Google Launches AI Chips Into Orbit to Test Space Data Centers

Google has launched four specialized AI chips, called TPUs, into orbit aboard a SpaceX Falcon 9 rocket. The mission, called Project Suncatcher, is a first step toward testing whether AI hardware can operate in space. The chips are packed into a small satellite the size of a refrigerator, built with Planet, and launched from California on October 1. This isn't about running a full data center in space yet—it's about seeing if the chips can survive the harsh conditions of orbit.

Data centers on Earth are struggling to keep up with AI's massive appetite for electricity, land, and water. Google thinks space could help: solar panels in orbit could generate up to eight times more power than on the ground because they get near-constant sunlight. If successful, some AI computing could eventually happen in space, reducing pressure on Earth's resources. But first, Google must solve big problems: cooling the chips (they overheat in about 15 minutes without air to carry heat away), protecting them from radiation, and building fast laser links between satellites.

Even if the technology works, scaling up to a full orbital data center would be extremely expensive and risky. Launch costs, space debris, and solar storms are just a few hurdles. Some researchers estimate it would take five to six years of operation for a space data center to match the emissions of a green Earth-based facility. So while the idea is exciting, it's far from replacing your local data center anytime soon. Google's experiment will provide valuable data either way—showing what's possible and what's not.