Google to test AI chips in space with first Suncatcher satellite mission

Google’s initial test will therefore focus on a basic but critical question: whether its AI chips can reliably operate in the environment required for such an orbital computing architecture.Google is set to test its artificial intelligence hardware in space for the first time, with its Project Suncatcher research programme scheduled to launch a prototype satellite aboard SpaceX’s Transporter-18 rideshare mission.The satellite, developed in partnership with Planet, will evaluate how Google’s tensor processing units (TPUs) withstand the physical, radiation and thermal challenges of operating in orbit, the company said in a blog post.Google CEO Sundar Pichai also announced the test on X, saying the project would determine whether its TPUs can “survive and operate in space”.— sundarpichai (@sundarpichai)Project Suncatcher is a long-term research effort to explore whether machine-learning infrastructure can eventually be scaled in space.

Google is studying networks of solar-powered satellites equipped with TPUs that could process AI workloads in orbit.The company said satellites in low Earth orbit can access near-constant sunlight and potentially generate up to eight times more solar power than comparable installations on Earth.

In the longer term, multiple satellites could be linked into constellations to handle larger AI workloads.Google explores orbital AI infrastructureThe first mission is intended primarily as an engineering test rather than a commercial deployment.

Google said it wants to determine what works in orbit, identify failure points and use the findings to guide subsequent missions.The company has already subjected its TPU hardware to radiation testing and conducted vibration tests designed to replicate launch stresses.Google said its Trillium TPUs withstood radiation exposure exceeding what they would encounter during a five-year space mission in initial ground testing.Cooling remains another challenge.

Unlike terrestrial data centres, satellites operate in a vacuum where heat cannot be removed through conventional airflow-based systems.

Google is therefore testing heat pipes and radiators to dissipate heat generated by the TPUs.The company also plans to test high-bandwidth laser links between satellites as it works towards a future architecture in which multiple spacecraft could operate as a distributed AI computing system.

It expects to test two satellites and their interconnectivity in 2027.Google first announced Project Suncatcher in November 2025, describing it as a research moonshot to explore interconnected, solar-powered satellites carrying TPU chips.

At the time, it said its initial learning mission with Planet was expected to launch by early 2027.SpaceX also explores AI compute in orbitGoogle’s plans come as SpaceX separately develops its own concept for space-based AI infrastructure.SpaceX’s Starmind programme envisages orbital satellites carrying localised AI compute, using solar power and the thermal properties of space while transmitting results back through high-bandwidth laser links to the Starlink constellation.

He said the company is examining dawn-dusk orbits, in which satellites receive sunlight for almost the entire time they are in orbit.Google’s initial test will therefore focus on a basic but critical question: whether its AI chips can reliably operate in the environment required for such an orbital computing architecture.Google also said the project will proceed through incremental engineering tests, with lessons from the first mission informing future designs rather than representing an immediate plan to build commercial data centres in space.