SpaceX will build its future AI infrastructure exclusively around Nvidia hardware, with Elon Musk calling the chipmaker’s upcoming Vera Rubin architecture the best AI computing platform currently available.
Speaking during SpaceX’s earnings call, Musk said the company plans to use Nvidia’s Vera Rubin NVL72 systems both on Earth and eventually in space. The first optimized systems are expected to launch aboard SpaceX’s Starmind satellites next year.
The scale of the plan is significant. SpaceX expects to have more than 2 gigawatts of computing capacity online by the end of 2026, with that figure potentially approaching 10 gigawatts by the end of 2027.
The Vera Rubin NVL72 combines 72 Rubin GPUs with Nvidia’s Vera CPUs and high-speed networking into a single rack-scale AI system. SpaceX wants to adapt that infrastructure for orbit, where AI workloads could eventually be processed directly in space rather than sending huge amounts of data back to Earth.
For Nvidia, the commitment is another major endorsement of its next-generation AI platform. It also puts additional pressure on rivals such as AMD, Cerebras, SambaNova and Tenstorrent, which are trying to win a larger share of the rapidly growing AI accelerator market.
SpaceX’s ambitions go far beyond a few experimental satellites. The company has discussed building a much larger network of orbital AI infrastructure, potentially turning its expertise in reusable rockets and satellite deployment into a new way of expanding computing capacity.
