a claim about the company's own speed, offered by its chief executive, and the filing summary offers nothing to test it against. Still, it explains why a rocket company thinks it has an edge in data centers. Moving heavy things and solving thermal problems is, loosely, its trade.
Starship, still in the picture
The rocket has not been forgotten. Musk said the next Starship test flight, expected in November, will attempt to recover both the first-stage booster and the second stage. Catching both would be a "massive technology milestone," he said, though the real milestone is reflying both stages.
Starlink, meanwhile, now has 13.6 million consumer subscribers. Musk predicted that the spread of AI agents will push demand for connectivity higher. It is a tidy narrative, in which the AI business sells the need and the satellite business sells the pipe.
The forecasts
Here the tone shifts from accounting to ambition.
- Musk's long shot: He predicted that SpaceX's annual recurring revenue will reach "somewhere between $500 billion and $1 trillion" by the end of 2027, driven by the Super Intelligence segment. The range is wide enough that nearly any outcome other than disaster could be described as inside it.
- Johnsen's guidance: Total revenue of $33 billion to $37 billion in the first quarter of 2027. That is a concrete figure, and it would be more than double this quarter's $14.6 billion.
Reading the two forecasts side by side is instructive. One is a number a CFO has to defend in a filing cycle. The other is a number a founder is free to enjoy.
What to make of it
The takeaway is that SpaceX now reports like an AI infrastructure company that also happens to launch rockets. Half its income is AI, its capital spending is about to leap, and its chief executive is living in Memphis to get a data center switched on.
Whether that is wise depends on things the report does not settle: how much of the $4.4 billion quarter will recur, what the AI segment earns once the one-time costs fade, and whether $35 billion to $40 billion of quarterly spending turns into revenue on the schedule being promised. For now, the company has a profitable quarter, an unprofitable year, and a trillion-dollar target. Investors get to decide which of the three to put on the slide.