Google to test AI tensor chips in orbit with Project Suncatcher satellite
Google's Project Suncatcher is preparing to send a prototype satellite on SpaceX's Transporter-18 rideshare mission, scheduled for October 1 aboard a Falcon 9. The satellite, co-designed with Planet, will carry four Google tensor processing units to study how they handle radiation, temperature swings, and other conditions in orbit. The mission will also test a cooling system using heat pipes and radiators, since a vacuum prevents the convective cooling used in Earth-based data centers.
Google's Project Suncatcher aims to learn whether AI accelerators can operate beyond Earth. A prototype, built with Planet, will ride SpaceX's Transporter-18 rideshare on a Falcon 9, carrying four tensor processing units. The experiment will gather data on radiation exposure, thermal extremes, and launch stresses.
The spacecraft will also trial a cooling design combining heat pipes and radiators, since vacuum conditions rule out the air-based convection used in terrestrial data centers. Ground tests suggested the chips tolerate radiation beyond a five-year mission's expected dose, and the concept rests on low Earth orbit's near-continuous sunlight, which can yield far more solar power than ground panels.
If orbital computing proves viable, it could reshape how cloud and AI services are powered, potentially affecting data-center operators, energy grids, satellite manufacturers, and researchers who rely on large-scale computation. Space-based solar power and cooling constraints may create new engineering markets, while radiation-tolerant chips could influence future hardware design. For everyday users, benefits may arrive indirectly through cost, latency, or sustainability changes—though this first mission is only a small technical test, so near-term societal effects are likely limited.