1. SpaceX earnings report at a glance
SpaceX reported net income of $4.4 billion for the third quarter of 2026, the three months ended September 30, on revenue of $14.58 billion, a 161% increase from a year earlier. The SpaceX earnings report shows the gain came mostly from renting AI computing capacity, and Anthropic was its largest single customer, providing about 34.5% of quarterly revenue.
The result marks a sharp reversal for the company. In the third quarter of 2025, SpaceX recorded a net loss of $1.05 billion. Twelve months later, the same reporting period produced a multibillion-dollar profit, a swing driven far less by rockets and satellites than by data centers full of graphics processors.
The quarter is also one of the first that public shareholders have been able to judge in detail. SpaceX listed its shares in June 2026, and its quarterly disclosures are now read not only by space-industry watchers but by investors weighing how much of the company's value rests on artificial intelligence rather than on launch and connectivity.
2. Where the $14.58 billion came from
The company now reports three main business lines, and the balance among them has shifted dramatically. The segment SpaceX calls Super Intelligence, which covers the rental of AI computing power, brought in $7.56 billion. That was about 52% of total revenue and roughly nine times the $813 million the same segment generated a year earlier, a gain of around 830%.
Starlink, the satellite internet service that for several years was viewed as SpaceX's main commercial engine, produced $5.25 billion. Starlink sells broadband to households, businesses, airlines, ships and government agencies through a constellation of low Earth orbit satellites, and it remains a large, recurring source of income.
The space and launch business, the operation most people associate with the SpaceX name, contributed about $1.8 billion. That covers rocket launches for commercial and government customers along with related space services. Measured against the rest of the company, it is now the smallest of the three revenue lines.
Put simply, a company built on reusable rockets now earns more than half of its revenue from leasing computing hardware to developers of artificial intelligence models. Satellites and launches still matter, but the quarter's growth and profit came overwhelmingly from the compute side.
3. Why is Anthropic SpaceX's biggest customer?
Anthropic is SpaceX's biggest customer because it rents a large block of the company's GPU computing capacity to train and run its Claude models. In the third quarter that arrangement was worth about $5 billion, or roughly 34.5% of SpaceX's total revenue, making Anthropic the single largest contributor to the company's sales.
The capacity Anthropic uses consists of Nvidia graphics processing units housed in SpaceX's data centers in Memphis, Tennessee. GPUs were originally designed to render images, but their ability to perform huge numbers of calculations in parallel makes them the workhorse chips for building and operating large AI models. Companies that develop frontier models need enormous fleets of them, and few have the capital or the time to build every facility themselves.
That gap has created a rental market. Firms with data centers, power and chips lease that capacity to AI developers, who pay for the computing time rather than owning the hardware outright. SpaceX has become one of the largest suppliers in that market, and Anthropic has become its most important tenant.
The scale of the relationship extends well beyond one quarter. Anthropic's contractual commitments for this computing capacity could reach as much as $120.5 billion through 2029. If the arrangement runs at or near that ceiling, it would provide SpaceX with a long stream of revenue from a single counterparty.
4. How AI compute rental works
Renting computing power to AI developers resembles a utility business more than a traditional technology product. The supplier builds or leases facilities, secures electricity and cooling, installs thousands of high-end processors and then sells access to that capacity, usually under contracts that run for months or years. The customer gets the processing power it needs without having to build every data center itself.
For a model developer such as Anthropic, the appeal is speed and flexibility. Training a frontier model and serving it to millions of users requires vast amounts of processing, and demand can rise faster than any one company can construct new facilities. Renting capacity from a partner like SpaceX lets the developer scale up while keeping its own capital free for research, staff and products.
For the supplier, the model brings large, predictable payments once the hardware is installed, but it also requires heavy upfront investment. That trade-off is visible throughout SpaceX's third-quarter results, where soaring compute revenue sits alongside capital spending that exceeded the quarter's total sales.
5. Rivals on the stock market, partners in the data center
The relationship carries an unusual twist. Anthropic is itself a publicly traded company, and many investors treat it and SpaceX as competing bets on the future of artificial intelligence. Both stocks are often discussed side by side, and comparisons between the two have grown more intense since each completed its initial public offering this year.
Yet the financial results show that the two businesses are also deeply intertwined. Every dollar Anthropic spends on SpaceX compute shows up as revenue in SpaceX's Super Intelligence segment, while for Anthropic it is a cost of developing and serving its models. Investors who hold one company as a hedge against the other may in practice be holding two sides of the same transaction.
This kind of arrangement is not unusual in the AI industry, where model developers, cloud providers and chip suppliers frequently act as customers, partners and competitors at once. What stands out in SpaceX's case is the size of the exposure relative to the company's total business: a single client accounting for more than a third of a quarter's revenue.
6. The fine print behind the profit
A single profitable quarter does not yet amount to a profitable year. Across the first nine months of 2026, SpaceX still recorded a net loss of $413 million. The third quarter's $4.4 billion profit was large enough to offset most, but not all, of the losses booked earlier in the year.
Spending is also running at a very high level. SpaceX's capital expenditures reached $19.6 billion in the third quarter alone, with the bulk going toward AI infrastructure such as data centers, power systems and GPUs. Capital spending of that size exceeds the quarter's entire revenue, which shows how heavily the company is investing ahead of expected demand.
For investors, the combination is important to understand. The compute business is generating rapid revenue growth, but it requires the company to keep pouring money into physical infrastructure. Whether those investments earn an attractive return depends on customers continuing to rent the capacity at today's scale and prices.
7. What the customer concentration means for investors
Customer concentration is the clearest risk in the report. When one client accounts for about 34.5% of revenue, that client holds considerable bargaining power. Any change in Anthropic's computing needs, its finances or its choice of suppliers would flow through to SpaceX's results quickly, and replacing that volume of demand would not be simple.
The $120.5 billion figure cuts both ways. On one hand, it gives SpaceX unusual visibility into future demand, the kind of long-dated commitment that many infrastructure businesses would welcome. On the other, it deepens the company's dependence on a single customer. It is also worth stressing that an up-to figure is a ceiling, not a guarantee: the amount Anthropic ultimately spends could come in below that maximum.
The arrangement also ties SpaceX's fortunes more closely to the health of the AI sector as a whole. Demand for rented computing power depends on model developers continuing to raise money, attract users and expand their products. If that cycle slows, suppliers of compute capacity would likely feel it early.
8. A different kind of SpaceX
The quarter crystallizes a change in what SpaceX is as a business. For most of its history, the company was defined by reusable rockets, and later by the Starlink network that those rockets made possible. Today, the largest share of its revenue and the main source of its profit comes from operating AI data centers.
That shift reshapes the questions investors are likely to ask. Instead of focusing mainly on launch cadence or Starlink subscriber growth, they now also need to track GPU deployments, data center capacity in Memphis and elsewhere, and the spending plans of a handful of AI customers, above all Anthropic.
The third-quarter numbers show how powerful that shift can be when demand is strong: revenue up 161%, a $4.4 billion profit and a compute segment that now produces more than half of all sales. They also show the trade-offs that come with it, from heavy capital spending to a reliance on one customer for roughly a third of revenue.
9. What to watch next
The next quarterly report will show whether the third quarter's profit was the start of a trend or a high point. Key indicators include whether Super Intelligence revenue keeps growing, whether SpaceX can add large compute customers beyond Anthropic to reduce its concentration risk, and whether capital spending begins to translate into consistent profits across a full year.
Investors will also be watching how much of Anthropic's commitment of up to $120.5 billion through 2029 is actually drawn down, since that spending will go a long way toward determining how durable SpaceX's new AI business proves to be.
