AI and the Move Off Earth: Why Companies Are Testing Data Centers in Space

AI model training and cloud services have driven strong growth in data centre work. The IEA has found that global data centre electricity use was between 240 and 340 TWh in 2022, and ML workloads have grown rapidly while energy efficiency has improved. Since demand for large models has risen, some firms have sought new ways to supply power and cooling.

Space-based data centres are being developed by startups and industry groups. Starcloud plans a demonstrator mission in 2025 and has said it raised about $21 million in seed funding. The European ASCEND feasibility study was published in June 2024 and examined whether orbital centres could lower carbon footprints. In orbit, solar arrays can produce more energy per area and radiators can reject heat without water; however, launch costs, battery life during eclipses, and orbital traffic are real constraints.

Therefore, space is one option among many. Terrestrial ideas have been tested too: Microsoft's Project Natick placed a sealed undersea module in 2018 and learned lessons about reliability. Furthermore, smaller orbital tests will show whether the high cost of access to space is outweighed by long-term energy savings.