Google launches its first Suncatcher satellite to test AI chips in space

Google is preparing the first in-orbit test of Project Suncatcher, its research effort to explore whether AI computing could one day move to space. A prototype satellite carrying Google's own AI chips, TPUs (Tensor Processing Units), will be launched next week on SpaceX's Transporter-18 rideshare mission. The satellite was developed with satellite imaging company Planet.

What Project Suncatcher is

Announced last year, Project Suncatcher is a long-term research moonshot exploring whether space could host large-scale AI infrastructure. According to Google, satellites in low Earth orbit get near-constant sunlight and can generate up to eight times more solar power than on Earth. The long-term idea is to link clusters of satellites so they can run large AI workloads in orbit.

What the first test will measure

According to Google, the question for this first mission is simple: can its AI hardware work in space? The satellite will gather real data on how the TPUs handle the physical stress of launch, space radiation and extreme temperature changes.

• Launch stress: The ride to low Earth orbit takes about 10 minutes, with forces up to 10 times gravity, and individual parts like the chips can see even more. The team shook the satellite on all three axes to mimic a launch, and Google says the hardware held up.
• Radiation: Solar events and cosmic rays can damage electronics. The team ran AI workloads on TPUs while exposing them to a proton beam at UC Davis's Crocker Nuclear Laboratory. According to Google, its Trillium TPUs survived a total radiation dose higher than they would receive in a five-year space mission.
• Cooling: TPUs produce a lot of heat in a small area, and with no air in space, heat can only be shed through radiators. Google built a new cooling system using heat pipes and radiators and tested it in a thermal vacuum chamber. The real test will come in orbit.

What comes next

Google plans for future satellites to carry dozens of TPUs each and fly in clusters. To do that, they must talk to each other by laser at very high bandwidth over very short distances, which Google compares to hitting a coin-sized target from miles away while both are moving. Two satellites that Planet will launch in 2027 will test these laser links.

Google describes this first launch as a test to see what works and find points of failure. AI data centres in space are still a research idea, and answers to core problems such as cooling and links between satellites will only become clear over the coming years. To judge how realistic the project is, it is worth following the data from this first mission and evaluations from independent space engineers.

Sources: Google: Behind Project Suncatcher, our moonshot to put AI in space, Google Research: space-based AI infrastructure system design, Planet announcement.