Google's Project Suncatcher Hits Heat Roadblock in Quest for Space-Based AI Compute

Google's Project Suncatcher is an ambitious research effort to explore the feasibility of hosting large-scale machine learning compute in space. On October 1, 2026, a prototype satellite carrying four Trillium Tensor Processing Units (TPUs) will launch aboard SpaceX's Transporter-18 rideshare mission. The test aims to find out whether solar-powered satellite constellations can host AI workloads, but the project faces a significant bottleneck: heat.

The 15-minute ceiling on how long those TPUs can compute before shutting down due to overheating is the real constraint facing Project Suncatcher's engineers. In space, there's no air to convect through and no ambient atmosphere to absorb waste heat, so the only way to shed heat is to radiate it away as infrared light.

Google built the MVP satellite bus with Planet, an Earth-imaging satellite company, rather than developing its own spacecraft platform from scratch. The partnership allows Google to focus on the TPU payload and thermal system while leveraging established satellite manufacturing expertise.