Google has successfully powered up its first experimental orbital data center satellite and received an AI response from space. The prototype, part of Project Suncatcher, is now online and communicating with ground stations after launching earlier in October 2026.
The satellite, built in partnership with Planet Labs, carries four of Google’s Tensor Processing Units (TPUs). It launched on October 1 aboard a SpaceX Falcon 9 as part of the Transporter-18 rideshare mission. After about three weeks of commissioning and system checks, engineers sent the first prompt to the onboard AI model.
The First Message from Orbit
During a ground-station pass over Svalbard, Norway, Google uplinked a request to the satellite’s pre-loaded Gemma 4 model: write a haiku about being in space. The response came back cleanly:
Blue marble below
Data streams in silent grace
Starlight feeds my core.
✨
Google also tested space-themed limericks and telemetry prompts that reported the technical status of the payload systems. The company described the successful exchange as its “Hello World” moment for running a data-center-class TPU in orbit.
What the Satellite Can (and Cannot) Do
The prototype generates up to about 1 kilowatt of power from its solar panels. Because of that limited power and the challenges of cooling electronics in the vacuum of space, only one TPU can typically run at a time, and only for roughly 15–20 minutes before it needs to cool down. The satellite is designed as a test platform rather than a full operational data center.
The main goals of this first mission are to verify that Google’s TPUs can survive launch stresses, operate under orbital radiation, and handle thermal cycles. Future tests will look more closely at performance under sustained thermal stress and radiation exposure.
Why Google Is Doing This
Project Suncatcher is a long-term research effort exploring whether large-scale AI computing can eventually move into orbit. The idea is that satellites could take advantage of near-constant sunlight for power and avoid some of the land, energy, and cooling constraints that affect Earth-based data centers.
Google has already outlined next steps. If this prototype continues to perform well, the company plans to launch two more advanced satellites in 2027. Those follow-on craft are expected to carry more power and demonstrate high-speed optical (laser) links between satellites, which would be essential for any future distributed computing network in orbit.
The Bigger Picture
Putting AI hardware in space is still very early and experimental. Cooling remains one of the hardest technical challenges, and the economics of large orbital data centers are far from proven. Still, Google’s successful first response marks a concrete milestone: the company’s custom AI chips have now run workloads in orbit and returned results to Earth.
For now, the satellite is continuing its test program. The haiku may have been a light first message, but the underlying achievement is more serious — proving that data-center-class AI hardware can function beyond the atmosphere.