Google Wants to Send Data Centers to Space—and Just Launched an Experimental Satellite to Test Whether Its Ambitious Plan Will Work

The tech giant aims to use the near-constant sunlight available in low-Earth orbit to power data centers. The experimental satellite, launched October 1, holds four computer chips that can run Google’s Gemini A.I. for 15 minutes at a time

Sara Hashemi | Daily Correspondent

October 2, 2026 4:01 p.m.

Share

Copy linkEmailSMSFacebookXRedditLinkedInBlueskyPrint

Add as preferred source

On October 1, Google sent a refrigerator-size test satellite to low-Earth orbit, taking a major step in its goal of putting data centers in space. The tech giant aims to put the infrastructure to run its artificial intelligence models there so the systems can be powered by sunlight.

Although far from achieving this goal, the company’s current test—part of Project Suncatcher—will assess how Google’s hardware performs in the harsh conditions of space.

“This first launch is about seeing what works, identifying points of failure and applying those findings to future missions,” writes Travis Beals, Google’s senior director of paradigms of intelligence and head of Project Suncatcher, in a September 24 statement.

The experimental satellite, which hitched a ride on a SpaceX rocket, was developed in partnership with the satellite imagery company Planet. It’s carrying four custom chips called tensor processing units (TPUs), which have the computing power of one server in a data center, report Kate Conger and Cade Metz for the New York Times. (The largest data centers contain more than one million servers.) The TPU chips will run for 15-minute stretches, during which they can run Google’s Gemini A.I. and process short queries. Then, they’ll shut down to cool off.

Eventually, the company wants to launch fleets of satellites, each carrying a dozen TPU chips and communicating with one another and with Earth via lasers. The spacecraft clusters will serve as space-based data centers.

The idea of building data centers far above the planet’s surface isn’t unique to Google. Elon Musk, Jeff Bezos and others have proposed the concept as A.I.’s need for computing power grows in the face of community opposition and a strained grid. In 2024, data centers accounted for about 1.5 percent of the world’s electricity consumption—and the percentage might double by 2030, according to estimates by the International Energy Agency.  

In space, these energy monsters can take full advantage of the sun. The near-constant sunlight there can provide up to eight times more solar power than on Earth, according to the statement.

“We have direct access to a wonderful renewable source of energy,” says Alexander Wyglinski, an electrical engineer at Worcester Polytechnic Institute, to Peter Hall at Scientific American.

However, operating a data center in space comes with many challenges. Advanced computer processing causes chips to heat up. In Earth-based facilities, chips are cooled using water and fans. But fans don’t work in space, per the Times, and we can’t send huge amounts of water there.

“In a satellite in particular, using more power to do the computations means dumping more heat into the confined environment inside of the satellite,” Brandon Lucia, a computer engineer at Carnegie Mellon University, tells NPR’s John Ruwitch and Geoff Brumfiel.

So, Google is testing a combination of pipes and radiators to dissipate the heat. “We’ll see how our new TPU cooling system works in space and refine our designs as we learn more,” Beals writes in the statement.

The satellites also need to be resilient to space debris and potential electrical-circuit errors caused by radiation, which blasts anything beyond the protection of Earth’s atmosphere. “When we talk about satellite platforms, and you put a computer into them, you don’t put [in] a computer like the one you have on your desk,” Wyglinski tells Scientific American.

Despite these challenges, Google is optimistic. The company plans to launch two more satellites next year, and it has proposed launching clusters of more than 80 satellites.

“If, five years from now, everything we’ve done has worked perfectly, it probably means we’ve not taken enough risk, and we’ve not learned as much as we could,” Beals tells the Times. “If we’re really successful with this in the long run, this will ultimately be boring, and people won’t think anything at the fact that their Gemini query might be getting served in space.”

Sara Hashemi is a science writer and fact-checker currently based in New York City. Her work has appeared in Sierra, The Body, Maisonneuve magazine and more.