We’ve received our first response from the Project Suncatcher prototype satellite.

On October 1, Google launched a prototype satellite for Project Suncatcher, initiating research into powering AI compute in space. The satellite carried a data center-class TPU and a Gemma 4 open model. Engineers completed commissioning tasks, including powering up the host and TPU trays, and conducted early machine learning tests. The team then prepared for the first operational milestone.

This week, the satellite achieved its "Hello World" moment by responding to an AI prompt while passing over Svalbard, Norway. Ground controllers uplinked a request for a haiku about being in space, which the onboard model generated and downlinked successfully. The response confirmed the TPU’s ability to execute AI workloads in orbit. Additional tests followed, including requests for space-themed limericks and telemetry prompts.

Following the initial success, engineers began evaluating the TPU’s performance under orbital conditions. The tests focus on thermal stress and radiation exposure to assess reliability. These assessments are critical for validating the hardware’s suitability for sustained space operations. The findings will guide future design iterations.

Google plans to advance the project with twin interconnected satellites scheduled for 2027. The successful operation of the prototype’s TPU in orbit represents a key achievement. The milestone demonstrates progress toward deploying AI compute resources in space, a long-term goal for the project.