Google's Project Suncatcher Puts AI Chips in Orbit

Google has put its AI chips in orbit. On October 1, at around 2:32 PM EDT, a SpaceX Falcon 9 launched the first Project Suncatcher prototype satellite into low-Earth orbit aboard the Transporter-18 rideshare mission, which carried 130 payloads. The satellite — internally called MVP, for "minimum viable prototype" — is about the size of a refrigerator, was built for Project Suncatcher in partnership with Planet Labs, and now sits in a dawn-dusk sun-synchronous orbit roughly 650 km (400 miles) above Earth, running Gemini models on custom AI hardware. According to Scientific American, the mission's defining challenge is not what most people would guess: the chips survived radiation testing with flying colors, and it is heat, not radiation, that now determines what the mission can do.

Project Suncatcher is Google's moonshot answer to the AI energy crisis: data centers are colliding with stressed power grids and community backlash, so the company is testing whether the models Gen Z uses daily could one day run on constant sunlight above the atmosphere. The project was unveiled in November 2025 and Google explicitly frames it as a decade-long research effort, not a commercial service. Read more science stories on our Science beat and the GenZ NewZ homepage.

Heat, not radiation, is the real enemy

MVP carries four of Google's Trillium-generation Tensor Processing Units — custom AI accelerator chips — delivering roughly the compute of one terrestrial data-center server for the Project Suncatcher test, with solar panels generating about 1 kilowatt of power. The chips run inference on Google's Gemini models in bursts of roughly 15 minutes, then shut down so radiators can shed heat. The problem is basic physics: in the vacuum of space there is no air for cooling, so heat can only escape as infrared radiation. That means the satellite must rest between compute bursts, throttling the very workload Project Suncatcher is meant to prove.

Radiation, the threat everyone assumed would be the showstopper, turned out to be a non-issue. Before launch, Google tested the chips at UC Davis's Crocker Nuclear Laboratory with a 67 megaelectronvolt proton beam: they survived a 15 krad total ionizing dose — about 20 times the roughly 750 rad(Si) expected over a five-year mission — with no hard chip failures, even though Google chose not to use radiation-hardened chips. As reported by Tech Times, that result flipped the Project Suncatcher mission's focus: cooling, not radiation, now defines what the satellite can attempt. Google confirmed contact with the satellite on the evening of the launch, and Travis Beals, Google's Senior Director of Paradigms of Intelligence, said it was "operating as expected."

Why orbit at all

Google's case for space comes down to sunlight and physics. In a dawn-dusk sun-synchronous orbit, the satellite's solar panels get near-constant sunlight and can generate up to 8 times more power than equivalent panels on Earth, where night, clouds, and atmosphere eat into output. If AI's energy hunger keeps growing — and every major lab says it will — beaming compute down from a fleet of solar-powered satellites starts to look less like science fiction and more like the infrastructure planning Project Suncatcher is built for. Alongside the launch, Google published a peer-reviewed paper in the journal Joule laying out the research and the economics, a signal that the company wants the scientific community to take the idea seriously.

Still, this is step one of a very long road. A two-satellite mission in 2027 will test laser interconnects between satellites — a prerequisite for anything resembling an orbital data center, since chips in space need to talk to each other at high speed without melting. The 15-minute compute bursts are a reminder of how early the technology is: one server's worth of compute, napping between shifts to cool down. Project Suncatcher's MVP is only the beginning. If Project Suncatcher works, the physical footprint of AI could one day move off the planet. If it fails, it proves that even Google cannot escape the basic physics of keeping chips cool in space — and the data centers will keep being built where the power lines are.