SpaceX reported its first earnings as a public company this week, and the headline number had nothing to do with rockets. The company that launches more payloads than anyone on Earth made more money selling AI compute and satellite internet than it did from space. Revenue nearly doubled to $7.8 billion, yet the stock still fell sharply. The reaction underscored how differently the market now views a firm that built its name on interplanetary travel.
For a company literally named after space travel, the reordering is striking. SpaceX's AI division brought in $2.6 billion in the quarter, up 247% year over year. Its traditional space business generated just $962 million. As one industry publication put it, SpaceX made more revenue as an AI company than as a space company. The shift is not merely a matter of accounting labels; it reflects a deliberate strategic pivot that began even before the company went public.
The rocket company that became a neocloud
The AI revenue came almost entirely from a business SpaceX barely had a year ago: renting out GPUs. The company signed deals to sell compute to Anthropic in May and Google in June, and told investors that those contracts drove the growth. One buyer, identified in the filing only as “Customer B”, accounted for 19.5% of all SpaceX revenue, a figure that points directly to Anthropic. The concentrated revenue is both a sign of strength and a vulnerability, since the loss of a single customer could reshape the company's financial picture.
This reframing is more than cosmetic. In its record June IPO documents, SpaceX stated that most of its value would come from AI, not rockets. The accounts now read like the financials of an AI company in 2026: fast growth bought with enormous spending. The AI unit still lost $1.26 billion in the quarter, underscoring the gap between revenue generation and profitability. The company is essentially betting that the same infrastructure and deployment capabilities that made it a leader in aerospace can be transferred to the world of data centers and high-performance computing.
SpaceX is not alone in this pivot. A wave of so-called neocloud providers has emerged in recent years, renting out Nvidia GPUs and other accelerators to AI startups that cannot secure their own chips. What sets SpaceX apart is the scale of its ambitions and its existing physical infrastructure. The company already operates massive facilities for satellite manufacturing and rocket testing, and executives have argued that it can build and operate AI data centers with similar speed and cost discipline. Whether that argument holds will be tested by the billions of dollars now flowing into the AI division.
Starlink is still the only part that makes money
Beneath the AI story sits the business that actually pays the bills. Starlink, SpaceX's satellite internet division, saw revenue rise 66% to $4.3 billion. Its subscriber base doubled to 12 million, according to GeekWire, and it was the company's only profitable segment, with $1.7 billion in operating income. For years, Starlink has been described as the cash machine funding everything else, and this quarter confirmed that role. The network now covers remote areas, ships, aircraft, and emergency response teams, and it continues to expand into new markets.
During the earnings call, Elon Musk told analysts that people were “really underestimating Starlink” and predicted it could deliver “a majority of the world's internet” within ten years. President Gwynne Shotwell added that SpaceX would soon take on “the big three” US mobile carriers, a market she valued at $600 billion. That would require a significant expansion of Starlink's direct-to-device capabilities, which are still in their infancy. Competition exists, including from OneWeb and Amazon's Project Kuiper, but for now they remain far behind in both deployment and subscriber count.
Starlink's profitability is all the more important given the losses elsewhere. The AI division's $1.26 billion operating loss would have been far larger without the cash generated by satellite internet. This internal cross-subsidy is a deliberate strategy: Starlink provides steady revenue, while AI beta represents a high-risk, high-reward bet. If AI compute prices fall or competition intensifies, SpaceX can fall back on the satellite business. But if the AI bet pays off, it could dwarf even Starlink's long-term potential.
Why the stock fell anyway
None of the upbeat numbers soothed investors. Capital spending jumped more than sixfold to $18.4 billion, most of it on AI, and well above what Wall Street had expected. CNBC reported that the figure “unnerved” the market, even as executives promised that each AI investment pays for itself within a year. Shares fell as much as 11% during the trading session before recovering somewhat. The market's reaction fits a broader season in which investors have punished heavy AI spenders across the tech sector.
The spending is not just on GPUs. SpaceX is building data centers at a pace that executives concede is ambitious, and it is also pouring money into expanding Starlink's satellite constellation and ground infrastructure. The result is a company that is simultaneously investing for growth in two capital-intensive industries. Wall Street had expected capital expenditures of roughly $3 billion; the actual figure was six times that. Even for a company with SpaceX's track record of audacious bets, the scale of spending gave investors pause.
The stock now trades below its $135 IPO price and near half its June peak, according to the BBC. A separate squeeze is scheduled for Thursday, when a lockup period expires and approximately $101 billion of insider stock can be sold, as Bloomberg reported. Roughly 35% of the tradable float is already sold short. That combination of a looming supply increase and high short interest creates the potential for significant volatility in the coming days.
Musk one-upped his own executives
The earnings call followed a familiar pattern: Musk made vast claims, and his executives walked them back. TechCrunch counted several instances. When the CFO carefully framed a $100 billion revenue run-rate target for December, Musk cut in, saying it “is not a question mark. That's what we would achieve if we basically did nothing.” The remark was meant to project confidence, but it also underscored the gap between the company's internal ambitions and the more conservative guidance favored by its finance team.
Musk compared SpaceX's edge in data centres to “the New York Yankees going in and playing a Little League team”, and called building them “a trivial problem” next to reusable rockets. Then he described building factories on the Moon staffed by robots, which Fortune reported he admitted “sounds totally nuts”. These comments are classic Musk: sweeping, visionary, and sometimes at odds with the details of the financial report. His executives, meanwhile, focused on near-term milestones and operational metrics, a discrepancy that analysts have begun to note.
For all the theatre, one figure will decide whether the strategy works: the gap between revenue and spending. Musk said SpaceX's internal target for $1 trillion in annual revenue had moved up a year, to 2030, with a “non-zero chance” of 2029. Reaching that target would mean growing roughly tenfold in four years, from a company that is still losing money. It would require the AI division to scale far beyond its current revenue base, and Starlink to continue its rapid expansion into telecom markets. Musk has a history of hitting ambitious targets late, but this one is more aggressive than anything SpaceX has previously attempted.
The numbers at a glance
- Total revenue: $7.8 billion, nearly double the previous year
- AI division revenue: $2.6 billion, up 247%
- Space business revenue: $962 million
- Starlink revenue: $4.3 billion, up 66%, with 12 million subscribers
- Starlink operating income: $1.7 billion, the only profitable segment
- AI unit operating loss: $1.26 billion
- Capital spending: $18.4 billion, up more than sixfold
- Stock performance: fell 11% after the report; trading below IPO price
Investors now have a clearer picture of the dual identity at the heart of SpaceX. The company is at once the world's leading launch provider and a speculative AI infrastructure play. The next test is nearer. SpaceX aims to catch a returning Starship with its launch tower this month, a genuine engineering milestone that would demonstrate the same reusable-rocket expertise that executives believe gives them an edge in data centers. But the market has already priced the company as an AI-infrastructure business with a rocket company attached. Its first results did nothing to change that read; they only sharpened it, days before the lockup lifts.
Source:TNW | Spacex News