Nvidia Is Going to Space With Elon Musk's Million-Satellites Plan. Here's Why It Could Be the Start of a Trillion-Dollar Cycle.

Nvidia (NVDA) just extended its grip on artificial intelligence (AI) computing with plans to go where no chipmaker has gone before: orbit. On Aug. 4, SpaceX (SPCX) and Nvidia announced they are partnering to design the compute payload for the Starmind AI1 satellite, which will run heavy AI workloads from space using Nvidia’s newest hardware. The plan is not a one-off. SpaceX has outlined a long-term vision that could eventually scale to as many 1 million AI satellites. Together, these satellites would work as one giant data center circling the Earth.

Even more than the spectacle, what stands out for Nvidia is the commitment behind this partnership. SpaceX CEO Elon Musk said bluntly that the company has agreed to exclusively use Nvidia GPUs because “the Vera Rubin architecture is the best architecture.” That is about as strong an endorsement as it gets, especially because it comes from the one customer with every reason to build his own chips or look for alternatives.