Google prepares first orbital test for Project Suncatcher

Google is preparing the first in-orbit test of Project Suncatcher, a research programme that could eventually move some AI computing infrastructure into space. In a September 24, 2026 update, Google said a prototype satellite will carry Tensor Processing Units, or TPUs, into low Earth orbit on an upcoming Transporter-18 rideshare mission developed with Planet and SpaceX. The immediate goal is modest: learn whether the hardware can survive launch and operate in orbit.

The test follows ground work on three difficult engineering problems. Google shook the satellite on all three axes to reproduce launch loads, then exposed Trillium TPUs to proton radiation at the University of California, Davis. The company says the chips survived a total radiation dose greater than what they would receive during a five-year space mission. Those are Google’s own test results, not an independent qualification for a production data centre.

Cooling is another constraint. In a vacuum there is no air to carry heat away, so an orbital system must use heat pipes and radiators. Google says it has tested a cooling design in a thermal-vacuum chamber and will use the first mission to see how it performs in space. The company also plans to study high-bandwidth laser links between satellites. A later 2027 test is intended to put two satellites in orbit to examine that interconnection.

Google argues that low Earth orbit could offer near-continuous sunlight and up to eight times more solar energy per panel than installations on Earth. That is a technical rationale, not proof that space-based AI will be cheaper or cleaner. Launches, maintenance, latency, radiation, cooling, communications and orbital safety all remain open questions. Ars Technica reported that the first test is planned for October 1 and will use four TPUs, adding operational detail around Google’s announcement.

The news matters because AI data centres are running into limits on electricity, land and cooling. Suncatcher is not an orbital data centre yet; it is an experiment to establish whether the basic hardware can work there at all. For AI users and companies, the practical significance is longer term: if the experiment succeeds, it could broaden the infrastructure options behind future models, while also creating a new set of engineering and environmental trade-offs.