In May, SpaceX agreed to give Anthropic access to roughly 325,000 Nvidia GPUs across its Colossus data centers for $1.25 billion per month. Weeks later, it struck a similar agreement to provide Google with about 110,000 GPUs for $920 million per month. Together, the two contracts could generate about $26 billion for the company annually—more than SpaceX’s entire revenue last year.

With Musk’s ambitious plans for building orbital data centers and establishing a Mars colony likely years away, these deals may help explain how SpaceX can justify a $1.8 trillion valuation after its record-breaking IPO last month.

Renting its unused compute gives SpaceX immediate revenue from infrastructure it has already built, while giving investors another reason to view the company as more than a rocket manufacturer, as Musk has repeatedly positioned it, according to Sean Cray, a senior analyst covering telecom, media and technology at Moody’s.

The business model works because demand for AI compute continues to outpace supply as the AI race heats up and companies build increasingly powerful models. Meanwhile, building a large data center can require years of development and a large investment.

This makes SpaceX’s offering attractive right now, and allows it to charge a premium to customers that need that computing power immediately, Cray told Fortune.

For now, most of SpaceX’s money still comes from Starlink and its traditional launch business. The company generated $18.7 billion in revenue last year, including $11.4 billion from connectivity services and about $4.1 billion from launches and other space-related work. 

Its AI segment contributed another $3.2 billion, but it also recorded an operating loss of roughly $6.4 billion—even as Musk has touted AI as the company’s next money-maker after it acquired xAI in February.

Part of the reason may be that Grok, the company’s large language model, requires massive investment in GPUs, electricity, and other infrastructure to train and operate it. At the same time, Grok has so far lagged behind models created by Anthropic and OpenAI in terms of revenue and capabilities. 

Yet, even if Grok doesn’t manage to catch up to its competitors anytime soon, the compute renting business shows investors the company has a proven path to profitability thanks to its AI segment, said Cray.

“It shows that there’s just different pathways for them to generate revenue in their AI segment,” he said. “It doesn’t strictly have to come from Grok and their AI enterprise applications.”

Better yet, the company also has flexibility over what it does with the hardware. SpaceX can rent capacity to outside customers, but it can also use it to train newer versions of Grok, or redirect it to Starlink’s own internal uses.

A potential pitfall

This caveat could be a blessing and a curse, warned Sridhar Tayur, a professor of operations management at Carnegie Mellon University. Built into the contracts with Google and Anthropic is a 90-day cancellation provision that gives SpaceX the ability to reclaim its compute if Grok’s needs suddenly increase.

But the same clause could also allow its customers to walk away if cheaper compute capacity becomes available, raising the risk that the revenue from this business could be temporary.

“Is it a one-off thing that is not your main line of business?” he asked. “Or is infrastructure as a service going to become a permanent part?”

SpaceX’s competitors may have already taken note of the potential gains that come with renting compute.

Meta is reportedly in talks to lease its computing power to Anthropic in a deal that could be worth up to $10 billion over two years, Reuters reported. Meta is also in talks with Anthropic to lease its computing power, sources told the New York Times.

Still, SpaceX’s prospects seem bright for now. The Wall Street Journal reported Friday that the company was in talks to provide the Pentagon with data-center capacity potentially worth billions of dollars for running AI models.

The deal isn’t final, but if it goes through, it would strengthen the case that selling compute could become a long-term business rather than a temporary one.

Justifying its valuation

Musk’s foresight made it possible for SpaceX to sell its compute in the first place. Back in 2024, xAI brought its first major Colossus cluster online in just 122 days by converting an existing factory. Since then, the company has expanded its Memphis-area complex to roughly 2 million square feet across Colossus and Colossus II, which together provide about 1 gigawatts of compute power, with a plan to eventually include 1 million GPUs.

SpaceX’s valuation depends on enormous future growth. And critics have argued that its market value is highly dependent on long-shot bets that would require massive investments before generating meaningful revenue—if they pan out at all.

On the other hand, the agreements with Anthropic and Google show SpaceX can generate billions of dollars from infrastructure it already owns while it works toward those long-shot goals.

How lasting that business could be will depend on whether the shortage of AI infrastructure persists. OpenAI is working with partners through the Stargate initiative announced last year to build a large network of data centers across the U.S. to aid AI development.

Yet, for now, SpaceX has something nearly every major AI developer wants and can’t easily build fast enough.

“There will probably be other third parties that would be interested in that compute because pretty much at this rate it’s a race to secure that compute capacity,” said Cray.