Google Suncatcher space data center online with Gemma 4

Google Suncatcher is designed as a long-term bet on infrastructure that would be economical in the 2030s, with the current satellite serving as a testbed for key questions about running AI workloads in orbit. The main technical risk under examination is how space radiation affects TPUs. Radiation can cause "bit flips" in computer memory and corrupt data without triggering typical monitoring tools. Test results on the ground suggest the impact would be minimal for inference workloads and more of a problem for training a model. Ground testing also included particle accelerator trials, but the team now gets to assess those dynamics in reality.

Google's approach differs from other space data center visions. Suncatcher is largely alone in planning high-speed links between satellites to support coherent distributed workloads. SpaceX's data centers are also expected to use laser links, but they would orbit too far apart to share processing effectively. Starcloud and Cowboy Space expect to perform inference loads on individual satellites.

The project's timeline may depend on SpaceX's Starship, which both Google and SpaceX need for their orbital data centers. A peer-reviewed analysis by Suncatcher researchers estimated Starship would need 1,800 flights to achieve learning-driven cost reductions similar to those seen with earlier vehicles.