Elon Musk's SpaceX plans to launch satellites powered by Nvidia processors, in what would be one of the first serious attempts to run heavy-duty artificial intelligence workloads off the planet.
According to Tom's Hardware, the effort — tied to an entity it refers to as SpaceXAI — will deploy standalone Nvidia Vera CPUs to handle "agentic" workloads for Grok, the chatbot and AI model family associated with Musk. The report says the hardware will take the form of an optimized version of Nvidia's Vera Rubin NVL72 system, flying aboard a satellite called Starmind. Nvidia claims the chip finishes agentic, reinforcement-learning and data-processing tasks up to 1.8 times faster than conventional x86 processors.
On timing, Livemint reports that Musk said the first Nvidia-powered AI satellite is slated for launch in the fourth quarter of next year, with the fleet expected to reach significant scale by 2028. Livemint frames the broader ambition as an orbital data centre debut in 2027.
Coverage elsewhere has cast the move as a milestone for the chipmaker itself: Techzine Global's write-up is headlined "SpaceXAI sends Nvidia's Vera Rubin into space," while XTB.com summarized it as "Nvidia Goes to Space," noting SpaceX is building satellites powered by Nvidia's chips.
A few caveats are worth keeping in mind. These are announced plans and vendor performance claims, not delivered results, and the sources here do not spell out how many satellites are involved, what they cost, or how the power and cooling problems that make orbital computing hard would be solved.
It matters because AI's growth is increasingly limited by where you can physically put data centers — land, power, cooling — and putting the computers in orbit is a bet that space, not silicon, is the next bottleneck to break.