Source: TH
Subject: Science and Technology
Context: Google Research has unveiled Project Suncatcher, exploring AI datacentres in low-Earth orbit powered entirely by solar energy to tackle AI’s surging electricity demand.
About Project Suncatcher:
What it is?
- Project Suncatcher is a concept and research programme to place AI datacentres in low-Earth orbit (LEO), operating continuously on solar power to run energy-intensive AI workloads.
Launched by: Google (Google Research).
Aim:
- Cut AI’s energy footprint by using uninterrupted solar power.
- Decouple AI compute growth from terrestrial grids, land use, and water-intensive cooling.
How it works?
- Deploys densely clustered satellites (not a sparse global swarm) flying in sun-synchronous orbits to ensure constant sunlight.
- AI workloads are distributed across satellites using ultra-high-bandwidth inter-satellite links; Earth downlinks handle only inputs/outputs.
- Uses radiation-tolerant TPUs and specialised thermal designs to operate in vacuum.
Key features:
- Always-on solar energy – no atmosphere, no night cycles in chosen orbits.
- Petabit-scale inter-satellite networking to support distributed AI training/inference.
- Radiation-hard compute – tests show TPUs tolerate doses beyond multi-year mission needs.
- Minimal Earth bandwidth dependency compared to internal cluster bandwidth.
- Scalable constellation architecture, with satellites replaced as units age out.
Significance:
- Offers a new path to power AI sustainably as model sizes and training runs explode.
- Reduces pressure on grids, water, and land near terrestrial datacentres.
