Tech Giants Race to Build AI Data Centres in Space

The Orbital Computing Gold Rush

A new frontier in artificial intelligence infrastructure is opening up hundreds of kilometres above Earth's surface. Several major technology companies and ambitious startups are racing to build data centres in space, driven by the enormous and growing power demands of AI systems that are increasingly straining terrestrial electricity grids.

The biggest move came on February 2nd when SpaceX confirmed its acquisition of Elon Musk's AI company xAI, creating a combined entity valued at approximately one point two five trillion dollars. The merger was swiftly followed by an application to the Federal Communications Commission seeking approval to launch up to one million satellites designed to operate as orbital data centres powered by near-constant solar energy. Musk declared that space-based AI is "clearly the only method to scale" and predicted space would become the most economically advantageous location for AI within thirty to thirty-six months.

Google and the Global Field

Google unveiled Project Suncatcher in November 2025, a research initiative to deploy its Tensor Processing Units aboard solar-powered satellites in collaboration with Planet. Prototype satellites are planned for early 2027, aiming to harness sunlight that makes orbital solar panels up to eight times more efficient than ground-based alternatives.

Phantom Space, led by SpaceX co-founder Jim Cantrell, is developing Phantom Cloud, a sixty-six satellite constellation for cloud computing in orbit. Meanwhile, India entered the race when Chennai-based Agnikul Cosmos announced a partnership with cloud provider NeevCloud to host AI data centres in low Earth orbit, with a pilot launch planned before the end of 2026 and ambitions to scale to over six hundred orbital edge data centres within three years.

Sceptics Sound the Alarm

Despite the momentum, significant technical challenges remain. Cooling is perhaps the most fundamental barrier. In the vacuum of space, heat can only dissipate through radiation, requiring massive radiator arrays that add weight and complexity. Space debris expert Hugh Lewis called the million-satellite proposals "somewhat optimistic, premature, and potentially naive."

Analysts at MoffettNathanson calculated that SpaceX's vision could require annual capital expenditures reaching five trillion dollars. A one gigawatt orbital data centre might cost forty-two point four billion dollars, nearly three times its ground-based equivalent. Proponents counter that unlimited solar power, passive cooling, no land constraints, and faster-than-fibre laser links make the investment worthwhile in the long run.