Google Is Sending AI Chips Into Space for Project Suncatcher Test

Google plans to launch a prototype satellite carrying four Tensor Processing Units into low Earth orbit next week, giving its Project Suncatcher initiative its first real-world test in space. The mission, developed with Planet and flying aboard SpaceX’s Transporter-18 rideshare, is designed to measure how Google’s AI hardware handles launch stress, radiation, heat and other conditions in orbit.

Project Suncatcher is a long-term research effort exploring whether space could eventually support large-scale machine-learning infrastructure. Google says satellites in low Earth orbit can receive near-constant sunlight and generate as much as eight times more solar power than comparable panels on Earth.

The immediate goal is much narrower. Google is not trying to demonstrate a complete orbital data center on this flight. Instead, the company wants to collect operational data from its TPUs and use the results to guide later missions.

Travis Beals, a senior director at Google Research, said the first launch is intended to show what survives, expose failure points and help the team improve future designs.

Getting the chips safely into orbit is one of the first hurdles. Google says a roughly 10-minute rocket ascent can expose a spacecraft to sustained forces of up to 10 g, while individual components may briefly experience between 50 and 100 g.

To prepare for that environment, the team subjected the prototype satellite to vibration tests along three axes, reproducing the frequencies expected during launch. Google said the hardware performed better than anticipated.

Radiation is another concern. Outside Earth’s atmosphere, electronics are exposed to solar activity and cosmic rays that can interfere with computing hardware.

Google tested its Trillium TPUs at UC Davis’s Crocker Nuclear Laboratory using a proton beam while the chips were actively running AI workloads. The company monitored radiation-induced errors such as bit flips and said the processors withstood a total ionizing dose above what they would encounter during five years in space.

Thermal management may prove just as difficult. TPUs concentrate substantial heat into a small area, but orbit offers no airflow to help carry that heat away.

Google is testing a cooling design that uses heat pipes and radiators. The system has already been evaluated in a thermal vacuum chamber that reproduces the temperature and vacuum conditions expected in space.

The upcoming flight will provide the first opportunity to see how that cooling system performs in orbit.

Project Suncatcher’s longer-term design calls for satellites carrying dozens of TPUs and operating in clusters. Those spacecraft would need extremely fast links to coordinate AI workloads across the group.

Google plans to use lasers rather than radio for those connections. The challenge is maintaining very high bandwidth over short distances while multiple satellites are moving at the same time.

The company has compared the required precision to striking a coin-sized target from miles away while both points are in motion.

A separate test involving two satellites and a laser link is planned for 2027.

Google is not the only company exploring AI computing beyond Earth. Starcloud sent an Nvidia H100 GPU into space in November 2025, while SpaceX has also highlighted space-based AI computing as part of its broader plans.

For Google, however, the next launch is still an engineering experiment rather than a commercial deployment. Project Suncatcher first has to answer a more basic question: whether high-performance AI hardware can operate reliably enough in orbit to make larger systems practical.

See more here:

This analysis is based on reporting from Google.

Image courtesy of Google.

This article was generated with AI assistance and reviewed for accuracy and quality.