Google’s Project Suncatcher to put ML infrastructure in space

Google's Project Suncatcher is set to launch a prototype satellite with the goal of testing how its AI hardware can survive the extreme conditions of space. This mission aims to evaluate the performance of Google's AI chips under high radiation levels, intense vibrations, and cooling challenges in a vacuum environment. By following the progress of this mission, which is being showcased in a new video series, the public can gain insights into the engineering challenges ahead.

The primary objective of Project Suncatcher is to determine if space is a suitable location to run AI systems, given the ample solar power available to satellites. This ambitious research moonshot, which began with years of study, seeks to build powerful AI systems in orbit as part of a long-term plan. Once the prototype satellite is in orbit, it will gather data on how Google's Tensor Processing Units (TPUs) perform under the physical stress of spaceflight, including radiation and thermal extremes.

Early tests have shown that TPUs can withstand radiation levels far greater than those experienced during a five-year space mission. However, cooling the chips in space remains a critical challenge, as there is no airflow in a vacuum. The team is exploring various cooling techniques, such as a combination of heat pipes and radiators, to ensure the chips do not overheat.

Communication between satellites will also be a key focus, with lasers being tested to provide high-bandwidth connections over short distances. Future satellite designs will likely carry dozens of TPUs and operate in clusters, enabling larger AI workloads in orbit. The first launch of this mission is focused on gathering data and identifying potential issues, with the ultimate goal of enabling AI's transformative benefits in various fields, from healthcare to scientific discovery.