How Meta is using AI to make better concrete

Meta has developed an AI tool that formulates concrete mixes that are stronger, faster to make and more sustainable. According to the tech giant, AI can accelerate construction work and mitigate the environmental impact of concrete.

Traditional concrete flooring is optimised for strength, but increasingly construction projects are demanding concrete that is also sustainable, quick-curing and easy to work with. Current low-carbon concrete formulas often have slow curing speeds and quality issues. Some types of low-carbon concrete require novel materials that are not easy to source.

Meta’s AI tools have been developed in partnership with cement manufacturer Amrize and the University of Illinois Urbana-Champaign. AI solves the complex challenge of optimising concrete mixtures, which have several ingredients, including cement, low-carbon supplementary materials, water, binders, aggregates, and admixtures. Concrete performance is also affected by location and weather. To find the right ratio of concrete ingredients for a specific location normally requires tests, which can take weeks to find the best mix. Meta uses its AI Bo Torch and AI tools in open-source software that optimises concrete composition and forecasts compressive strength and sustainability.

Meta says:

“We believe AI can generate high-performance, low-carbon concrete mixes for major construction projects.”

Concrete developed by AI and Meta engineers has been successfully used at Meta’s data centre in Rosemont, Minnesota. The concrete mix contains 8.4% cement, 1.5% fly ash, 1.97% slag, 5.33% water, 35.9% fine aggregate, and 46.9% coarse aggregate.

Meta’s published concrete AI model is available to view on GitHub.