SpaceX plans to deploy its orbital data center concept, dubbed Starmind, with initial satellite launches slated for 2027. Revealed by Chief Financial Officer Bret Johnsen at a Goldman Sachs conference on Sept. 10, the initiative aims to scale heavy AI workloads into space, building upon the company’s acquisition of xAI in February 2026.
From Terrestrial Constraints to Orbital Scale
While xAI currently relies on traditional, land-based data centers, SpaceX is aggressively mapping out a parallel track in low Earth orbit. The company is targeting 5 to 10 gigawatts of terrestrial data center capacity by 2027, with further scaling scheduled for 2028 as the orbital infrastructure matures. This grand design broke cover in January when SpaceX filed an application with the Federal Communications Commission detailing a massive constellation of up to 1 million satellites dedicated to orbital computing.
Technical specifics emerged during a June 2026 company update featuring CEO Elon Musk. The presentation outlined the foundational satellite architecture, designated AI1. Each AI1 unit pairs a high-performance computing platform with sprawling solar arrays and large thermal radiators. Initial operational parameters point to an average compute power of roughly 120 kilowatts per satellite. Manufacturing of these specialized units is slated for a new Gigasat facility located in Bastrop, Texas.
The Physics of Heat Rejection in Space
However, Musk brushed off these concerns during a Terafab event in March, downplaying the difficulty of space-based thermal management.
“For some reason there’s been a bizarre debate about radiators in space,” Musk stated during the event. “It’s safe to say SpaceX knows how to do heat rejection in space with 10,000 satellites in orbit.”
That milestone was officially crossed in March, when SpaceX successfully launched its 10,000th Starlink satellite.
Economics and Industry Skepticism
OpenAI CEO Sam Altman has publicly criticized the orbital data center concept, questioning whether it can deliver meaningful AI compute capacity at a realistic price point. Altman highlighted runaway launch expenditures and the operational nightmare of on-orbit hardware maintenance as severe roadblocks.
Musk counters that space-based computation offers a distinct economic advantage rooted in infinite, unobstructed solar power. The financial equation hinges entirely on driving down launch costs via the heavy-lift Starship vehicle.
“As soon as the cost to orbit drops to a low number, it immediately makes extremely compelling sense to put AI in space,” Musk argued.
Despite the skepticism from competitors, enterprise validation is already materializing. SpaceX has secured multibillion-dollar contracts for AI computing power from prominent industry players, including Anthropic, Google, and one undisclosed customer. As capital pours into the venture, 2027 will test whether orbital infrastructure can successfully migrate from regulatory filings and promotional slide decks into the harsh reality of low Earth orbit.
