On June 8, Elon Musk presented AI1, the first satellite designed to run artificial intelligence workloads directly in orbit. Together with SpaceX engineer Ian Dahl, he broke down the spacecraft’s architecture, showcased two factories that will manufacture it, and outlined a timeline under which the constellation could eventually grow to roughly one million units.
The presentation took place during the week leading up to SpaceX’s stock market debut, a listing that is aiming to become the largest IPO in history. SpaceX is scheduled to go public on June 12 at $135 per share, offering 555.6 million shares and raising $75B — the largest IPO fundraising amount ever recorded in global markets. The company as a whole is being valued at approximately $1.77T.
At the IPO price, Musk’s personal net worth would rise to around $988B, bringing him within reach of becoming the first person in history with a $1T fortune.
How AI1 Is Designed
AI1 carries solar arrays with a span of 70 meters and supports up to 150 kW of computing power — roughly equivalent to the power consumption of a single Nvidia GB300 rack in a conventional ground-based data center. Musk described the spacecraft as simpler than Starlink: essentially a combination of solar panels, a radiator system, and several laser communication links.
The satellites will operate at an altitude of approximately 600 kilometers, connecting both with one another and with the Starlink constellation through laser communications. Excess heat will be released through deployable liquid-cooling radiator panels extended into space.
Production and Timeline
SpaceX plans to split manufacturing between two facilities. The Gigasat factory in Bastrop, Texas, will be responsible for solar panel production. Manufacturing there is expected to begin by the end of 2027, with output exceeding 1,000 satellites per year.
The second facility, Terafab, is being developed in partnership with Tesla and xAI. It is expected to produce between 100B and 200B AI chips annually using a 2-nanometer process technology. The first phase of the project is estimated to cost approximately $55B, while total investment could reach as much as $119B.
By the end of 2027, Musk expects the network to reach 1 GW of orbital computing capacity per year, followed by exponential scaling toward 1 TW. At the same time, he urged investors to treat those figures cautiously: the IPO prospectus presents a more conservative outlook and pushes the start of large-scale deployment to 2028.
Why Move Computing Into Space
The company says the shift of computing infrastructure into orbit is driven by growing energy constraints facing ground-based data centers. In its IPO filing, SpaceX estimates the AI market at $26.5T and argues that continued growth will eventually run into the rising cost of electricity and water on Earth.
According to the company, solar energy in orbit eliminates that bottleneck.