Google's Project Suncatcher satellite sent its first AI answer from orbit: what it is and what's next
Quick answer
On Oct 30, Google said its Project Suncatcher prototype satellite had produced “the first-ever AI response from a data center class TPU in orbit.” Engineers sent a prompt to the onboard Gemma 4 open model during a ground-station pass, asking for a haiku about being in space, and got one back as the satellite passed Svalbard, Norway (Google). Google calls it a “Hello World” milestone. It’s a test of whether AI chips work in space, not a working data center.

What happened
- Oct 1, 2026: the prototype launched (Google), on a SpaceX Falcon 9 rideshare with more than 100 other spacecraft (Dataconomy).
- Commissioning: Google powered up the host and TPU trays and ran early machine-learning tests.
- Oct 30, 2026: Google announced the first response (Google).
- The answer: “Blue marble below / Data streams in silent grace / Starlight feeds my core.” Google then tested limericks and prompts that report the payload’s technical status (Google).
What’s next
- Tests of how the TPUs hold up under heat stress and radiation in orbit (Google). Radiation can flip memory bits and quietly corrupt data (Dataconomy).
- If that goes well: twin interconnected satellites in 2027 (Google).
Why it matters
- AI data centers on Earth are fighting for power. The Suncatcher idea is to put the compute where sunlight is constant and use the sun to power it.
- It’s a long-range moonshot. This test shows one chip can run a model up there. It doesn’t show that space compute is cheap or ready.
FAQ
- What model ran in space? Gemma 4, Google’s open model, preloaded on the satellite.
- Is Google running AI data centers in space now? No. This is one prototype satellite in a test campaign.
- When is the next launch? Google is aiming for two interconnected satellites in 2027.
