SpaceX commits to exclusively using Nvidia GPUs including optimized Vera Rubin NVL72 chips for space deployment in 2027.

Elon Musk announced on Tuesday that SpaceX and xAI will exclusively use Nvidia GPUs going forward, stating during SpaceX's earnings call that "the Vera Rubin architecture is the best architecture" and "the best AI computer." Musk emphasized the company's commitment, noting: "We are exclusive to Nvidia," and cited the cable-less design of Nvidia's Vera Rubin NVL72 rack-scale system as significantly superior to standard rack-style designs. He highlighted that Nvidia's approach to compute trays—which eliminates manual cable and hose connections—greatly improves serviceability, assembly speed, and reliability while reducing installation errors and failures.

The exclusivity announcement, while presented as a formal decision, largely formalizes existing practice. xAI has historically used only Nvidia hardware, training multiple generations of Grok on Hopper and, more recently, Blackwell accelerators. Although AMD has demonstrated Grok-1 on its Instinct MI300X accelerators, there has been no public announcement or credible evidence that xAI has evaluated or deployed AMD Instinct or other non-Nvidia accelerators in production. xAI's Colossus supercomputers have relied exclusively on Nvidia accelerators for years.

Beyond terrestrial data centers, Musk revealed an ambitious space deployment plan. SpaceX plans to launch what it calls the Starmind AI satellite—an optimized Vera Rubin NVL72 computer into orbit next year. "With respect to the Starmind AI satellite, which will be essentially an optimized Vera Rubin NVL72 computer, this is not some sort of far future distant thing; we expect to start launching this next year," Musk said, adding that the design represents "a radical simplification of the normal NVL72 rack." SpaceX intends to deploy the system both on the ground and in orbit.

This approach significantly exceeds Nvidia's current space plans. Nvidia's Space-1 Vera Rubin Module is designed to deploy several—perhaps a dozen—individual Rubin AI accelerators in space, offering a total of 36 Vera CPUs and 72 Rubin AI GPUs. An NVL72-scale deployment presents substantially greater technical challenges, leaving open questions about cooling and reliability in the orbital environment.