Elon Musk's SpaceX is going all-in on Nvidia chips, and the scale of the commitment is unusual even by AI-boom standards.
According to marketscreener.com, SpaceX and Nvidia are teaming up on a new orbital AI data center — computing infrastructure designed to run in space rather than in a warehouse on Earth.
The size of the buildout is what makes it notable. Forkast.news frames the plan as a "10 GW Nvidia Vera Rubin bet," tying it to what it calls a "compute landlord thesis" that scales to $100 billion. Ten gigawatts is power-plant territory; Vera Rubin is Nvidia's chip platform underpinning the deal.
TechRepublic reports that SpaceX's commitment to Nvidia is exclusive — one vendor, one architecture, across the whole program. Its analysis cuts both ways: standardizing on a single supplier's AI infrastructure can accelerate deployments, because everything from the software stack to the hardware integration is uniform, but it also deepens supplier and supply chain risk. If Nvidia stumbles on delivery, there is no second source waiting in the wings.
Why it matters in plain terms: AI's bottleneck right now isn't ideas, it's electricity and silicon. Data centers on the ground compete for grid capacity, land, and cooling water, and that competition is already reshaping utility planning. Moving compute into orbit — where solar power is constant and space is not zoned — is an attempt to route around those limits entirely. A 10 GW commitment also tells the market that Nvidia's demand pipeline extends years out, and that the biggest buyers are willing to lock themselves to one supplier to get there.
The deal is a bet that the physical constraints on AI, not the software, will decide who wins the next decade.