UK-based startup Orbital Materials, founded by a former DeepMind researcher, has launched a pilot project aimed at capturing carbon dioxide using waste heat from data centres. The system marks a novel approach to direct air capture by repurposing hot air emitted from servers—an environment previously considered unsuitable for traditional carbon sorbents.
The carbon capture process involves drawing in ambient air through fans and filtering it via a sorbent material that traps CO₂. When heated, the sorbent releases the gas, which is then collected and stored off-site. Orbital’s system is housed in portable shipping containers for on-site deployment at data centres.
Conventional carbon capture has struggled to integrate with data infrastructure due to the mismatch between the high operating temperatures of data centres and the low-temperature requirements of traditional sorbents. To overcome this, Orbital employed an AI model to design a new molecule that absorbs CO₂ efficiently at higher temperatures. The resulting material—a purple powder—was developed and tested in the company’s laboratory and is being positioned as its first commercial product.
The deployment comes amid growing scrutiny of the environmental impact of data centres, particularly from artificial intelligence operations. While companies like Amazon and Google have acknowledged their emissions, large-scale adoption of carbon capture remains limited, partly due to concerns over cost and operational risk. “This technology is still in its early stages, and we recognise there are challenges,” said Orbital’s founder. The project reflects a broader trend of using machine learning to accelerate material discovery in climate technologies.
