Overview

Google has unveiled Project Suncatcher, a long-term research push to explore running large-scale machine learning in space. The idea is to create an interconnected constellation of solar-powered satellites carrying Google Tensor Processing Unit (TPU) chips, tapping abundant sunlight to expand computing capacity beyond Earth.

Announcement and early research

Announced November 4, 2025 on Google’s The Keyword blog, the effort follows the company’s tradition of “moonshot” projects, such as autonomous vehicles and quantum computing. Google released an initial preprint that lays out early findings on satellite constellation architecture, attitude control, and communications, as well as results from radiation testing of TPU hardware.

Prototype mission with Planet

The next milestone is a learning mission with Planet, the Earth imaging company. The partners plan to launch two prototype satellites by early 2027 to test hardware in orbit and validate core assumptions about performance, resilience, and operations. Insights from this flight will inform whether space-based compute can one day complement terrestrial data centers by easing power constraints and scaling AI training and inference.

Why it matters

Google emphasizes that Suncatcher is exploratory research with significant engineering challenges ahead. Still, the company sees promise in off-planet solar energy and distributed satellite networks as a pathway to massively scalable, energy-rich computation in the future.

If the prototypes succeed, they could pave the way for subsequent missions and, eventually, a new layer of AI infrastructure in orbit.