Behind Project Suncatcher, our moonshot to put AI in space

Project Suncatcher is preparing to launch a prototype satellite on SpaceX’s Transporter-18 rideshare mission to assess whether Google’s Tensor Processing Units (TPUs) can function in the extreme environment of low Earth orbit. The mission aims to gather data on how the chips withstand radiation, thermal extremes, and the physical stress of a rocket launch, including vibrations up to 100 times Earth’s gravity. The team has conducted rigorous ground tests, including shaking hardware to simulate launch forces and exposing TPUs to proton radiation at UC Davis’s Crocker Nuclear Laboratory, where initial results showed resilience to doses exceeding five-year space mission levels.

The satellite will also test thermal management systems, as TPUs generate significant heat in a vacuum where traditional airflow cooling is impossible. Engineers are exploring radiators and heat pipes to dissipate heat, having validated prototypes in thermal vacuum chambers that replicate space conditions. The mission represents the first in-orbit test of these cooling approaches, which are critical for scaling AI compute in orbit.

Beyond hardware survival, the project investigates whether space offers practical advantages for AI workloads, such as near-constant solar power in low Earth orbit, which can provide up to eight times more energy than on Earth. The team envisions future satellite constellations linked by high-bandwidth laser communications to distribute AI processing tasks across multiple orbiting units.

The initial launch is a foundational step toward a 2027 milestone, when Google plans to test inter-satellite laser communication at high bandwidth over short distances. This phase will assess whether satellites can maintain precise alignment and data transfer rates necessary for clustered AI processing. The broader goal is to determine if scalable, orbit-based AI infrastructure is feasible, with lessons from this mission shaping future designs.