Skip to content
Neural Daily

ResearchNews

Google's TPU satellite sends its first AI reply from orbit

🇭🇺 Olvasd el magyarul →

A Gemma 4 model running on Google's Project Suncatcher prototype satellite wrote its first haiku from space. Google calls it the first AI response from a data center class TPU in orbit.

By: Mikola Máté Patrik · · 2 min read

Google team members gathered around a laptop with the caption Testing AI in orbit
Image: Google · Press kit · Thumbnail of Google's official Project Suncatcher video · Source

What happened

Google launched the first Project Suncatcher prototype satellite on October 1 as part of a long-term research moonshot into powering AI compute in space. Over the following weeks the team worked through its commissioning checklist, powering up the host and TPU trays and running early machine learning tests (Google).

This week it reached its Hello World milestone. During a ground-station pass, the team uplinked a prompt to the onboard Gemma 4 open model asking for a haiku about being in space, and the poem came back as the satellite passed over Svalbard, Norway. Google calls it the first AI response from a data center class TPU in orbit. Space-themed limericks and telemetry prompts reporting the payload's status followed.

What comes next

The next tests will measure how the TPUs handle thermal stress and radiation in orbit. If the chips run workloads reliably, the next milestone is twin interconnected satellites in 2027. The Verge notes that the project could eventually put AI data centers in space (The Verge).

Why it matters

Power demand from AI data centers is becoming a political and infrastructure constraint in more and more places on Earth. Suncatcher is testing whether solar-powered compute in orbit can work. A haiku is a long way from a data center, but it is the first proof that a TPU and an open model actually run in orbit, and the team can now focus on reliability.

    Produced with the help of AI, reviewed by an editor. How we make our articles

    Reviewed and edited by the Neural Daily desk. About

    Related articles