Google says its Project Suncatcher prototype has returned its first AI response from orbit, a milestone in the company’s research into running machine-learning compute on satellites.

In an October 30 update, Google said a ground station sent a prompt to an onboard Gemma 4 model. The reply was sent back during a pass over Svalbard. Google described the test as an AI response generated on a data-center-class TPU in orbit.

On this page

From launch to first response

The prototype launched on October 1. Since then, Google says its team has powered up the host and TPU trays and completed early machine-learning tests as part of commissioning. The company did not give an exact date for the first response; its October 30 post said the milestone occurred that week.

The prompt asked the onboard model for a haiku about being in space. After that response, Google said it tested additional prompts, including requests for information on the payload’s technical status. The company reported the result in its own post; the announcement did not provide an independent benchmark or a quantitative reliability result.

The earlier launch report explains the prototype’s background and Google’s pre-launch testing.

What remains to be tested

The company says its next experiments will assess how TPUs perform under thermal stress and radiation exposure in orbit. Those are continuing tests, not evidence from the October update that long-term reliability has already been demonstrated.

Google also identifies a 2027 milestone involving two interconnected satellites. It presents that as a next step, rather than a system already operating. Project Suncatcher is a research effort exploring whether AI compute powered by sunlight could run in space; a single successful response does not establish that an orbital data centre is commercially viable or ready for production workloads.

Source

Google Research: First response from the Project Suncatcher prototype satellite, October 30, 2026.