AI IN SPACE: WHY ELON MUSK, GOOGLE AND AMAZON ARE LOOKING BEYOND EARTH FOR DATA CENTRES

Major technology companies are exploring the idea of deploying artificial-intelligence data centres in outer space.

- 

Google's “Project Suncatcher” remains at the conceptual stage, but research released by the company highlights how satellites equipped with solar-powered hardware and high-speed optical links could eventually host machine-learning compute loads. Other organisations cited include Nvidia Corporation-backed startups developing satellites with H100 GPUs, and SpaceX—led by Elon Musk—which has publicly indicated plans to use its Starlink V3 system as a platform for orbiting data-centre infrastructure.

The impetus for the shift arises as AI compute demands grow rapidly. On Earth, training large models and serving inference workloads consume increasing quantities of energy, require sophisticated cooling systems and use substantial land and water resources. Research from Google and others suggests that orbiting platforms could eliminate many of those constraints—by providing uninterrupted solar power, avoiding terrestrial grid limitations and reducing local resource consumption. If launch and maintenance costs fall sufficiently, the economics of space-based AI compute could eventually approach parity with Earth-based data centres.

WHAT DRIVES THE MOVE TO ORBITAL COMPUTING

The report explains that current data-centre growth is hitting physical and environmental limits. High-end GPUs and tensor-processing units now generate large amounts of heat, which must be managed through advanced cooling technologies that often require large amounts of electricity and water. At the same time, the land footprint of data centres is increasing, and in some regions availability of suitable power and space is becoming a bottleneck.