Data center in UK becomes proving ground for new carbon capture technology

In a lab in New Jersey, Orbital churns out its purple powder and sticks it to plastic sheets that are layered like lasagna to create a carbon-capturing air filter.

In the carbon capture process, air is drawn in through fans and pulled through the sorbent, which absorbs the carbon in the air. When the material is heated, it releases the CO₂ into large tanks that get hauled to wherever the carbon is being stored.

Orbital packs the machinery into shipping containers that can be sent to the data centers, like the one in the UK.

The system is similar to other carbon capture processes, except that it utilizes the power of the data center, which is constantly blowing fans and generating heat that can be reused to operate Orbital’s carbon capture machines, lowering costs.

The idea of direct air capture at data centers has been kicked around for years. One reason it hasn’t taken off is that the air coming out of the facilities is heated, while traditional sorbents require cold air.

Instead of trying to find a way to cool the air before it interacts with the sorbents, Orbital asked its proprietary AI model to find a molecule that would have better absorption at higher temperatures. Then, the company had to test the material in its lab. Founded in 2022, the purple powder is the first material it has marketed.

Companies like Amazon go through great pains to claim their data centers are not contributing to climate change, and have become major adopters of renewable energy technology, helping to finance solar and wind farms to power their facilities.

Amazon says it is the largest purchaser of renewable energy in the world, which it does mainly to offset the carbon emissions produced by its data centers.

Google, another major cloud provider, is planning to pay $100 per ton to a different direct air capture company. (The price is partially subsidized by federal incentives.)

If companies are able to capture carbon using their data centers, they could save money in the long run compared to purchasing carbon credits.

Godwin said the larger cloud providers have not yet deployed carbon capture projects at their data centers because they are more risk-averse.

“There are teething problems with any new technology,” he said. For instance, if something goes wrong with the system, it could cause public relations headaches for high-profile companies. Companies also worry the carbon capture technology could have a performance impact on the data center itself.

Orbital hopes to prove itself with its first pilot program at the Civo data center, which is about an hour from London by train.

The building, which houses a mixture of traditional compute and high-performance GPUs for handling the biggest AI models, will feed hot air to the Orbital containers shipped from New Jersey.

There is no guarantee that Orbital’s carbon capture system will perform to a high enough standard to make a dent in climate change, but it’s a worthy goal. And if it’s successful, it could mean the AI boom doesn’t have to speed up harmful environmental effects.

Orbital’s idea also provides a working model for how AI-assisted materials science can be employed.

There are many failed inventions that could become successful ones with the right chemistry breakthrough. From batteries to solar panels, an explosion of new molecules could lead to step changes in the efficiency of existing products.

That means every new technology could become an enabler for even more advances.

The conventional wisdom is that carbon capture is not a very practical solution to climate change because of its high cost. Even at $200 per ton, it may be too expensive. Another problem is that storing all of the carbon is a challenge. It can be stored underground, but much of the world, including Europe, is not suitable for that kind of storage, as this Reuters article points out.