Google Activates First AI Data Centre Satellite
Google has activated its first orbital data centre prototype, marking a milestone in its effort to bring AI computing infrastructure into space.
The satellite, developed under Google’s Project Suncatcher, has powered up its onboard Tensor Processing Units (TPUs), run an AI model and transmitted its first complete response back to Earth. The achievement demonstrates that AI computing hardware can operate in orbit and communicate with ground stations.
Project Suncatcher Completes Its First AI Test
Google launched the prototype satellite on October 1, 2026, aboard a SpaceX Falcon 9 rocket. Planet Labs built the spacecraft, which carries four Google TPUs and can generate up to 1 kilowatt of electricity.
After approximately three weeks of commissioning, Google’s engineering team successfully activated the computing hardware and began running machine learning tests.
For its first demonstration, engineers sent a prompt to Google’s Gemma 4 open AI model, asking it to write a haiku about space. The satellite generated the poem and transmitted it back to Earth.
Google also tested space-themed limericks and prompts asking the satellite to report its technical operating conditions. The company confirmed that the spacecraft was functioning normally.
Technical Challenges and Next Steps
Despite the successful demonstration, Google still faces several engineering challenges before it can establish operational data centres in orbit.
The prototype carries four TPUs, but it can operate only one at a time. Power limitations restrict each operating period to approximately 20 minutes. Engineers must also address heat management, radiation exposure and reliable communication between satellites.
Radiation presents a particular concern because it can cause memory errors known as bit flips. These errors could affect computing accuracy and data reliability.
Google will therefore continue testing the processors under real orbital conditions. The next milestone is to launch two interconnected satellites in 2027. These spacecraft will test higher power generation and optical links between satellites.
Google’s Vision for Space-Based AI Infrastructure
Project Suncatcher explores building solar-powered computing networks in orbit. Such networks could eventually support large AI workloads without relying entirely on conventional data centres.
Google also believes advances in satellite technology and launch economics could make orbital computing more practical in the 2030s.
However, significant obstacles remain. Launch costs, thermal management, radiation protection and satellite manufacturing must become economically viable. Consequently, this demonstration is an early research milestone rather than a commercially operational space data centre.
Google’s latest achievement offers an early indication that AI infrastructure could extend beyond Earth. Its next challenge is to prove that orbital computing can operate reliably, communicate efficiently, and scale economically.
