Google’s Space AI just wrote a Haiku from orbit; here’s how it did it

Google has reached an unusual milestone in its plan to explore artificial intelligence in space. A prototype satellite launched earlier this month has sent back its first AI-generated response, a haiku about life in orbit. The achievement marks the first successful machine learning response from a data center-class Tensor Processing Unit (TPU) operating in space. It is an early test, but one that brings Google's Project Suncatcher a step closer to its long-term goal of exploring AI data centers beyond Earth.

How did Google’s satellite write a haiku in space

Google launched its Project Suncatcher prototype on October 1 2026 aboard SpaceX's Transporter-18 rideshare mission. The satellite was built in partnership with Planet, and the team confirmed contact with it shortly after launch.


Over the following weeks, engineers completed the initial checks, powered up the onboard systems and ran early machine learning tests. The biggest milestone came during a ground-station pass, when the team sent a prompt to Google's Gemma 4 open model running on the satellite.

The instruction was to write a haiku about being in space.

As the satellite passed over Svalbard, Norway, it sent the response back to Earth:

“Blue marble below

Data streams in silent grace

Starlight feeds my core. ✨”

Google described this as the first AI response generated by a data center-class TPU in orbit. The team also tested space-themed limericks and prompts asking the satellite to report the technical status of its onboard systems.

What technology is powering Project Suncatcher

The prototype carries four Tensor Processing Units, or TPUs. Google designed these chips for machine learning and already uses them in data centers on Earth. The satellite runs an onboard version of the Gemma AI model. Because of heat-management limits, the chips are expected to run the model for about 15 minutes at a time while handling simple requests.

The next stage will focus on how the processors cope with the conditions of space. Google plans to test their performance under thermal stress and radiation exposure.

Radiation is a particular concern because it can cause what engineers call “bit flips”. These change stored values in computer memory and can corrupt data without necessarily setting off standard monitoring systems.

Why does Google want AI data centers in space?

Project Suncatcher is a long-term research effort to explore whether space could eventually support large-scale machine learning infrastructure.

Satellites in low Earth orbit can receive near constant sunlight and may generate up to 8 times more solar power than comparable systems on Earth. Google is exploring whether connected satellite groups could use that energy to handle larger AI workloads. The company plans to test twin interconnected satellites in 2027. The first successful response is a starting point, engineers continue checking whether the hardware can work reliably in orbit. The information is taken from Google's blog post and NPR.