Google Launches First AI Data Center Satellite Into Space

 

Google Launches First AI Data Center Satellite Into Space

Google has taken its first real step toward building data centers in space. On Wednesday, October 1, at around 2:32 p.m. ET, a SpaceX rocket carried an experimental satellite called Project Suncatcher into low-Earth orbit, marking the first in-space test of the company's plan to run AI computing from beyond Earth.

What Just Launched

The prototype satellite carries four of Google's Tensor Processing Units, or TPUs, the custom chips the company uses to train and run its AI models on Earth. According to Google, the chips will run a version of its open-weight Gemma AI model in short 15-minute bursts, answering simple queries before shutting down to cool off, since the satellite cannot rely on the constant cooling systems available in a ground-based data center.

Google developed the satellite in partnership with Planet, an aerospace and satellite imagery company. The mission is expected to stay operational for about a year, testing how the hardware holds up under the stress of launch and the radiation and temperature extremes of space.

Why Go to Space at All

The core idea behind Project Suncatcher is simple: the sun provides effectively unlimited solar power in orbit, without the land, water, and electricity-grid pressures that increasingly large AI data centers create on Earth. Data centers are projected to consume around 3 percent of the world's electricity by 2030, roughly double their current share, according to the International Energy Agency.

In a blog post, Google said the chips performed "remarkably well" during the test so far, and that early data suggests they could handle more radiation exposure than they would typically encounter over five years in orbit. Engineers also had to design a new approach to managing heat, since there is no air in space to carry it away the way data centers do on Earth.

The Bigger Vision

This single test satellite is a long way from an operational data center. Google's long-term vision involves clusters of satellites, eventually as many as 81 working together, linked by lasers to each other and to Earth, forming distributed computing networks in orbit. The company plans to launch two more satellites in 2027 specifically to test that inter-satellite laser communication.

Google is not alone in this race. The startup Starcloud launched a satellite carrying an Nvidia chip into orbit last November and used it to run a version of Google's own Gemini model from space. Separately, Elon Musk's SpaceX has said it expects to begin deploying its own orbital AI computing satellites as early as 2028, with ambitions for a constellation that could eventually include up to a million satellites.

The Challenges Ahead

Experts caution that space-based computing faces real hurdles. Launch costs, radiation damage, and the difficulty of physically servicing satellites all add significant cost and complexity compared to building on the ground. One analyst estimated those factors could multiply costs by 10 to 100 times, meaning any payoff would need to be substantial.

There are also environmental and scientific concerns. Astronomers have criticized growing satellite constellations for contributing to light pollution that already makes observing the night sky more difficult. And while orbital data centers would run on solar power, the rockets that launch them burn carbon-based fuel, and the satellites will eventually re-enter Earth's atmosphere and burn up, a process that could add its own pollution.

Bottom Line

Project Suncatcher is still an early experiment, not a working data center. But it marks a concrete first step by one of the world's largest AI companies toward a strikingly ambitious idea: moving some of AI's growing computing needs off the planet entirely. Whether that idea proves practical will depend on costs coming down, technical problems getting solved, and the broader debate over the environmental trade-offs of putting more hardware into orbit.