Google launched a satellite into orbit last week to test whether artificial intelligence hardware can function in space. Project Suncatcher carries four TPU chips designed to run the Gemma AI model outside Earth's atmosphere. This experiment helps engineers understand the physical limits of data processing in a vacuum. The results will determine if space-based computing becomes a viable alternative to ground servers.
Satellite prototype carries four TPU chips running Gemma AI model in orbit
The satellite launched aboard SpaceX's Transporter-18 mission on October 13, 2024. Google developed the hardware in partnership with Planet Labs. The prototype entered orbit successfully, and ground teams have established contact with the device. Operations are currently performing as expected according to the mission report.
The onboard TPU chips face strict thermal constraints due to the vacuum environment. Each computation cycle lasts only about 15 minutes before the system must pause to cool down. Engineers estimate that SpaceX Starship would need to complete roughly 1,800 launches over ten years to make such space data centers cost-effective. The satellite's physical size is comparable to a standard refrigerator, though exact dimensions remain unconfirmed.
Google plans to test high-bandwidth laser communication between satellites in 2027. This next phase aims to resolve the data transfer bottleneck created by the limited compute windows. The current mission serves as a technology validation test for TPU physical tolerance in space. The project demonstrates early steps toward distributed AI infrastructure beyond Earth.
Discussion
0 comments
Log in to join the thread with a thoughtful take, question, or correction.