SpaceX announces plans for $74B in US AI data center campuses

In short: SpaceX says it plans to build more than $74 billion worth of new U.S. supercomputing campuses for artificial intelligence.

What happened

SpaceX, best known for rockets, is expanding further into artificial intelligence infrastructure after acquiring xAI. The company said on November 12 that it plans more than $74 billion in new U.S. “supercomputing campuses,” which are large data centers built to run AI.

SpaceX said it plans to spend over $24 billion at Starbase, Texas, and about $50 billion in Cochise County, Arizona. The Texas site is planned to generate its own electricity instead of relying on the public power grid. The Arizona site is a proposal and is not an operating facility.

The spending is not only a future plan. SpaceX reported $16.44 billion in capital spending in the third quarter by its Super Intelligence unit, which includes AI models and the computing hardware that runs them. It also reported 1.7 gigawatts of installed GPU capacity as of September 30, 2026. GPUs are specialized chips that act like high-powered engines for AI (they do the heavy math fast).

SpaceX is also financing equipment for these facilities. It has a reported $40 billion equipment lease commitment tied to Nvidia’s next-generation platform, with deliveries expected to begin in 2027. SpaceX has a company target of reaching 2.5 to 3 gigawatts of compute capacity by the end of 2026, but that target is not the same as a completed build.

Why it matters

AI systems run on huge “warehouses” of computers, plus power and cooling to keep them running, like a factory that needs both machines and electricity. SpaceX’s numbers show how expensive and energy-hungry this buildout is, and they also show the gap between announcements and what is already built.

Source: NYTimes