Google will send its own AI chips into orbit for the first time next week, in the first flight test of Project Suncatcher. A prototype satellite carrying its Tensor Processing Units (TPUs) will fly on SpaceX’s Transporter-18 rideshare mission.
Google built the satellite with Planet, the Earth-imaging company. The mission will measure how the chips cope with the stress of launch, the radiation of space and its extremes of heat and cold.
Suncatcher, announced last November, explores whether AI computing could one day run on fleets of satellites. Google says panels in low Earth orbit can collect up to eight times more solar power than on the ground.
The hardware has already passed tests on the ground. A trip to orbit takes about ten minutes and can put the chips under forces of 50 to 100 times gravity. To copy that, the team shook the satellite hard on all three axes, and the hardware held up.
For radiation, Google ran AI workloads on its Trillium TPUs inside a proton beam at UC Davis’s Crocker Nuclear Laboratory. The chips withstood a larger total dose than they would receive over a five-year mission, Google said.
With no air in space, cooling is harder, and heat can only escape through radiators. Google is testing a mix of heat pipes and radiators, which has so far worked in a vacuum chamber on Earth.
The next step comes in 2027, when two satellites will test the laser links that clusters of TPU-carrying satellites would need. Google has previously described that test as a joint mission with Planet.
Google is not the only company enrolled in this race as Nvidia-backed startup Starcloud launched an H100 chip into orbit last November and has used it to run Google’s Gemma model.
Cost remains the main obstacle, as Google’s own research estimates that launch prices would need to fall to about $200 per kilogram for space computing to compete on cost.
Get the TNW newsletter
Get the most important tech news in your inbox each week.