Google’s Project Suncatcher Aims to Launch AI Data Centres in Space

In a move that blends science fiction with cutting-edge innovation, Google has announced Project Suncatcher, an ambitious research initiative to build AI data centres in space. The project envisions solar-powered satellites housing high-performance AI hardware like TPUs (Tensor Processing Units) orbiting the Earth. The goal is to overcome the energy and sustainability limits faced by ground-based data centres, particularly as AI demands continue to soar.

Why AI Data Centres Need Space

Modern AI data centres are energy-intensive, consuming vast amounts of electricity and water—especially for cooling servers. With sustainability pressures and infrastructure limitations mounting, placing AI hardware in orbit presents an attractive alternative,

• Uninterrupted access to solar energy in space
• No water required for cooling
• Potential for greater computing efficiency with lower emissions
• Scalability without exhausting terrestrial resources

According to Google, solar panels in space can be up to 8 times more efficient than those on Earth, offering near-continuous power and a sustainable solution for the data-heavy future.

Project Suncatcher: How It Works

Google’s approach involves deploying a constellation of satellites, each equipped with,

• TPUs (Trillium v6e chips)
• Solar panels for uninterrupted energy
• Free-space optical communication systems capable of tens of terabits per second

These orbiting nodes would form a distributed computing network, much like Starlink’s satellite-based internet service but optimized for AI data processing. The data centres would communicate with each other through high-speed optical links and operate in a tight orbital formation, possibly less than a kilometre apart.

Key Challenges and Technical Barriers

Despite its promise, Project Suncatcher faces formidable engineering hurdles,

High-Speed Satellite Communication

• Maintaining multi-terabit wireless communication between orbiting satellites remains technically complex.
• Early lab tests have shown 1.6 Tbps bidirectional speeds, but consistency in orbit is unproven.

Satellite Proximity and Station-Keeping

• Satellites need to maintain precise distances of a few hundred meters.
• Requires sophisticated station-keeping technology to avoid collisions or drift.

Radiation Resistance of AI Chips

• Google’s TPUs have undergone radiation testing under 67 MeV proton beams, with no critical failures detected up to 15 krad(Si).
• However, long-term space exposure still poses a risk.

Hardware Adaptation

• Most AI hardware is not designed for vacuum, temperature extremes, or cosmic radiation.
• Reusing Earth-based components in space will need custom engineering.

Cost and Launch Logistics

• Launching satellites is expensive, but Google expects costs to fall to $200/kg by 2030s.
• The initial prototype satellites are expected to launch by early 2027.

Strategic Significance

• Scalability: Space data centres could overcome real estate and resource limitations on Earth.
• Sustainability: Offers a low-emission path to meeting AI’s growing energy needs.
• Geopolitical Tech Race: Positions Google alongside other tech titans in defining space-based digital infrastructure.
• Disaster Resilience: Orbital networks could provide backup infrastructure in case of terrestrial network failures.