Understand
Google's Project Suncatcher explores cooling AI chips in space
Overview
Google's Project Suncatcher is an experimental moonshot aiming to put AI compute into space, powered by sunlight. A major challenge of the project is thermal engineering, specifically how to cool high-powered AI chips (TPUs) in the vacuum of space where air cooling is not possible. Researchers are utilizing heat pipes, pumped fluid loops, and large radiators to transfer and dissipate heat, and are testing their solutions in thermal vacuum chambers on Earth.
Key points
- Project Suncatcher is a Google moonshot to operate solar-powered AI compute in Earth's orbit.
- Without air in space, traditional terrestrial data center cooling methods cannot be used.
- Google researchers rely on radiation heat transfer, heat pipes, and pumped fluid loops to move heat from the spacecraft's interior to external radiators.
- TPUs generate thousands of watts in a small area, while radiators dissipate a few hundred watts per square meter, requiring large surface areas.
- Solutions are tested on Earth using thermal vacuum chambers that simulate space environments.