Google Completes First Orbital AI Test Under Project Suncatcher

Google has successfully executed an artificial intelligence model on an orbiting satellite and received a response on Earth. This marks a significant milestone for Project Suncatcher, which investigates the viability of space-based machine learning infrastructure.

Google has achieved a key technical milestone in its Project Suncatcher initiative by successfully running an artificial intelligence model on a satellite in orbit and receiving a complete response back on Earth. The development was announced by Google CEO Sundar Pichai in a post on X.

According to the company, a prototype satellite was launched on October 1 as part of a long-term research effort to examine whether space can support scalable machine learning infrastructure. Following its launch, the satellite completed initial commissioning checks, including powering up its host and Tensor Processing Unit (TPU) trays and running early machine learning tests.

During a ground-station pass over Svalbard, Norway, the team sent a command to the satellite, booted up its TPU, and prompted the onboard Gemma 4 open model to generate a haiku about space. Google subsequently tested the system with space-themed limericks and prompts requesting the technical status of the satellite's systems.

Project Suncatcher aims to explore whether satellites could eventually be linked into constellations to handle larger AI workloads. Google notes that satellites in low Earth orbit could access near-constant sunlight and potentially generate up to eight times more solar power than equivalent infrastructure on Earth.

The next phase of the project will involve testing how the TPUs perform under thermal stress and radiation exposure in orbit. These evaluations are intended to determine if the chips can reliably sustain computing workloads in a space environment. Successful testing is expected to pave the way for the next planned milestone, which involves two interconnected satellites in 2027.

"The successful execution of an AI model on an orbiting satellite marks an interesting step in exploring alternative computing infrastructure. As energy demands for artificial intelligence continue to rise globally, investigating space-based solutions powered by solar energy opens up new technical possibilities. However, the true commercial and operational viability will depend heavily on the upcoming radiation and thermal stress tests in orbit, as well as the long-term scalability of satellite constellations." — Dr. Shishir Gupta, Founder & CEO, StartupLanes