Google tests its plan for AI data centers in space with Project Suncatcher

Google's Project Suncatcher launched a test satellite carrying four AI processors to run Google's Gemini models for 15 minutes at a time before requiring shutdown for cooling, marking an early step toward its goal of building an orbital data center constellation. The initiative is driven by surging terrestrial energy demands—data centers are projected to consume about 3% of global electricity by 2030—and the appeal of abundant, free solar power in space where panels generate up to eight times more energy than on Earth, plus the political advantage of moving data infrastructure away from populated areas. Significant technical and economic hurdles remain: scaling to rival current AI data centers (which use hundreds of thousands of chips and draw up to a gigawatt) would require solar arrays spanning kilometers, while vacuum heat dissipation, space debris, cosmic radiation, and the bottleneck of transmitting data back to Earth all pose challenges. The economics are also uncertain, as launch and replacement costs combined with limited bandwidth may erode savings from free solar power, suggesting that the company ultimately achieving space-based data centers at scale may still be years away.

(Source：Scientific American)