In November 2025, Google LLC announced ‘ProjectSuncatcher,’ a research moonshot initiative to develop aspace-based, solar-powered Artificial Intelligence (AI) infrastructure systemusingTensor Processing Units (TPUs)aboard satellites to explore the scalability and feasibility of machine learning (ML) compute systems in low-Earth orbit (LEO).
About ProjectSuncatcher:
Project Overview:It aims to build asolar-powered, space-based AI infrastructure using Trillium TPUs proven to withstand up to 15 krad(Si) radiation, operating in a dawn–dusk orbit with advanced thermal radiators to manage heat in space.
Launch & Partnership:Two prototype satellites, developed with Planet Labs, will launch in early 2027 to test AI performance and optical communication systems.
Solar-powered satellites in low Earth orbit (LEO):The idea is to place satellites in a dawn–dusk sun-synchronous orbit where solar panels can generate up to 8 × the power compared to ground installations, offering near continuous energy exposure.
Communication Test:Bench-scale validation achieved 800 Gigabits per second (Gbps) each way (1.6 Terabits per second, (Tbps) total) using Dense Wavelength Division Multiplexing (DWDM)-based free-space optical links, crucial for high-speed inter-satellite communication.
Constellation & Hardware:An 81-satellite constellation (within 1 kilometer, km) will form a distributed AI network, connected by Free-Space Optical (FSO) links achieving up to tens of Tbps using DWDM technology.
Economic Outlook:Falling launch costs below USD 200/kilogram (kg) by the mid-2030s could make space-based data centers cost-competitive with Earth-based ones, supporting scalable and sustainable AI growth.
IAS-2026 - OPTIONAL / GEOGRAPHY / PUBLIC ADMINISTRATION / SOCIOLOGY / ANTHROPOLOGY / ORIENTATION ON 03 & 04-10-2025 
