SpaceX pulled the curtain back on its first-generation orbital data center satellite late Monday, posting a video to X that showed a craft the company calls AI1, a machine designed not to beam internet to your phone, but to run artificial-intelligence computation from orbit.
CEO Elon Musk, speaking from SpaceX's facility in Bastrop, Texas, framed the satellite as a natural extension of hardware the company already builds. And he offered a characteristically blunt engineering assessment: the AI satellite is, in his view, a less complicated machine than the Starlink birds SpaceX has been launching by the thousands.
The reveal lands days before SpaceX's anticipated market debut, described by Yahoo Finance as slated for this Friday. The timing is no accident. SpaceX is showing prospective investors that its ambitions stretch well beyond broadband, into the red-hot market for AI compute infrastructure, where demand has outstripped supply on the ground for years.
The satellite is big. AI1 has a wingspan of 70 meters, roughly 230 feet, and a deployed height of 20 meters, or about 66 feet. It carries a single compute module, a 150-kilowatt solar array, laser links for satellite-to-satellite communication, and a deployable liquid-radiator thermal-management system to dump heat in the vacuum of space.
SpaceX says the solar array delivers 250 watts per square meter, manufactured using solar technology produced at a facility the company identifies as CEW. Peak compute power hits 150 kilowatts; average output runs at 120 kilowatts. The company quotes efficiency at 70 kilowatts per ton.
Those are modest numbers by terrestrial data-center standards. A single warehouse-scale facility on the ground can draw hundreds of megawatts. But the pitch is scale through volume: AI1 satellites, in theory, would form an orbital data center constellation that SpaceX says could eventually reach terawatts of compute.
That is an enormous claim. And Musk, to his credit, hedged it.
Musk described AI1 as SpaceX's "first best effort" to tackle AI compute in space. He made no promises about when the satellites would launch or how large the initial deployment would be. That kind of restraint is worth noting from a figure not famous for underselling.
What Musk did emphasize was the engineering overlap with Starlink. As he told viewers in the video posted to X:
"The AI satellite is essentially a lot of solar cells; you still need some laser links, but you don't have all of the super complex antennas that you have on a Starlink satellite. The easier one to design for is the AI satellite... a lot of this is technology we've already made with the Starlink V3 satellites."
The argument is straightforward. Starlink V3 satellites already use laser inter-satellite links and advanced solar arrays. AI1 drops the complex phased-array antennas that beam broadband to consumers and replaces them with a compute module. Fewer moving parts. Simpler radio architecture. A satellite that borrows its bones from a proven platform.
Whether that simplicity translates to fast, affordable production at scale remains to be seen. SpaceX has a strong track record of iterating hardware rapidly, Starlink went from prototype to a constellation of thousands in a few years, but orbital computing is a different problem set with different thermal, latency, and reliability constraints.
SpaceX also announced a new factory called Gigasat in Bastrop, Texas. The company said it owns or has under contract more than 1,000 acres at the site, with potential for over 11 million square feet of building space.
That is a staggering industrial footprint. For context, 11 million square feet approaches the floor area of the Pentagon, and SpaceX is describing it as potential capacity, not what exists today. But the land is secured, and the company is clearly signaling that it intends to manufacture AI1 satellites at volume, not as one-off demonstration units.
Bastrop sits about 30 miles southeast of Austin, in a part of central Texas that has become a magnet for aerospace and technology investment. SpaceX's decision to anchor a major satellite factory there is a vote of confidence in the state's business environment, and a reminder that when companies need to build big things fast, they tend to pick states that let them.
The reveal was heavy on specifications and light on timelines. SpaceX has not said when AI1 will launch, how many satellites will be in the first tranche, or what compute provider or chipmaker will supply the hardware inside the module. The company has not disclosed costs for the satellite or the Gigasat factory.
Those gaps matter. An orbital data center is only useful if it can compete on price and latency with terrestrial alternatives, or serve use cases that ground-based facilities cannot. Military and intelligence applications come to mind immediately: compute infrastructure in orbit is harder to target, harder to sanction, and harder to cut off from power grids. But SpaceX has not publicly identified its target customers or use cases.
The market debut this Friday will likely force more specifics into the open. Investors tend to want numbers, not just wingspan measurements.
What SpaceX is attempting here is worth taking seriously, regardless of how long it takes to mature. The company has a demonstrated ability to drive down launch costs, iterate satellite hardware at speed, and operate constellations at a scale no competitor has matched. If any private company can put meaningful compute capacity into orbit, SpaceX is the obvious candidate.
The AI compute shortage on the ground is real. Power constraints, permitting delays, water-cooling demands, and political resistance to new data-center construction have slowed buildouts across the country. An orbital alternative, powered by the sun, cooled by radiators in the cold of space, and free from local zoning battles, has a certain elegant logic to it.
Musk himself called AI1 "much simpler than a Starlink satellite." Whether that holds true at production scale, under the thermal and radiation stresses of low Earth orbit, with paying customers demanding uptime, is the question that separates a compelling demo from a viable business.
SpaceX has answered that kind of question before. The company that made reusable rockets routine has earned the benefit of the doubt on hard engineering problems. But doubt, in measured doses, is what keeps engineering honest, and Musk's own refusal to overpromise on AI1's timeline suggests he knows it.
In a country where it takes years of permitting just to break ground on a data center, the fact that a private company is building satellites to bypass the problem entirely tells you everything about where the real bottlenecks are.