Google tests orbit-based AI compute with Project Suncatcher

Google has unveiled Project Suncatcher, a research initiative to test whether artificial intelligence compute can be operated from orbit, with two prototype satellites planned for launch in early 2027. The company said the effort would place Tensor Processing Unit AI accelerators on solar powered satellites in a low Earth orbit designed to receive near continuous sunlight, and link the nodes with high speed optical beams to form a distributed, space based data centre.

Google said the project responds to rapidly rising demand for compute and the growing environmental and infrastructure costs of terrestrial AI data centres. The proposed configuration would use a dawn dusk sunsynchronous orbit, where solar panels can remain illuminated most of the time. Google researchers estimate that panels in that orbit could be up to eight times more productive than equivalent ground panels, reducing reliance on batteries and fossil fuel backstops. The firm described the vision as a constellation of compact satellites, on the order of dozens to roughly 80 spacecraft, flying in close formation and connected by freespace optical links capable of tens of terabits per second.