SpaceX is stepping into the AI infrastructure race in a big way. According to The Information, Elon Musk’s rocket company has struck a compute deal with Anthropic worth up to $40 billion, a tie-up that puts one of the world’s most valuable private companies behind the maker of Claude.
The arrangement is striking on its face. Musk runs xAI, a direct Anthropic competitor, yet SpaceX is now funneling resources into the rival camp. That tells you how decoupled SpaceX’s treasury decisions are from Musk’s other AI bets, and how hungry every frontier lab has become for guaranteed compute capacity.
What the deal signals
Compute is the new oil, and the contracts keep getting bigger. A few reference points:
- Anthropic recently signed a roughly $30 billion compute commitment with Microsoft and Nvidia.
- OpenAI has stacked multi-hundred-billion-dollar infrastructure pledges across Oracle, Nvidia, and AMD.
- Google and Amazon are already deep into Anthropic’s stack as investors and cloud partners.
Layering SpaceX on top of that pile means Anthropic now has compute relationships spanning every hyperscaler plus a private space and satellite giant. That’s an unusual moat.
Why SpaceX, of all places
This is the part worth watching. SpaceX has cash, Starlink revenue, and a strategic interest in connectivity infrastructure. Funding Anthropic compute, whether through prepaid capacity, direct investment, or a Starlink-adjacent data layer, gives SpaceX a foothold in the AI buildout without launching its own foundation model lab. The Information’s reporting frames the agreement as a long-dated compute commitment, the same structure Anthropic has used with Microsoft, Google, and Amazon.
The Musk angle is impossible to ignore. xAI just raised tens of billions to expand its Memphis Colossus cluster. SpaceX writing checks that benefit Anthropic complicates the narrative that Musk’s empire moves as one. Shareholders and board governance at SpaceX clearly see Anthropic as a defensible bet on its own terms.
What it means for the industry
A few implications worth tracking:
- Compute scarcity is real. Labs are locking in capacity years out because spot availability won’t cover frontier training runs.
- Non-cloud capital is entering AI infrastructure. When a rocket company underwrites model training, the financing pool for compute is officially broader than the Big Three clouds.
- Anthropic’s runway just got longer. Stacked commitments from Microsoft, Nvidia, Amazon, Google, and now SpaceX put the company in a different financial weight class heading into its next model cycle.
- Expect more cross-industry deals. Energy, telecom, and aerospace players sitting on cash and infrastructure are obvious next entrants.
What to watch next
The structure of the deal matters. If it’s pure prepaid compute, it’s a customer relationship. If it includes equity or revenue sharing, SpaceX becomes a strategic stakeholder in Anthropic’s roadmap. Either way, the next round of disclosures, from Anthropic’s revenue updates to any SpaceX filings, will sharpen the picture.
For AI practitioners, the takeaway is simple. Frontier model access isn’t going to get cheaper any time soon. The labs spending tens of billions on multi-year compute contracts are signaling that the scaling curve still has room to run, and they’re willing to pay to stay on it.
Full details at The Information.