Google's Suncatcher satellite returns its first AI response from an in-orbit data center TPU, a haiku
Google's Project Suncatcher prototype satellite launched on October 1, and over the following weeks the team worked through an in-orbit test and commissioning checklist that included powering up the host and TPU trays and running early machine learning tests.
Google's Project Suncatcher prototype satellite launched on October 1, and over the following weeks the team worked through an in-orbit test and commissioning checklist that included powering up the host and TPU trays and running early machine learning tests. On October 30 Google announced a Hello World milestone: a data-center-class TPU running in orbit returned its first AI response. This week ground engineers uplinked a prompt asking the preloaded Google DeepMind model Gemma 4 to write a haiku about being in space, and the model answered with the lines Blue marble below, Data streams in silent grace, Starlight feeds my core. Alongside the haiku, engineers also tested limerick generation and telemetry tasks such as returning orbital environment data. The satellite carries four Google TPUs capable of producing up to one kilowatt of power, intended as groundwork for tightly clustered satellite networks that run large inference workloads in orbit. Limited by heat rejection and energy budget, each TPU can only run for about 20 minutes at a time, and the test focused on verifying chip stability and error correction in microgravity, sharp temperature swings and space radiation.