KEY POINTS
-
A concrete mix at a Minnesota data center reached full structural strength 43% faster using an AI model and U.S.-sourced materials.
-
Meta released BOxCrete, an open-source AI model that helps concrete producers design mixes using domestically sourced cement, reducing reliance on imports that currently account for nearly one in four tons of U.S. cement consumption.
-
Amrize, the largest cement and concrete manufacturer in North America, announced close to $1 billion in 2026 capital investments to increase domestic cement production and launched a Made in America cement label.
According to a March 30 post from Meta Engineering, the company released an open-source AI model designed to help concrete producers reformulate mixes around American-made materials, cutting the time and cost of switching away from imported cement.
Nearly one in four tons of cement used in U.S. concrete production is imported, the company wrote. This presents a supply chain vulnerability for a $130 billion domestic sector that supports roughly 600,000 jobs.
Modern Concrete's Complexity
Modern concrete must meet competing demands for strength, workability, durability, and sustainability, that make mix design increasingly complex. AI and machine learning models have shown promise for predicting compressive strength and guiding optimization, but most existing tools rely on proprietary datasets and closed-source code.
Meta released BOxCrete (Bayesian Optimization for Concrete), an AI model that uses adaptive experimentation to navigate possible concrete formulations and identify high-performing mixes faster than traditional trial-and-error lab methods.
BOxCrete is trained on a publicly available dataset of more than 500 strength measurements, from 123 mixes (69 mortar and 54 concrete), and tested at five curing ages from 1 to 28 days.
Why It Matters for Construction
The U.S. pours roughly 400 million cubic yards of concrete annually. Concrete is formed when cement is mixed with water and aggregate, such as sand and rock, and are allowed to harden. While most ready-mix concrete is produced domestically, 20–25% of the cement required for it is imported.
Cement chemistry varies by source. A concrete mix designed for an imported cement may not perform the same way with a domestic substitute. Therefore, producers seeking to use domestic sourced cement must determine a new formula and test it, a process that can take months in a lab.
BOxCrete artificial intelligence addresses this by learning from existing mix data, proposing high-potential formulations, incorporating technical constraints upfront, and refining predictions with each lab result.
Meta says the tool does not replace lab validation, field trials, engineering sign-off, or code compliance. However, it accelerates the discovery phase, helping engineers find better starting points with fewer tests.
On-the-Ground Results from Meta's Minnesota Data Center
One of the model's proof points comes from Meta's data center build in Rosemount, MN. In partnership with Amrize, Mortenson, and the University of Illinois at Urbana-Champaign, BOxCrete generated a concrete mix for the massive foundation slab, engineered to support thousands of servers and cooling systems.
Using domestically sourced materials, Meta Engineering reported that the AI-optimized concrete:
-
Reached full structural strength 43% faster than the original formula.
-
Reduced cracking risk by nearly 10%.
The mix, developers said, qualified for use in additional areas of the data center.
Shown in a rendering is the Meta data center in Rosemount, MN, presently under construction. The 715,000-square-foot project is home to a BOxCrete-designed concrete mix poured in the foundation slab. The open-source AI model is designed to help concrete producers reformulate mixes around American-made materials, cutting the time and cost of switching away from imported cement. Image: Meta
Industry Partnerships and Scale
Meta's primary industry partner is Amrize, the largest cement and concrete manufacturer in North America. From its headquarters in Chicago, Amrize operates 18 cement plants, 141 cement terminals, and 269 ready-mix concrete sites across North America.
The company recently launched a Made in America cement label that guarantees U.S. manufacturing standards and a domestic workforce. Amrize also announced close to $1 billion in capital investments in 2026 to increase domestic cement production.
In Pennsylvania, Quadrel, whose software provides concrete performance and quality solutions to the ready-mix industry, has adapted Meta's AI framework into its software platform.
Quadrel has integrated the AI into its daily mix design and quality control workflows, training the models on customer-specific data that improve over time as field results come in.
The Reshoring Context
The push to reduce cement imports aligns with a broader reshoring trend. Reshoring and related foreign direct investment have brought more than 1.1 million jobs back to the U.S. since 2020, Meta says in its post.
Every $1.00 spent in domestic manufacturing adds $2.69 to the U.S. economy, according to the National Association of Manufacturers.
Additionally, U.S.-produced cement meets performance and environmental standards that imported cement may not match consistently.
What's Next
BOxCrete is open-source and freely available. Producers and software platforms can integrate it without changing existing workflows.
Meta says it plans further collaboration with the construction industry to develop additional AI tools and continue academic work on concrete sustainability and performance.
Stay Connected
Stay connected with ConstructConnect News for construction industry news and construction market analysis to stay ahead of what’s building next.
About ConstructConnect
At ConstructConnect, our software solutions provide the information that construction professionals need to start every project on a solid foundation. For more than 100 years, our keen insights and market intelligence have empowered commercial firms, building product manufacturers, trade contractors, and architects to make data-driven decisions, streamline preconstruction workflows, and maximize their productivity. Our newest offerings—including our comprehensive, AI-assisted software—help our clients find, bid on, and win more projects.
ConstructConnect operates as a business unit of Roper Technologies (Nasdaq: ROP), a constituent of the Nasdaq 100, S&P 500, and Fortune 1000.
For more information, visit constructconnect.com
