Elon Musk is about to launch an Nvidia data center into orbit — and he’s betting everything on its chips

SpaceX’s first orbital data center is now due to launch in the fourth quarter of 2027 — about a year earlier than the company told investors when it went public. The company had earlier noted plans to start deploying satellites “as early as 2028” in its initial May Prospectus. Then on the company’s first earnings…


Elon Musk is about to launch an Nvidia data center into orbit — and he’s betting everything on its chips

SpaceX’s first orbital data center is now due to launch in the fourth quarter of 2027 — about a year earlier than the company told investors when it went public.

The company had earlier noted plans to start deploying satellites “as early as 2028” in its initial May Prospectus. Then on the company’s first earnings call on Aug. 4, Musk said launches would start next year. And on Aug. 24 he put a quarter on it: “SpaceX, in partnership with Nvidia, has designed a space-optimized Vera Rubin NVL72 system for launch to orbit in Q4 next year, with significant scale in 2028,” he wrote on X.

Must Read

SpaceX went public in June and is valued at more than $1.8 trillion, up from $7.8 billion in revenue last quarter — a price that assumes a lot goes right. And Nvidia, which reported earnings Aug. 26, is the company selling it the parts.

A Vera Rubin NVL72 hardware is a server cabinet a little taller than a refrigerator, and it’s the newest thing Nvidia sells to data centers. SpaceX wants to strip one down, bolt it to solar panels and fly it in an orbit that keeps it in sunlight nearly all the time. From there, whatever it computes comes back to Earth over the same laser links Starlink satellites use to talk to each other. The company says each satellite will draw about 250 kilowatts at peak, and it has already asked the FCC for permission to launch up to a million of them.

What SpaceX gave up when it went exclusive with Nvidia

Back in June, Musk was saying the satellite’s design could work with Nvidia’s older GB300 chips, and that SpaceX would draw up a version for Google’s TPUs as well. Two suppliers means a buyer can play one off the other on price. But on SpaceX’s first earnings call as a public company, Musk dropped the second one.

“We’ve decided to build exclusively on Nvidia because we think the Vera Rubin architecture is the best architecture,” Musk said. “We’re exclusive to Nvidia.”

That is a lot to give up on a satellite that hasn’t flown, though it may not last. SpaceX and Tesla are planning a chip plant called Terafab, and SpaceX says one of the things it will make is chips for the space data centers, which would end the exclusivity once Musk’s own factory is running.

Source link