Google to Test TPUs in Space With Project Suncatcher Launch
The company's research moonshot aims to explore whether low Earth orbit satellites could one day support scalable AI workloads with abundant solar power.
AI illustrationGoogle is preparing to launch the first orbital test of its Project Suncatcher, a long-term research effort examining whether space could host scalable machine learning infrastructure.
The prototype satellite carrying Google Tensor Processing Units, or TPUs, is scheduled to fly aboard SpaceX's upcoming Transporter-18 rideshare mission in partnership with Planet. The test will gather data on how the hardware performs under the conditions of spaceflight.
According to a post by Travis Beals, Google's senior director of Paradigms of Intelligence, satellites in low Earth orbit can access near-constant sunlight. This could generate up to eight times more solar power than on Earth. Future plans could involve linking constellations of satellites to manage larger AI workloads in orbit.
The initial mission focuses on basic questions about hardware operation in space. Google conducted vibration testing that simulated a rocket launch, with acceleration loads up to 10 times the force of gravity for the spacecraft and up to 50 to 100 g for individual TPU chips. The hardware held up during these tests.
Radiation poses another challenge. The team tested Trillium TPUs in a proton beam facility at UC Davis’s Crocker Nuclear Laboratory while running AI workloads. Initial results showed the chips can survive a radiation total ionizing dose greater than what they would receive during a five-year space mission.
Cooling represents a key engineering hurdle. TPUs generate significant heat in a small area, but space lacks airflow. Heat can only be diffused via radiators in a vacuum. Google has tested a system using heat pipes and radiators in a thermal vacuum chamber that simulates space conditions. The upcoming launch will provide in-orbit data on this approach.
Future satellite designs would carry dozens of TPU chips in orbiting clusters. These satellites would communicate via lasers to maintain precise positioning and the high bandwidth needed for AI processing over short distances. Google plans to test two satellites in orbit in 2027 to evaluate this technology.
The Project Suncatcher team produced a video series discussing the science, what they hope to learn from the first test, and remaining engineering challenges. The first launch is intended to identify what works, find points of failure and inform future missions.