The Project Suncatcher M1 prototype has reached low Earth orbit and entered the commissioning phase after Planet Labs established communication with the spacecraft. Launched on a SpaceX Transporter-18 rideshare mission, the refrigerator-sized satellite carries four Trillium Tensor Processing Units, the first of Google's custom AI accelerators to be tested in space to determine if terrestrial hardware can withstand radiation and execute machine learning workloads in orbit.

Google aims to bypass ground-based power constraints by leveraging near-constant orbital sunlight, which can generate 8x more solar power than terrestrial sources. To manage heat in a vacuum, the prototype uses heat pipes and radiators to operate the chips in roughly 15 minute bursts. The company estimates that orbital compute costs could approach terrestrial energy expenses if launch prices drop to approximately $200/kg by 2035.