Google’s Project Suncatcher prototype satellite has run an AI model on one of its data-centre chips in orbit and sent the answer back to Earth: a three-line haiku. Travis Beals, who leads the effort, called it in a Google blog post on 30 October “the first-ever AI response from a data center class TPU in orbit”.
It is an early engineering milestone for a research project, not a working space data centre. The satellite is a single prototype, launched on 1 October, and the hard tests on radiation and heat have only now begun.
In brief
- Google’s Project Suncatcher prototype ran its onboard Gemma 4 model on a TPU in orbit and downlinked a haiku over Svalbard, Norway.
- Commissioning is complete. Tests on how the chips cope with thermal stress and radiation come next.
- Google plans twin interconnected satellites in 2027. That is a plan, not a scheduled launch.
What Project Suncatcher has done so far
The satellite, built with Earth-imaging company Planet, went up on SpaceX’s Transporter-18 rideshare mission on 1 October, according to Google’s launch post. Google said on X that day that the prototype carries four TPUs, its own Tensor Processing Units, the chips it designs for AI work in its data centres.
Over the following weeks the team worked through a commissioning checklist, which Beals said included “powering up the host and TPU trays and successfully running early machine learning tests”. Then, during a ground-station pass, engineers uplinked a prompt asking the onboard Gemma 4 open model for a haiku about being in space. The reply came down as the satellite passed Svalbard:
Blue marble below / Data streams in silent grace / Starlight feeds my core.
Google chief executive Sundar Pichai described the sequence on X: “we’ve now successfully sent a command up to orbit, booted up the TPU, ran the prompt, and received our first complete response back on earth”. The team also tried space-themed limericks and prompts asking the payload to report its own technical status, the blog says.
TechCrunch, which spoke to Google, reported that the satellite generates up to 1kW of electricity and that the chip can only run for about 20 minutes at a time because of its power demands. A spokesperson told the outlet the satellite was operating nominally.
Radiation and heat tests come before any 2027 twins
The next phase is the one that matters. Google says it will run “a series of tests to evaluate how TPUs perform under thermal stress and radiation exposure in orbit”. Planet flagged one of the core problems when it promoted the mission on 7 October: “how to cool AI chips in the vacuum of space”, where there is no air to carry heat away.
If those tests go well, Beals wrote, they pave “the way for our next milestone of twin interconnected satellites in 2027”. That is a stated plan. Google has given no launch date, budget or design details for the pair in these posts.
Google’s wider Earth and space work has mostly come down to the ground so far, through tools such as its Earth AI geospatial reasoning agent, which analyses satellite data on Earth rather than in orbit. Suncatcher flips that: the computing itself goes up.
A haiku proves the chip runs, not that space compute pays
The milestone is real but narrow. Running a small open model and returning a few lines of text shows that a commercial AI chip survived launch, powered up and completed a full command-and-response loop through a ground station. That is the minimum any orbital computing scheme needs, and it is more than most proposals for space data centres have shown in hardware.
What it does not yet show is whether the economics can ever work. The questions that decide that are the ones Google has listed as next: whether radiation corrupts results over months, whether the chips can shed heat without air, and whether enough solar power can be collected to run them for longer than short bursts. TechCrunch’s report of roughly 20-minute operating windows is a reminder of how far a single prototype is from continuous data-centre work. Bandwidth is another limit: the haiku came down during a ground-station pass, not over a permanent high-speed link.
The project also sits within a broader race to put computing where the energy is. Elon Musk’s SpaceX has set compute targets measured in gigawatts, while the UK Space Agency’s AI roadmap names edge computing on board spacecraft as one of its priority technologies. Projects such as Ai2’s open satellite embeddings show the more immediate use of AI and satellites: processing imagery on the ground. Google’s test is a credible first step in hardware, but the twin-satellite flight in 2027 will say far more than a poem about a blue marble.
Sources
- Google: Google receives response from Project Suncatcher prototype satellite in space (Travis Beals, 30 Oct 2026)
- Sundar Pichai on X, 30 Oct 2026
- Google: Google’s Project Suncatcher prototype satellite is in orbit (1 Oct 2026)
- Google on X, launch post, 1 Oct 2026
- Planet on X, 7 Oct 2026
- TechCrunch: Google’s first data center satellite is online and writing poetry (30 Oct 2026)
Related stories
- Google Expands Its Experimental Earth AI Geospatial Reasoning Agent
- Ai2 Releases Open Satellite Embeddings for Every 10m Pixel on Earth
Follow @tsnmediahub on X for breaking AI news as it happens.
Follow on X
