Google is moving forward with Project Suncatcher, its program to operate computing hardware in space and eventually train AI systems off Earth. In collaboration with Planet, Google plans to send some of its Trillium TPUs on the SpaceX Transporter-18 rideshare mission.
What Google has tested
Google has performed stress tests on the TPUs to evaluate their resilience to the high g-forces experienced during launch. The company also ran radiation tests and reported that the Trillium TPUs “hold up remarkably well, and can survive a radiation total ionizing dose greater than what they would receive during a five-year space mission.”
A remaining engineering challenge is cooling: computer chips generate significant heat, and radiating that heat away in the vacuum of space is a difficult problem.
Why this matters
Although Project Suncatcher sounds like science fiction, its rationale is practical. AI training and inference demand ever-growing compute capacity and energy. Space offers abundant area and solar power, making it a plausible location to move substantial amounts of computation into orbit.
Google frames Suncatcher as an effort to explore off-Earth locations for AI training and inference as part of longer-term infrastructure planning.
Next steps
The Transporter-18 ride will give the company an opportunity to test hardware under actual space conditions. Google will continue work on space-specific engineering challenges such as thermal management to support longer-duration operations.
Summary
Project Suncatcher is now in a practical test phase: Trillium TPUs have passed multiple environmental tests, including radiation tolerance above a five-year mission dose, but thermal control in vacuum remains unresolved. The initiative could be a step toward relocating significant AI compute into orbit to take advantage of available solar energy and space capacity.
