Skip to main content
Silicon Starts Rhyming

Google's TPUs produce first AI response from orbit

Google's experimental satellite has reached a milestone few expected so soon: a data center class chip launched into orbit has started producing language. The result came from a simple request, and it was poetic. But the real test begins now, as engineers prepare to push the hardware into space's harsh conditions.
Google's TPUs produce first AI response from orbit
Google's TPUs produce first AI response from orbit

Google has received the first AI response from a data center class TPU in orbit, a “Hello World” milestone for its Project Suncatcher prototype satellite. The satellite launched on October 1 as the first step in a long-term research moonshot to explore powering AI compute in space. Over the following weeks, the team completed its commissioning checklist, which included powering up the host and TPU trays and running early machine learning tests.

During a ground-station pass, Google uplinked a prompt asking its onboard Gemma 4 open model to write a haiku about being in space. As the satellite passed over Svalbard, Norway, the downlinked poem confirmed the system worked:

Blue marble below
Data streams in silent grace
Starlight feeds my core. ✨

After the haiku, the team tested space-themed limericks and telemetry prompts that reported the technical status of the payload systems. Next, Google will run a series of tests to evaluate how its TPUs perform under thermal stress and radiation exposure in orbit. Google says that verifying its TPUs can reliably execute workloads in space paves the way for its next milestone: twin interconnected satellites in 2027.

Key Features at a Glance

  • Onboard model: Gemma 4 open model running on the satellite – shows the payload can run a real generative AI model in orbit, not just simple test code.
  • Data center class TPU: the first such chip confirmed producing AI output in orbit – suggests the hardware used in terrestrial data centers can be carried to space.
  • Commissioning checklist: host and TPU trays powered up and early machine learning tests completed – confirms the system is operational before deeper testing.
  • Telemetry prompts: payload status reported through test prompts – demonstrates the model can work with the satellite's own technical data.
  • Thermal and radiation tests: planned next – will show whether the TPUs hold up under the harsh conditions of orbit.
  • Twin satellites: interconnected pair planned for 2027 – the next major step toward the moonshot.

Availability

Project Suncatcher is a research moonshot, not a commercial product. The prototype satellite launched on October 1. The next planned milestone is the twin interconnected satellites in 2027. No pricing or public availability was announced.

Felipe Santos

“Artificial intelligence can process the world in milliseconds, but only the human heart can give meaning to every second lived” – Mr. Santos