Google is set to launch a refrigerator-sized satellite on October 1 to test the viability of running AI infrastructure in orbit using solar power.

Google's 'Suncatcher' project will reach a milestone on October 1 with the launch of an experimental satellite named 'MVP.' The refrigerator-sized craft is scheduled to lift off from Vandenberg Space Force Base in California aboard a SpaceX Falcon 9 rocket. The mission aims to test the performance of Google's Tensor Processing Units (TPUs) in space, marking a step toward the company's goal of running AI infrastructure in orbit.

Technical Challenges in Orbit

Operating AI hardware in space presents significant obstacles, including cosmic radiation, extreme temperatures, and the lack of air for cooling. While traditional space hardware is built to withstand these conditions, Google is testing its standard ground-based TPUs. To manage heat, the satellite utilizes a thermal interface material connected to aluminum and copper heat pipes, which transfer heat to a radiator. Due to cooling limitations, the system is designed to run in 15-minute intervals before requiring a shutdown.

Long-Term Viability and Scaling

Experts and company officials acknowledge that orbital data centers face substantial economic and logistical hurdles. Google manager Travis Beals noted that approximately 10,000 satellites would be required to match the capacity of a single 1-gigawatt ground-based data center. Additionally, launch costs would need to decrease to roughly $200 per kilogram to make the model financially viable. Jeff Bezos has estimated that it could take up to 20 years for orbital data centers to become more cost-effective than those on Earth.