Behind Project Suncatcher, our moonshot to put AI in space
Source
blog.google
Date
Key takeaways · AI-distilled
Google says satellites in low Earth orbit get near-constant sunlight and up to eight times more solar power than on Earth, which is the core rationale for testing whether AI compute can run in orbit.
Launch loads reach about 10 g over roughly 10 minutes, and individual TPU chips can see 50 to 100 g. Google says its prototype held up in three-axis vibration tests that mimicked rocket launch frequencies.
In proton-beam tests at UC Davis while running AI workloads, Google reports initial results showing Trillium TPUs tolerated a total ionizing dose greater than a five-year space mission would deliver.
With no airflow in a vacuum, heat can leave only through radiators. Google is testing heat pipes plus radiators, so far only in a thermal vacuum chamber, and plans to refine the design from in-orbit data.
Future clusters need laser links at very high bandwidth over very short distances, the opposite of today's long-range, low-bandwidth space lasers. Google plans to test this with two satellites in 2027.
Why it matters
If AI compute can survive and run reliably in orbit, it reframes long-term assumptions about the power and cooling limits currently bounding AI datacenter growth on Earth.
