Project Suncatcher: Revolutionizing AI Computing with Solar Power in Space

Project Suncatcher is Google's latest moonshot project, which looks to a future where machine learning computing can extend beyond Earth — literally into space. By combining power from solar-powered satellites, Google TPU processors, and high-speed free-space optical links, the project aims to explore ways to build AI infrastructure in space that could revolutionize both computing efficiency and energy sustainability in the future.

The vision behind Project Suncatcher 

Artificial intelligence (AI) continues to be a powerful force driving scientific discoveries and solving humanity's biggest problems, but scaling up AI in a sustainable way on Earth is becoming increasingly complex. 

Project Suncatcher thus proposes an ambitious approach — launching a constellation of small, sun-powered satellites into orbit that can continuously process machine learning models in space.

In orbit, satellites can generate up to eight times more power from sunlight than on Earth and operate virtually 24/7 without the need for large batteries, making space an ideal environment for large-scale AI processing while reducing the use of resources on Earth. 

System Architecture: Satellites Built for AI Processing 

Project Suncatcher envisions a network of solar-powered satellites in low-Earth orbit (LEO), each equipped with a Google TPU (Tensor Processing Unit) and connected by space-based optical links.

The system would function as a “space data center,” distributing AI processing across hundreds of satellites in a network. 

This modular structure makes the system scalable, redundant, and sustainable — the cornerstones of future AI systems. 

Solving key engineering problems 

Building AI processing systems in space faces several challenges, including: 

1. Data center-level satellite connectivity 

Machine learning processing requires high-speed connectivity between compute units to achieve performance comparable to Earth-based data centers. Each satellite must support data transmissions in the order of terabits per second (Tbps).

By using Dense Wavelength Division Multiplexing (DWDM) and Spatial Multiplexing techniques, the Suncatcher satellites can maintain a powerful connection even from just a few kilometers apart. Laboratory test results have achieved combined speeds of over 1.6 Tbps, a positive sign for this approach.

2. Orbital control and movement mechanism (Orbital Dynamics) 

Moving dozens of satellites within a few hundred meters of each other requires precise physics calculations, so the research team used the Hill-Clohessy-Wiltshire equations in conjunction with JAX-based differentiable models to simulate a stable constellation of 81 satellites at an altitude of 650 km.

3. Radiation resistance of TPU 

Space hardware must withstand ionizing radiation that can cause system failures. Google tested TPU v6e (Trillium) under a 67MeV proton beam and found no serious damage even at doses up to 15 krad(Si), which is three times the radiation intensity of a real mission.

Economic value 

The cost of launching satellites, once a major barrier, is falling rapidly. Google's analysis predicts that by the mid-2030s, launch prices could fall below $200 per kilogram, making space-based AI processing as economically viable as Earth-based data centers in terms of annual energy consumption.

Future Directions and Cooperation 

The next step for Project Suncatcher is a joint test mission with Planet, which aims to launch two prototype satellites by 2027, with the aim of:

• Testing the performance of TPU in space 
• Experimenting with sharing machine learning data between satellites via optical links 
• Analyze the effects of heat and radiation in real-world conditions. 

Longer term, Google sees a future of gigawatt-scale satellite constellations designed to integrate energy storage, processing, and cooling in harmony, much like how system-on-chip integration revolutionized smartphones. 

Why Project Suncatcher is important 

Summary 

Project Suncatcher is Google's bold move to redefine the frontiers of AI infrastructure. Harnessing the power of endless solar energy and the vastness of space, it envisions a future of AI that is truly sustainable, scalable, and global.

Like past moonshots like Quantum Computing and Autonomous Vehicles, this project is about pushing the limits of technology and, if successful, could usher in a new era of “out-of-this-world” AI systems powered by the sun and human innovation.