On October 1, Google’s refrigerator-sized Project Suncatcher satellite reached a sun-synchronous orbit. The company hopes the research project will pave the way for solar-powered AI data centers in space, US public broadcaster NPR reports.
A SpaceX Falcon 9 rocket carried the spacecraft to orbit. Planet, a satellite imaging company that helped Google develop it, will handle commissioning. Google will then power up and test four of its own Tensor Processing Units (TPUs) — machine learning chips the company designed in-house and already runs in its data centers on Earth. The mission is planned to last a year. Project Suncatcher lead Travis Beals called it “a very minimal test” meant only to confirm the chips can run in space. Google announced the mission on September 24, a week before launch.
Heat management limits mean the chips will run Google’s open-weight Gemma model in 15-minute sessions, answering only simple queries. In a blog post, Google said the mission will show how the TPUs handle “the physical stress of spaceflight and the radiation and thermal extremes of space.” Before launch, the prototype went through testing in a thermal vacuum chamber that simulates the space environment.
Demand for AI services has surged in recent years, and with it the need for data centers. At Google’s developer conference in May, CEO Sundar Pichai said AI demand is outstripping supply. He expects the company’s capital spending to reach $180 billion to $190 billion this year — more than six times its 2022 level. At the same time, public pushback against power-hungry data centers on the ground keeps growing. That makes Google one of a handful of companies looking to move data centers into orbit, where solar energy is virtually unlimited and free, and there are no protesters.
Follow Mediasat on Google News: the latest in TV, telecom and tech
“The sun puts out almost all of the power in our solar system. All of the other power sources that humanity has tapped into are just a tiny fraction of a percent,” Beals explained. In a sun-synchronous orbit, the satellite’s solar panels are almost never in Earth’s shadow, so they generate power with few interruptions. That let engineers skip heavy batteries and backup power on board.
Google isn’t the first to test the idea of computing in orbit. In November 2025, startup Starcloud launched a spacecraft carrying an Nvidia H100 chip and ran a version of Google’s Gemini model in space. Elon Musk’s SpaceX is also working on space data centers and expects to start deploying “orbital AI compute satellites” as early as 2028.
That same November, Google unveiled the Project Suncatcher concept, and the new prototype is meant to put it to a real-world test. Down the line, Google wants to fly satellites in clusters that work together as data centers in space. Each satellite would carry dozens of TPUs and use lasers to communicate with the others and with Earth. Next year, Google plans to launch two more satellites to test the links between them.
Brandon Lucia, a professor of electrical and computer engineering at Carnegie Mellon University, says the idea sounds “very sci-fi,” but much of the technology has already been proven at the research level. Heat, he says, is one of the biggest hurdles. On Earth, chips are cooled with air or water — something that’s impossible in the vacuum of space. Google has also called cooling “a crucial research challenge” and is working on approaches that include pipes and radiators to draw heat away from the chips. Beals notes that the radiators are among the heaviest components on the current spacecraft, and heavier payloads cost more to launch. His team is now looking for more efficient solutions.
Lucia also points to maintenance. In a data center on the ground, a technician can swap out a faulty chip. Doing that hundreds of miles above Earth is far harder. “Those all have their cost and complexity amplified by a factor of 10, maybe a factor of 100. And so there has to be a big payoff,” Lucia said.
There are environmental questions, too. Space data centers would run on solar power, but the rockets that launch them burn hydrocarbon fuel. And once retired, the satellites will reenter the atmosphere and burn up, potentially adding to pollution.
Musk has called space the “only way to scale” data centers to meet future AI demand. But the economics remain uncertain. Launches are expensive, repairs are difficult, and laser links across hundreds of miles and through the atmosphere are slower. On top of that, it would take a large constellation of satellites to match the computing power of a single data center on the ground.
Beals says his small team thinks about these questions a lot. “It doesn’t help a lot if this is technically possible, if it’s always going to be too expensive to be practical,” he said. He won’t predict when space data centers will make economic sense. “I don’t see this being something where it’s cheaper to do this in the next five years. I think it will take longer than that,” the project lead added.
Beals expects data centers on the ground and in orbit to coexist for a long time, since some workloads make more sense on Earth and others in space. He called the whole project a “moonshot” and warned that the road ahead will be long and far from straightforward.
