Google Sends Tensor Processing Units into Orbit for Space-Based AI Compute

Google is launching its first Tensor Processing Units (TPUs) into orbit as part of SpaceX's Transporter-18 rideshare mission. The prototype satellite, built in partnership with Planet, will test how Google's TPUs handle the physical stress of spaceflight and radiation.

In a five-year ground-test, Google ran Trillium TPUs under a proton beam at UC Davis's Crocker Nuclear Laboratory and found they can survive a radiation total ionizing dose greater than what they would receive during a five-year space mission. The team subjected the satellite to sustained acceleration loads of up to 10 times the force of gravity during launch, with individual components seeing 50 to 100 g.

Google's Project Suncatcher aims to explore whether space could host scalable machine learning infrastructure. In low Earth orbit, satellites can generate up to eight times more solar power than on Earth, making it an attractive location for AI compute. The company plans to fly in clusters with dozens of TPU chips and use laser interconnects for communication.

The prototype satellite will not carry any of Google's designs for later satellites, which will have the capacity for scalable machine learning infrastructure. Google publishes no date or time for the launch, but says only that it will occur next week.