The first TPU in space is now online — and it’s writing poetry. Google’s Suncatcher prototype space data center has completed commissioning and is communicating with the ground after launching earlier this month on a SpaceX Falcon 9 alongside more than 100 other spacecraft.
Planet Labs (Credits: TechCrunch)
Google Suncatcher satellite: first TPU in space runs Gemma 4
The satellite carries four Google Tensor Processing Units and generates up to 1 kW of electricity. It’s intended to pave the way for future networks of close-flying satellites performing large AI inference workloads in orbit. The power-hungry TPUs mean engineers can only switch one on for about 20 minutes at a time — complicating the already tricky business of space-to-ground communications.
After a roughly three-week commissioning process, the ground team sent its first prompt this week, asking the pre-loaded model — Google DeepMind’s Gemma 4 — to write a haiku about being in space. The satellite responded:
“Blue marble below / Data streams in silent grace / Starlight feeds my core. ✨”
Google also tested the model with space-themed limericks and asked the spacecraft to report on orbital conditions around the satellite. A spokesperson told TechCrunch the satellite was operating nominally.
Radiation, bit flips and the path to orbital data centers
The ongoing test campaign will examine how the chips hold up under thermal stress and space radiation. Radiation can cause “bit flips” that silently corrupt memory values — ground testing suggested the impact would be minimal for inference workloads, but the team will now assess the real dynamics in orbit rather than in simulation.
If the TPU proves capable, Google plans to launch two more satellites in 2027 with higher power generation and inter-satellite optical links. Notably, Suncatcher’s vision stands apart from rivals like Starcloud and Cowboy Space in planning high-speed links between satellites for coherent distributed workloads — a moonshot aimed at making orbital compute economical in the 2030s.
Also Read: Nanbeige Releases Nanbeige4.3-33B, a 33B Open Agentic Coding Model With 512K Context | NVIDIA Releases Nemotron 3.5 Bolt E4B 4GB Open Small Language Model
Source: TechCrunch AI — read the full story there.
