Google plans space-based AI data centers powered entirely by sunlight

Written by

Published 6 Nov 2025

Fact checked by

We maintain a strict editorial policy dedicated to factual accuracy, relevance, and impartiality. Our content is written and edited by top industry professionals with first-hand experience. The content undergoes thorough review by experienced editors to guarantee and adherence to the highest standards of reporting and publishing.

disclosure

google space ai data centers

Google wants to build artificial intelligence (AI) data centers in space powered by the sun, launching its first test satellites with partner Planet by early 2027 to escape the energy crisis plaguing tech companies on Earth.

Project Suncatcher would place Google’s custom AI chips, known as Tensor Processing Units (TPUs), on solar-powered satellites orbiting 650 kilometers above the Earth. The tech giant announced the research project on Monday.

    “In the future, space may be the best place to scale AI compute,” wrote Travis Beals, Google’s senior director for Paradigms of Intelligence, in a blog post.

    Solar panels work eight times better in space than on Earth, according to Google’s research. They also generate power almost constantly in the right orbit. This endless clean energy could help tech companies dodge a growing problem: AI data centers consume massive amounts of electricity, driving up power plant emissions and utility bills.

    But for this to work, satellites must fly within kilometers of each other to maintain the ultra-fast connections AI processing requires. Google says it needs speeds of “tens of terabits per second” between satellites.

    “Achieving this kind of bandwidth requires received power levels thousands of times higher than typical in conventional, long-range deployments,” Google stated, having already achieved 1.6 terabits per second in ground tests.

    Space radiation also poses another threat to electronics. Google blasted its Trillium TPU chips with proton beams to test their durability. The chips survived nearly three times the radiation dose expected during a five-year mission before showing memory problems.

    Money remains the biggest obstacle. Launch costs need to drop from current prices to about $200 per kilogram by the mid-2030s for space data centers to compete with Earth-based facilities.

    Planet, a satellite company, will build and operate two prototype satellites for the 2027 test mission. The demonstration will verify whether Google’s AI chips can function in space and validate the optical communication links between satellites.

    Google is not the only one contemplating the idea. Jeff Bezos predicted last month that gigawatt-scale data centers will orbit Earth within two decades. Elon Musk announced Saturday that SpaceX plans to build orbital data centers.

    Hewlett Packard Enterprise already sent computers to the International Space Station. Its Spaceborne system operated for nearly two years but suffered failures in power supplies and solid-state drives.

    Google’s moonshot recalls its previous ambitious projects, as the company spent over a decade developing quantum computers and fifteen years creating self-driving cars that became Waymo.

    Environmental concerns and community opposition to terrestrial data centers continue mounting as AI’s appetite for power grows. Space offers unlimited solar energy without the land use, carbon emissions, and neighborhood conflicts that plague Earth-based facilities.