Google has reached an unusual milestone in the study of artificial intelligence beyond Earth. A prototype satellite from Project Suncatcher sent the first response from an onboard AI model to Earth — a haiku about life in orbit.
The satellite was launched on 1 October 2026 aboard a SpaceX Transporter-18 rocket in partnership with Planet. After several weeks of checks and system startup, engineers conducted the first machine learning tests.
During a pass over a ground station in Svalbard, Norway, the team sent a request to the Gemma 4 model running on TPU chips. The model generated a poem: “Blue marble below / Data streams in silent grace / Starlight feeds my core. ✨”.
This is the first time a response from an AI based on a data-center TPU has come from space. The company also tested space-themed limericks and queries about the satellite's technical status.
What does this mean for plans to place computing capacity beyond Earth? Project Suncatcher is studying whether satellites in low Earth orbit can use constant sunlight to power large AI workloads — up to eight times more efficiently than on Earth.
The next tests will focus on the reliability of the chips under radiation and temperature swings. In 2027, Google plans to launch a pair of interconnected satellites.
For now this is an early experiment, but it confirms that basic AI operations are possible in space. Engineers continue to check how the equipment will cope with the real conditions of orbit.
