Google has received its first artificial intelligence response from space, a small poetic milestone in its long-term experiment to see whether data centres could one day run in orbit.
The company said on 30 October that its Project Suncatcher prototype satellite had returned its first answer from a data-centre-class Tensor Processing Unit (TPU) in orbit, which it described as a “Hello World” moment for the project. During a ground-station pass, engineers sent a prompt to the satellite’s onboard Gemma 4 open model, asking it to write a haiku about being in space. As the spacecraft passed over Svalbard in Norway, it sent back: “Blue marble below / Data streams in silent grace / Starlight feeds my core.”
The prototype was launched on 1 October aboard a SpaceX Falcon 9 rideshare mission, Transporter-18, alongside more than 100 other spacecraft, and was built in partnership with satellite company Planet. Google spent the following three weeks working through a commissioning checklist, powering up the host and TPU trays and running early machine-learning tests, before the first full prompt-and-response test succeeded.
After the haiku, engineers also tested space-themed limericks and prompts asking the satellite to report on the technical status of its own payload systems. A Google spokesperson told TechCrunch the satellite was operating nominally.
The prototype carries four of Google’s TPUs, the custom chips that power its data centres on Earth, but it generates only about one kilowatt of electricity. That tight power budget means only one chip can run at a time, for roughly 20 minutes, which makes timing each contact with a ground station a careful exercise.
The idea behind Suncatcher is that satellites in the right orbit sit in near-constant sunlight, so their solar panels could run AI computing on clean power around the clock. Google’s longer-term vision is not a single machine working alone, but networks of close-flying satellites sharing large inference workloads over high-speed optical links, rather than sending every job back to Earth.
That vision is still a long way off. This test proves only that a data-centre-class chip and an AI model can run a small workload in orbit and return an answer. The next phase will examine how the TPUs cope with thermal stress and radiation in space. Radiation can flip bits in computer memory and corrupt data without triggering ordinary monitoring tools, and ground tests in particle accelerators suggested the effect would be small for inference — running an already trained model — but larger for training one. Those findings now have to be verified in real orbital conditions.
If the chips prove reliable, Google plans to launch two more satellites in 2027 with higher power generation, designed to demonstrate the inter-satellite optical links that would let spacecraft genuinely share a workload. The company is explicit that this is a research moonshot aimed at infrastructure that might only become economical in the 2030s, not a service customers can use today. Rival visions exist, including from Starcloud and SpaceX, but Google’s plan is distinctive in aiming for satellites close enough together to process work jointly.
For now, the haiku is the whole achievement — and, fittingly, it is about the view from up there.
Sources: Google, “Google receives response from Project Suncatcher prototype satellite in space”, https://blog.google/innovation-and-ai/models-and-research/google-research/project-suncatcher-satellite-response/?utm_source=about.google&utm_medium=referral&utm_campaign=homepage ; TechCrunch, “Google’s first data center satellite is online and writing poetry”, https://techcrunch.com/2026/10/30/googles-first-data-center-satellite-is-online-and-writing-poetry/
