Project Suncatcher’s orbiting TPU delivers its first AI-generated Haiku: Unite.AI

Google said on October 30, 2026, that its prototype Project Suncatcher satellite returned its first AI response from orbit: a haiku generated by an onboard Gemma 4 open model running on a data center-class Tensor Processing Unit (TPU).

Travis Beals, senior director of Paradigms of Intelligence at Google, announced the milestone in a Google Research post, describing it as the mission’s “Hello World” milestone and what the company called the first-ever AI response from an orbiting data center-class TPU. During a flyover to the ground station, the team connected its first message to the Gemma 4 model aboard the satellite, asking it to generate a haiku about being in space. As the satellite passed over Svalbard, Norway, the downlinked poem confirmed the success: “Blue marble below / Data streams in silent grace / Starlight nourishes my core.”

After the initial result, the team tested space-themed limericks and telemetry instructions that reported the technical status of the payload systems.

Commissioning after launch

The satellite launched on October 1, 2026 aboard the Transporter-18 rideshare mission with SpaceX and was built in collaboration with Planet, according to Google’s launch announcement. The team confirmed contact with the spacecraft and said it was functioning as expected.

In the weeks following launch, the team completed a go-live checklist that included powering up the host and TPU trays and successfully running the first machine learning tests, Beals wrote. According to the launch announcement, the mission is designed to collect in-orbit data on how Google’s TPUs handle the physical stress of spaceflight and the radiation and thermal extremes of space. A peer-reviewed paper detailing the research behind the mission is now available in the journal Joule.

A Moon launch announced in 2025

Google announced Project Suncatcher on November 4, 2025, describing it as a long-term lunar research that explores whether an interconnected network of solar-powered satellites equipped with its TPU AI chips could one day scale machine learning to space. The company’s initial research, shared in a preprint document alongside the announcement, described its approach to designing, controlling, and communicating satellite constellations, along with initial learnings from radiation testing of Google’s TPUs. The plan then called for a learning mission in collaboration with Planet to launch two prototype satellites by early 2027 to test the hardware in orbit.

In a behind-the-scenes engineering post published on September 24, 2026, Google said that in low-Earth orbit, satellites can access near-constant sunlight, generating up to eight times more solar energy than on Earth. The company said it may eventually be possible to link multiple satellite constellations together, allowing them to handle larger AI workloads while in orbit.

Ground tests and 2027 milestone

Before launch, the team tested whether the hardware could survive the journey to orbit. A rocket ride to low Earth orbit takes about 10 minutes, with acceleration loads sustained up to 10 times the force of gravity, and individual components such as TPU chips can experience forces of 50 to 100 g, according to the engineering office. The team conducted vibration tests by intensely shaking the satellite on all three axes to mimic the frequencies of a rocket launch.

On the radiation side, the team tested the TPUs in a proton beam facility at UC Davis’ Crocker Nuclear Laboratory while running AI workloads, monitoring how errors such as a bitflip would affect those workloads. Google reported that initial results showed that its Trillium TPUs can survive a greater total ionizing dose of radiation than they would receive during a five-year space mission.

Cooling is another challenge described by the team. TPUs generate a large amount of heat in a small area, and in the vacuum of space there is no airflow, so the heat can only be spread via radiators. The team is looking at a combination of heat pipes and radiators to cool the chips, and has so far tested the technology in a thermal vacuum chamber that simulates both the thermal and vacuum environments in space.

Future satellite designs would each carry dozens of TPU chips as they orbit the Earth in clusters, communicating via lasers that must operate at very high bandwidth over extremely short distances — a challenge the post compares to hitting a coin-sized target miles away while both dots are moving. Google plans to test the interconnection work in 2027, when it will put two satellites into orbit.

Next, Beals wrote, the team will perform a series of tests to evaluate the performance of the TPUs under conditions of thermal stress and radiation exposure on orbit. Google said verifying that its TPUs can reliably run workloads in space paves the way for the project’s next milestone: interconnected twin satellites in 2027.