Google Tests AI Computer Parts in Space with Project Suncatcher

Google is testing a bold new idea: putting computer parts that power artificial intelligence into space. The company's project, called Project Suncatcher, launched four AI chips into orbit today aboard a SpaceX rocket. Scientists want to know if these chips can survive the harsh conditions of space. The big dream is to one day run full data centers in orbit, powered by the sun's energy.

Data centers are buildings packed with computers that power things like internet searches, videos, and AI tools. They use enormous amounts of electricity. By 2030, data centers could use about three percent of all the electricity in the world — roughly twice as much as they use today. That is a huge strain on power grids and the communities that support them.

Google thinks space could be part of the answer. Solar panels in orbit can collect up to eight times more energy than the same panels on the ground. That is because there are no clouds, no night cycle, and no atmosphere blocking the sunlight. 'We have direct access to a wonderful renewable source of energy,' said Alexander Wyglinski, a professor of electrical engineering at Worcester Polytechnic Institute.

The test satellite carries only four processors. These chips will run Google's Gemini AI models for 15 minutes at a time before they need to shut down and cool off. The end goal is a giant network of thousands of satellites that share AI tasks and beam answers back to Earth. But engineers admit that getting there will be very, very hard.

Space creates some tricky problems for computers. Even though space is extremely cold, heat from computer chips has nowhere to go because there is no air to carry it away. That is why Suncatcher's chips must take breaks, and why the satellites need special cooling panels called radiators. Radiation from the sun can also scramble computer calculations, so systems must constantly check and fix errors.

Before the launch, Google spent months testing its hardware on the ground. Engineers shook the spacecraft to copy the vibrations of a rocket launch. They also fired beams of tiny particles called protons at the chips to see how they handled radiation. 'It's about seeing what works, identifying points of failure, and applying those findings to future missions,' said Travis Beals, a senior director at Google who worked on the project.

Even if the chips work well in space, sending data back to Earth is another big challenge. It is expensive and slow to beam large amounts of information down from orbit. During the Cold War, U.S. spy satellites actually dropped their film in parachutes to get it back to Earth because there was no good way to send it digitally. Today, lasers can send data between satellites, but getting it all the way down to the ground is still very difficult.

Scale is the biggest question of all. The largest AI data centers on Earth today use hundreds of thousands of chips and need up to one gigawatt of power. That is several thousand times more power than the solar panels on the International Space Station can produce. Building enough space-based solar arrays to match that would mean going from panels a few meters wide to panels several kilometers wide — a massive engineering challenge.

There is also the question of cost. Launching satellites is expensive, and they wear out over time and must be replaced. If the cost of launching and maintaining all those satellites is higher than the savings from free solar power, the whole idea may not make financial sense. Unlike the space race of the Cold War — driven by governments competing for national pride — this effort is led by companies that need to make a profit.

Google's engineers say they started out by looking for reasons why the idea would never work. Over time, they began to believe it might be possible after all. Still, experts warn that space almost always turns out to be harder than it looks. 'It's usually the late entrants who finally figured out how to do it effectively,' said MIT professor Kerri Cahoy. The company that truly solves the puzzle of data centers in space may not even exist yet.