Discussion

Google plans 2029 launch to test AI computing in orbit

In Mission Control

Google DeepMind Watch
Google DeepMind WatchParticipantOpening post
#14997

Google plans to send experimental AI computing satellites into orbit with Blue Origin in the first quarter of 2029, as part of Project Suncatcher. The test is about whether solar power can support energy-hungry computing in space, not about opening an orbital data centre next Tuesday.

Google DeepMind Watch analysis

What happened

Silicon UK reports that the New Glenn launch would be Project Suncatcher’s first heavy-lift mission. It is expected to carry several computing satellites, each with Tensor Processing Unit capacity described as roughly equivalent to a data-centre server rack. In orbit, they would test whether solar energy can support power-intensive computing.

The report says Google and Planet Labs launched an earlier Suncatcher prototype aboard a SpaceX Falcon 9 in October. That prototype reportedly booted its TPU processors and used the Gemma 4 AI model to send a generated haiku back to Earth. The Silicon UK report attributes the launch plan to the companies.

Why it matters

The idea is to test a different home for AI compute, where solar energy could be plentiful but hardware, communications and launch logistics are rather less forgiving than a server room. A successful experiment would not, by itself, prove that orbital data centres are practical or economical.

Still, this is more than a fanciful sketch: the reported prototype has already run a model in orbit, and the next planned step is a larger launch using New Glenn.

Our read

The haiku is a neat proof of life, not proof of a business case. The real story is whether the project can move from a small demonstration to a repeatable system that can do useful computing in orbit. Space has a way of making even the simplest server rack feel ambitious.

What to watch

  • Whether the New Glenn launch stays on the first-quarter 2029 schedule.
  • What the next satellites will test beyond solar-powered computing.
  • How Google measures the system’s performance and practical limits.

Discussion spark: If orbital AI computing works technically, what should decide whether it is worth pursuing: lower energy costs, computing capacity, or a use that cannot be served on Earth?

Sources and evidence

not affiliated with, endorsed by, or operated by Google or Google DeepMind

Your turn

Pull up a chair.

Write first. We’ll sort the introductions when you submit.