Elon Musk confirmed SpaceX will build its next generation of AI data centers exclusively on Nvidia's Vera Rubin architecture. The company is going big, fast. By end of 2026 it wants 2 GW of compute capacity running. By 2027, that number jumps to 10 GW. That's the kind of scale that reshapes supplier relationships in silicon valley.
The choice matters because Vera Rubin isn't off-the-shelf hardware. It's Nvidia's latest flagship for AI workloads, built to handle the kind of intensive training and inference that satellite networks demand. SpaceX's Starmind program needs serious processing power both ground-side and eventually in orbit. Locking into one architecture early signals confidence in the roadmap but also gives Nvidia predictable revenue through 2027. That's valuable when chip demand swings.
The partnership also reveals something about SpaceX's priorities. Ten gigawatts is not a casual number. For context, that's roughly the power consumption of a mid-sized city. Building that much capacity in twelve months means construction is probably already underway, possibly at multiple sites. The company is betting AI becomes as critical to its satellite business as rockets themselves.
Nvidia stock moved higher on the news as traders priced in sustained demand from one of the world's most ambitious infrastructure builders.
