Summary
Google says a Project Suncatcher prototype will carry Tensor Processing Units on a SpaceX Transporter-18 rideshare to collect data on how they handle spaceflight. The test is an early step towards the company’s proposed satellite clusters for AI computing.
Google says Project Suncatcher is preparing for its first orbital test: a prototype satellite carrying Google Tensor Processing Units (TPUs) is scheduled to fly on SpaceX’s upcoming Transporter-18 rideshare mission. Developed with Planet, the mission will collect data on how the chips respond to launch forces and the radiation and thermal conditions of space.
The flight is an early experiment in a much larger research idea: using satellites to host machine-learning infrastructure. Google’s proposed future system would connect groups of satellites to handle AI workloads in orbit. The current mission is intended to test hardware and gather engineering data, rather than operate such a cluster.
A test of hardware in orbit
A rocket journey to low Earth orbit takes about 10 minutes, Google says. During ascent, a spacecraft can experience sustained acceleration loads of up to 10 times the force of gravity; individual components such as chips can face forces of 50 to 100 g. Google reports that it shook the satellite along all three axes in vibration tests designed to reproduce launch conditions, and that the hardware held up.
Radiation is another hazard for electronics beyond Earth’s atmosphere. Before the flight, Google tested TPUs under a proton beam at the Crocker Nuclear Laboratory at UC Davis while running AI workloads. The team monitored how radiation-related errors, including bit flips, affected those workloads. Google says initial results showed its Trillium TPUs could withstand a total ionizing dose greater than the amount it estimates they would receive during a five-year space mission. Those results come from laboratory testing; the upcoming flight will provide measurements from orbit.
Keeping the chips cool poses a different problem. On Earth, airflow can carry heat away from electronics. In a vacuum there is no airflow, so heat must be conducted to radiators and released as thermal radiation. Google says it is developing a cooling approach combining heat pipes and radiators, and has tested it in a thermal-vacuum chamber. The orbital test will help the team assess how the cooling system performs in space.
From one satellite to a proposed network
Google’s longer-term concept calls for future satellites to carry dozens of TPUs and operate in clusters. The company argues that satellites in low Earth orbit could access near-constant sunlight and generate up to eight times more solar power than on Earth. That is part of the rationale for exploring orbital computing; the first test is focused on whether the hardware and supporting systems can operate there.
Connecting satellites for demanding workloads is another engineering challenge. Google plans to use laser links, but says its intended system needs very high bandwidth over short distances, requiring satellites to maintain precise alignment while moving. The company plans to test this work by putting two satellites in orbit in 2027.
Project Suncatcher is therefore advancing through staged tests: first checking hardware and cooling in orbit, then working towards satellite-to-satellite links. Whether that approach can support scalable AI computing depends on the results of those engineering steps.