Google is set to launch a small AI data center into space
Google is launching AI compute into orbit. MVP satellite carries four TPUs, runs for 15 minutes per test cycle. Full constellation comes 2027.
Why it matters
Project Suncatcher represents a concrete shift in how frontier labs approach compute scarcity and power constraints — moving AI inference to space to bypass terrestrial data-center gridlock and energy limits. For enterprises and practitioners, this signals that orbital compute is transitioning from concept to testbed; the operational model (15-minute duty cycles, laser inter-satellite links) and timeline (years to productization) matter for long-term infrastructure planning.
The key facts
16 to knowOctober 1, 2026: Google's Project Suncatcher MVP satellite launch date
MVP satellite: refrigerator-sized, built by Planet Labs
Four Google TPU accelerators onboard
Solar panels generate ~1 kilowatt
TPUs can run 15 minutes at a time during testing due to power and cooling limits
High-speed laser links planned for satellite-to-satellite communication
Additional launches planned for 2027
Google expects several years before productization
Workload: Gemini inference in space
Launch date: October 1, 2026
Satellite MVP built by Planet Labs, refrigerator-sized
Payload: 4 Google TPU accelerators running Gemini
Solar panels generate ~1 kilowatt; TPUs can operate 15 minutes at a time during testing
Target architecture: network of satellites with high-speed laser links
Timeline to product: several years
Story key: google-project-suncatcher-tpu-satellite-launch
The story so far
Earlier coverage of this storyline
Go to the source
Computerworldcomputerworld.com
Publisher excerpt: On October 1, Google’s first test satellite for Project Suncatcher is set to be launched. This is the next step in the company’s plan to build AI data centers in Earth’s orbit, according to The New York Times. The idea is, among other things, to harness solar energy in space instead of building…