SpaceX now earns more from AI compute than from its rockets
SpaceX's AI compute arm tripled to $2.6bn in a quarter, outearning its rockets and even supplying Anthropic and Google. What it signals for India's own GPU crunch.
The News
Elon Musk's SpaceX has quietly become one of the largest sellers of artificial-intelligence computing power, and the takings now outstrip what the company earns from launching rockets. In its latest quarterly earnings, SpaceX reported AI revenue of $2.6 billion, more than three times what the division brought in a year earlier, according to The Verge.
The surge came almost entirely from renting computing capacity to other AI firms. SpaceX signed an agreement with Anthropic in May and a second with Google in June, supplying both with the horsepower they need to train and run large models. That casts Musk's company as an arms dealer to the very rivals of his own AI venture, xAI.
Profit has not kept pace. The AI division lost $1.5 billion during the quarter, only slightly narrower than its loss in the same period a year earlier. In the documents SpaceX filed ahead of a planned public listing, the company described this unit as the source of most of its value, a striking claim for a business still burning through cash.
Why It Matters
The figures mark a reversal that would have sounded absurd two years ago: a rocket company earning more from selling compute than from spaceflight. It underlines how completely the AI buildout has reshaped corporate priorities, pulling capital, land and power towards data centres regardless of a firm's founding purpose.
The pattern echoes the neocloud wave, in which specialist providers such as CoreWeave rode surging GPU demand to enormous valuations. The last comparable shift in infrastructure economics was the 2006-2010 cloud transition, when Amazon turned spare server capacity into AWS and eventually into its most profitable arm. SpaceX appears to be attempting a similar pivot, wagering that whoever owns the compute owns the margins, even if today's losses are heavy.
It also signals that AI compute is now scarce enough that firms will buy it from almost anyone. When Anthropic and Google, both of which run vast infrastructure of their own, turn to a rocket maker for capacity, the shortage is real.
Indian Angle
For India, the story lands squarely on its biggest AI bottleneck: compute. Domestic model builders such as Sarvam and Krutrim have repeatedly flagged the shortage of high-end GPUs, and the government's IndiaAI Mission has responded by subsidising access to tens of thousands of graphics processors through empanelled providers. SpaceX's results show just how lucrative that scarcity has become for whoever can supply it.
Indian neocloud and data-centre players, from Yotta Data Services and E2E Networks to CtrlS and the compute ambitions of Reliance's Jio, are chasing the same rent-the-GPU model on home soil. The read-across from SpaceX is twofold: demand is deep enough to build a multi-billion-dollar line of business, but thin or negative margins can persist while capacity is scaled. For Indian investors weighing data-centre bets, that is a caution as much as an invitation.
There is a regulatory thread too. Starlink only recently secured clearance to operate in India, and SpaceX's growing reliance on AI compute revenue reframes what an Indian licensee is really backing. MeitY and TRAI may increasingly find that a firm licensed as a satellite operator is, on its books, an AI-infrastructure company.
FAQ
How much AI revenue did SpaceX report?
$2.6 billion for the quarter, more than three times the figure from a year earlier. The bulk came from selling computing capacity to other AI companies rather than from its core space business.
Is the AI division profitable?
No. It lost $1.5 billion during the quarter, only slightly less than its loss in the same period last year, even as revenue tripled year-over-year.
Which companies is SpaceX supplying?
It signed compute deals with Anthropic in May and Google in June, providing infrastructure to two of the biggest names in artificial intelligence.
Why does this matter for India?
It highlights how valuable AI compute has become, the exact resource Indian startups and the IndiaAI Mission are scrambling to secure, and offers both a template and a warning for domestic data-centre firms.
Where can I read the original report?
The Verge published the earnings breakdown; the link is in the attribution below.
This story was reported by The Verge. Read the full original coverage at The Verge.