Google achieves first AI response from satellite in Project Suncatcher
The company successfully ran an AI model in orbit and received a haiku response back on Earth.
Google has successfully operated an artificial intelligence model on an orbiting satellite and received its first complete response on Earth, according to the Free Press Journal. This development is part of Project Suncatcher, an initiative to study the feasibility of running AI computing infrastructure in space.
Google CEO Sundar Pichai announced the milestone on X, stating that the team sent a command to the satellite, activated its Tensor Processing Unit (TPU), and received a haiku in return. The prototype satellite was launched on October 1 to research whether space can support scalable machine learning infrastructure.
During a ground-station pass over Svalbard, Norway, the team prompted the onboard Gemma 4 open model to write a haiku about space. Following this, the company performed further tests using space-themed limericks and requests for the satellite to report its technical status.
The next stage of the project will examine how the TPUs handle radiation exposure and thermal stress while in orbit. These tests aim to determine if the chips can consistently manage computing workloads in a space environment.
Google stated that successful results will lead to a 2027 milestone involving two interconnected satellites. The project is investigating if satellite constellations in low Earth orbit could eventually manage larger AI workloads by using the near-constant sunlight available in space.