First-of-Its-Kind AI-powered Tech Captures Carbon Using Heat From Data Centers

Artificial intelligence’s rapid expansion has raised significant environmental concerns due to its substantial energy consumption and associated carbon emissions.

Training large AI models demands considerable computational power, leading to increased electricity usage and greenhouse gas emissions.

But a new startup thinks data centers can be part of the solution.

Orbital Materials, founded by a former DeepMind researcher, is launching a first-of-its-kind pilot program to capture CO₂ using waste heat from servers.

For this, the startup plans to use an AI-designed material that can work more efficiently in data center conditions.

Orbital Materials would be testing its pilot at a Civo-operated data center in the UK.

>> In Other News: US Against Plan for Levy on Carbon Emissions From Ships, Leak Suggests

Capturing heat, cutting carbon

Direct air capture — which filters air through materials called sorbents — is a costly technique where expenses can go as high as $1,000 per ton of captured carbon.

Jonathan Godwin, Orbital co-founder and CEO, said early tests suggest the cost at the UK facility could drop to around $200 per ton.

"That’s lower than what we expected," Godwin told Semafor.

The key, he said, is a new molecule—a purple powder that operates at different temperatures than conventional sorbents, making it better suited to the heat-intensive environment of data centers.

"The integration of our ability to use AI and an extraordinary chemistry team allows us to innovate on the sorbents way more than anybody else," he said.

During the carbon capture process, fans draw in air and pass it through a sorbent material that traps carbon dioxide.

Once heated, the sorbent releases the CO₂, which is collected in large tanks and transported to storage sites.

Portable, scalable deployment

Orbital houses this system in portable shipping containers that can be deployed directly to data centers.

The idea to capture direct air has been floated around for years. However, it never materialized as data centers emit hot air, while traditional sorbents require cold air.

To tackle this challenge, Orbital turned to its proprietary AI model to design a molecule capable of absorbing carbon more efficiently at higher temperatures. The resulting material—a purple powder—was then tested in the company’s lab. The startup is now marketing this as its first commercial product.

Several tech giants, including Amazon and Google, have publicly acknowledged the carbon footprint of their data centers but have not fully addressed the impact of their AI operations.

According to Godwin, major cloud providers have yet to adopt carbon capture technologies at scale, largely because they tend to be more risk-averse.

"There are teething problems with any new technology," he said.

Orbital’s carbon capture system is still in its early stages, and it remains to be seen whether it can operate efficiently and at scale.

The company’s use of AI to design new materials reflects a broader trend in applying machine learning to accelerate scientific discovery.

While challenges such as cost and carbon storage remain unresolved, Orbital’s pilot could serve as a test case for how data centers might play a role in carbon removal.

Its success or failure may offer valuable insights for future climate-related technologies.