Google reports first operational test of AI inference computing on satellites under Project Suncatcher, moving select latency-tolerant workloads to orbital infrastructure.

Google is preparing to launch a prototype satellite next week as part of the first orbital test of Project Suncatcher, a research initiative designed to explore whether space can provide a suitable environment for large-scale AI computing infrastructure. The mission will assess how Google's AI hardware withstands the extreme conditions of launch, radiation, and temperature fluctuations in orbit.

The satellite will carry Google Tensor Processing Units into low Earth orbit to evaluate their operational performance in space. A key focus of the mission is testing Google's cooling design for power-intensive chips in the vacuum of space, where conventional airflow cooling is impossible. The company plans to use heat pipes with radiators to manage thermal loads. Suncatcher is structured to gather comprehensive data and identify potential failure points rather than demonstrate a fully operational orbital data center.

SpaceX and Starcloud are also pursuing similar plans to deploy data centers in low Earth orbit, aiming to leverage continuous sunlight for powering energy-intensive AI computing and to circumvent terrestrial constraints on electricity availability.

Google has already conducted preliminary testing at a facility at the University of California, where its TPUs demonstrated the ability to withstand high levels of G-force and radiation. However, certain satellite components can only be properly evaluated in orbital conditions. By successfully deploying and testing this prototype, Project Suncatcher would advance Google's longer-term goal of establishing a constellation of chip-equipped satellites in space.