Elon Musk unveils AI1 satellite design to power SpaceX's orbital data centers

Elon Musk unveiled the first detailed design of SpaceX's AI1 orbital data center satellite on June 8, revealing a spacecraft with a 70-meter wingspan, 150 kilowatts of peak AI compute capacity and an interchangeable hardware design that would allow different chip providers to supply the processing payload.

The disclosure came during a 30-minute video shared on Musk's social media platform X, in which the SpaceX chief executive detailed the company's near-term plans across its Starship rocket program, its Terafab chip manufacturing joint venture with Tesla and its ambition to move a significant share of global AI computing infrastructure out of terrestrial data centers and into low-earth orbit.

The AI1 satellite, billed as the first generation of SpaceX's orbital AI compute platform, is designed to operate at 120 kilowatts on average and deliver 70 kilowatts of compute per ton of mass — performance figures the company says make space-based AI compute economically competitive with ground-based alternatives within a few years. The interchangeable compute provider design means the satellites are not locked to any single chipmaker, a deliberate architectural choice that keeps the platform open to whoever produces the most competitive hardware.

Alongside the AI1 reveal, Musk announced the Gigasat factory in Bastrop, Texas — a manufacturing site spanning more than 1,000 acres with over 11 million square feet of planned building capacity — that SpaceX intends to use to produce AI satellites and solar components, including solar cell ingots, wafers and finished panels, by the end of 2027. Two prototype AI satellites are scheduled to launch in early 2027, with a full commercial constellation to follow.

The physics case for orbital data centers, which Musk has been making publicly since February, rests on two structural advantages that space has over any location on earth. Solar power in orbit is constant and abundant, with no weather, no night cycle and no grid dependency. Thermal management, one of the most expensive engineering challenges in terrestrial data centers, is solved passively in space, where heat radiates freely into the cold vacuum without fans, chillers or cooling towers. Musk said in February that within two to three years, space will likely be the most cost-effective environment for generating AI compute.

"You're power constrained on Earth," Musk said at an earlier event. "Space has the advantage that it's always sunny."

The orbital data center concept is the commercial centerpiece of the merged SpaceX-xAI entity that Musk created when the two companies completed their combination on February 2, 2026, in a deal valuing the combined group at approximately $1.25 trillion. The merger gave SpaceX direct ownership of the AI workloads that its satellites would host — closing a vertical integration loop that now spans rocket manufacturing, satellite production, network operations, chip fabrication and AI model development within a single corporate structure.

The Terafab project, announced jointly by SpaceX and Tesla on March 21, anchors the chip supply chain for the orbital data center constellation. The facility, projected at approximately 100 million square feet and designed to be roughly 10 times the size of Tesla's Gigafactory in Austin, aims to produce one terawatt of AI compute annually — a figure Musk said is 50 times the combined current output of all manufacturers of advanced AI chips globally. About 80 percent of Terafab's output is targeted at space applications, with the remaining 20 percent destined for terrestrial use.

The primary chip destined for the satellites is the D3, a processor SpaceX is designing specifically for space environments — built to run hotter than conventional chips and hardened against the radiation levels found in low-earth orbit. Musk described chip production as "the major piece of the puzzle" in the broader orbital computing ambition, saying the Terafab was a prerequisite rather than an option.

The unveiling serves a dual commercial purpose. It provides technical substance to the orbital AI compute narrative that SpaceX has been building since early 2026, and it lands precisely as the company prepares for its initial public offering — reported to be targeting approximately $75 billion in proceeds at a valuation of around $1.75 trillion, which would make it one of the largest IPOs in stock market history. Investors evaluating the offering are being asked to price a company that is simultaneously the world's most active launch provider, the operator of the dominant low-earth orbit satellite internet network, and now the developer of an orbital AI computing infrastructure that does not yet exist at commercial scale.

Google is pursuing a parallel concept it calls Project Suncatcher, an 81-satellite cluster developed in partnership with Planet Labs. China's state aerospace company has separately announced plans for gigawatt-scale space-based data processing infrastructure. The orbital AI computing race, once the exclusive territory of science fiction, has acquired a competitive industrial dimension that multiple governments and corporations are now treating as strategically significant.