Story
October 2, 2026
Google’s AI chip is in orbit, but its space-data-center dream still needs a Starship breakthrough
Google’s first TPU satellite has reached orbit, proving its Project Suncatcher is no longer just a white paper. Yet the company’s own analysis says affordable orbital computing depends on a Starship launch tempo that remains far from demonstrated.
Google’s Project Suncatcher moved from laboratory ambition to an in-orbit experiment when SpaceX’s Transporter-18 mission lifted a Planet Labs prototype carrying four Google tensor processing units, or TPUs, from California. The launch was the first live test of whether the company’s AI chips can work reliably amid radiation, harsh temperatures and tight power constraints in low Earth orbit.1
Google’s public line is hopeful but measured. Sundar Pichai called the successful flight evidence of progress — “how far we’ve come...and how far there is to go” — while Planet reported that it had made contact with Project Suncatcher and that the satellite was healthy.
2
The immediate mission is modest: after commissioning, the satellite is expected to run its TPU in 15-minute bursts, testing continuous power delivery, cooling and model performance without overwhelming its platform. Google executive Travis Beals put the rationale simply: “there’s no test that’s completely as good as the real thing.”3
That experiment is a long way from Google’s eventual vision: 81 satellites flying in close formation, linked to process AI workloads in parallel. Solar energy is the attraction; in low Earth orbit, satellites can receive near-constant sunlight. But the central obstacle is not merely whether a chip survives space. It is whether rockets can deliver enough hardware cheaply enough.1
Google’s newly published analysis sketches a demanding condition for that future. To push launch prices toward roughly $200 per kilogram by 2035, it estimates Starship would need to loft 370,000 tons over a decade — about 1,800 launches, or 180 annually, assuming 200 metric tons per flight. That is an enormous leap for a vehicle that has never flown more than five times in a year.3
Musk remains focused on Starship’s expansive potential, saying it will enable much larger telescopes in orbit and on the Moon.
4 Google’s launch demonstrates that orbital AI has begun; its own numbers show that turning it into data-center scale still requires a launch revolution.