Can AI help create less carbon-intensive concrete?

Yes, according to an experiment from researchers at Meta, IBM, and several universities.
article cover

Stockseller_ukr/Getty Images

· 3 min read

Stay up to date on emerging tech

Drones, automation, AI, and more. The technologies that will shape the future of business, all in one newsletter.

Concrete production and carbon emissions are—unfortunately for the environment—a match made in heaven.

Each year, more than 10 billion tons of concrete are produced worldwide, and the main environmental offender is concrete’s key ingredient: cement. A study from researchers at IBM, Meta, and Ozinga, as well as at USC, University of Illinois Urbana-Champaign, and University of Chicago, recently suggested AI may be able to help reduce emissions—and then they tested it out in real-world data-center construction.

  • The researchers found that the new concrete formulations could roughly cut the global warming potential in half, when compared to the average of other formulations with similar 28-day strength.

AI-assisted cement

Cement is a popular binding and fortifying agent with a high production cost (and we’re not talking about $$): For every ton of cement produced, at least one ton of CO2 is released into the atmosphere—adding up to at least 8% of annual global emissions.

That’s where AI comes in. The researchers trained a generative AI model on environmental impact data and a small public dataset. Using semi-supervised learning, the model sought out concrete formulas that checked all of the researchers’ boxes: 1) lower carbon footprint, 2) significant compressive strength, and 3) similar durability and other qualities.

Once the AI model came up with lower-carbon recipes for concrete, the researchers tasked a concrete supplier with whipping up the new batches. And since the new formulas had performed well in testing, Meta used them in construction on its data center in DeKalb, Illinois—specifically a guard tower and an office building for the construction crew.

  • The authors wrote, “Results from field experiments as part of this real-world deployment corroborate the efficacy of AI-generated, low-carbon concrete mixes.”
  • They also note there's room to improve on their approach, by building a model that better optimizes for things like cold-wearther curing and logistical considerations (e.g., availability of certain materials).

Zoom out: Data centers have long been criticized for their staggering environmental costs, primarily around their energy and water consumption. To tackle that, companies like Microsoft and Google have announced plans to slash water usage and switch to carbon-free energy sources. Low-carbon concrete could be another tool in the toolbox for those building out data centers—and potentially for builders more broadly.

Stay up to date on emerging tech

Drones, automation, AI, and more. The technologies that will shape the future of business, all in one newsletter.