SpaceX is presenting orbital data centers as a major part of its future business, according to Ars Technica, which cited the company’s financial filing ahead of a planned public offering in June.
The plan is not modest. SpaceX says it wants a constellation of 1 million satellites that together could generate 120 gigawatts of power for data-center services. The company says that power could support tens of millions of frontier-class GPUs, and possibly as many as 100 million.
That is the pitch: move compute into orbit, then sell data-center capacity from a satellite network rather than from buildings packed with servers on the ground. The public numbers, as described by Ars Technica, come from SpaceX and should be treated as company claims until the hardware exists and the service is operating.
The AI1 satellite is the first proposed step
Elon Musk, SpaceX’s founder, had previously described the broader idea, but the shape of the individual satellites remained unclear until June. In a promotional video, Musk and Ian Dahl, SpaceX’s director of satellite engineering, discussed the company’s first orbital data-center design, which SpaceX calls the AI1 satellite.
According to Ars Technica, that video gave the company’s figures for the satellite’s size and power capabilities. Those per-satellite numbers are the part that turns a space-computing slogan into an engineering target, because a million-satellite architecture depends on how much compute and power each unit can actually carry.
The disclosed aggregate goal is also a constraint. A 120-gigawatt orbital network would require SpaceX to launch, maintain and coordinate a constellation far larger than its current Starlink internet network. The filing described by Ars Technica puts data-center services, rather than launch services or spacecraft sales, at the center of SpaceX’s claimed future value.
What SpaceX has actually said
- SpaceX envisions a constellation of 1 million satellites for orbital data-center services.
- The company says the network could generate 120 gigawatts of power.
- SpaceX says that capacity could support tens of millions of frontier-class GPUs, with a possible upper figure of 100 million.
- Musk and Ian Dahl discussed the first version, the AI1 satellite, in a June promotional video.
The unresolved part is execution. The available details establish SpaceX’s ambition and the scale of the claimed system. They do not show that SpaceX has built an orbital data center, launched one, or demonstrated the economics of running GPU infrastructure in space.
This story draws on original reporting from Ars Technica.
