• Meta continues its long-term plan to help the construction industry use AI to produce higher quality, more sustainable concrete mixes and those made exclusively in the United States.
• Coinciding with the American Concrete Institute (ACI) Spring 2026 Meeting, Meta is releasing a new AI model for concrete mix design – Bayesian optimization for concrete (BOxCrete)as well as the fundamental data used to develop award-winning concrete mixes.
• Metas Open source model for sustainable concrete is available today on GitHub.

YearlyThe United States pours about 400 million cubic meters of concrete, enough concrete to pave a two-lane highway that circles the globe several times. IIt is the backbone of our bridges, data centers, highways and homes. Although we produce the majority of our ready-mix concrete domestically, we are Import almost a quarter of the cement from which it is made. Meta’s AI is helping to change that.

Concrete consists of a mixture of cement and cementitious materials, aggregates, water and chemical additives. Concrete suppliers must design concrete mixes that meet competing requirements: strength, speed, ease of use, cost and sustainability. Traditional concrete mix designs rely heavily on trial and error in the laboratory, the engineer’s intuition, and decades of accumulated knowledge—a workflow that is slow and expensive to adapt.

Cement is a key ingredient in concrete, so imported cement can have a significant impact on U.S. suppliers and suppress U.S. production, jobs and investment. While ready-mixed concrete is typically produced domestically, the cement required for it is imported on a large scale About 20-25% of U.S. cement consumption is met by imports. Additionally, cement made in the USA meets U.S. performance and environmental standards not uniform internationally.

At the same time, it must be ensured that the products are manufactured domestically – a process that is often called this Relocation – generally increases manufacturing jobs in the United States. Reshoring and related foreign direct investment (FDI) have returned over 1.1 million jobs to the U.S. since 2020, and manufacturing has one of the highest economic multipliers; Every dollar spent in manufacturing increases the U.S. economy by $2.69. The cement and concrete sector alone more than contributes $130 billion annually and supports about 600,000 jobs – yet imports still meet about 23% of total domestic demand. To capture more of this value at home, U.S.-based concrete manufacturers are looking to incorporate more U.S.-made materials into their mixes.

Different cements have different chemical properties and a mixture that works perfectly with one cement may fail completely with another cement. Therefore, manufacturers need a way to quickly research and validate new formulations without having to spend months in the lab.

Real world implications in the US

Meta and its partners have already received multiple awards for these innovations in concrete design, including: 2025 Building Innovation Award for the best partnership (shared with Amrize) and a Slag Cement Award 2025 for the sustainable concrete project of the year (together with Amrize and the University of Illinois at Urbana-Champaign). However, the impact of this model is also being felt in local collaboration across multiple states through partnerships with major concrete manufacturers and software companies.

Illinois

Meta works closely with the University of Illinois at Urbana-Champaign and Amrize, the largest cement and concrete manufacturer in North America, headquartered in Chicago, Illinois the implementation of AI for sustainable and domestically produced concrete. Amrize operates 18 cement plants, 141 cement terminals and 269 ready-mix concrete sites across North America. Their size makes them an ideal partner to demonstrate how AI can transform mix design at industrial scales. Amrize recently introduced a “Made in America” cement labelwhich guarantees that the cement meets strict US standards and is made in the USA by local workers using American materials. The company also recently announced nearly $1 billion in capital investments in 2026 to increase domestic cement production, among other things.

Meta and Amrize will present at the American Concrete Institute (ACI) Spring Meeting alongside researchers from the University of Illinois Urbana-Champaign to further showcase our partnership Using AI for lower-emission, domestically produced concrete.

In parallel with the event, Meta is releasing a new AI model for concrete mix design, Bayesian Optimization for Concrete (BOxCrete). BOxCrete is an improvement over Meta’s previous models as it is more robust to noisy data and has new features including the ability to predict concrete slump (an important indicator of concrete workability).

In conjunction with BOxCrete, Meta publishes the basic data used for development the new concrete mix used in ours Data center in Rosemount, MN. This basic data is the best systematic basic data for concrete mix performance compared to other published open source datasets.

Meta researchers submitted an article about BOxCrete for publication It describes the new model, the data and the associated methodology.

Minnesota

In partnership with AmrizeMortenson and the University of Illinois at Urbana-Champaign, BOxCrete was used to create a stronger, faster-setting concrete mix that was used on a large scale in a site support area in one of our data center buildings in Rosemount, MN.

The AI-optimized mix was designed for one of the most demanding parts of construction: the massive concrete foundation that supports the weight of thousands of servers and cooling systems. By using domestic materials, the blend reached full structural strength 43% faster than the original formula while reducing the risk of cracking by nearly 10% – proof that AI can help American manufacturers quickly transition to U.S.-made materials without sacrificing quality. Because the data confirms that all structural requirements are met, the mix is ​​now suitable for use in additional areas of the data center.

Pennsylvania

In 2023, Meta released his Concrete optimization AI framework as open source software under the MIT License, allowing wide acceptance from academia to commercial software providers.

In an effort that reflects how AI-driven mix design is becoming part of standard concrete production infrastructure, Pennsylvania-based Quadrela leading SaaS enterprise platform for the ready-mix concrete industry, has adopted Meta’s AI framework into its software. Quadrel has applied it to real-world use cases including data preprocessing, batch and test normalization, feature engineering, and custom model training. The models, which continually improve over time with the incorporation of field test results, have been embedded into daily mix design and quality control workflows, informing day-to-day quality control and operational decisions.

How Meta uses AI for concrete mixing

Metas AI for concrete model can help suppliers incorporate U.S. materials into their blends more quickly through an approach called adaptive experimentation.

This is how it works:

Metas Adaptive Experimentation (Axe) Platform uses Bayesian optimization to intelligently navigate the vast space of possible concrete formulations. Instead of randomly testing mixtures or relying solely on human intuition, AI can:

1. Learns from existing data: Historical mix designs, lab results and performance metrics train the model on what works
2. Suggests candidates with high potential: The AI ​​suggests new blends most likely to meet target specifications and can compare performance between U.S.-made and foreign materials
3. Considers limitations in advance: The user specifies technical requirements and the ingredients to be used.
4. Refines with each test: Each lab result improves the model’s predictions and leads to an automatic improvement loop.

While incorporating AI and adaptive experimentation does not change the process of lab validation, field testing, technical release, and code compliance, it significantly increases the speed of discovery and helps engineers find better starting points with less testing.

Source: University of Illinois at Urbana-Champaign

Building an AI-powered future for concrete

Meta’s AI for Concrete is part of a broader commitment to applying machine learning where it can deliver measurable, real-world impact. While collaborations with Amrize, the University of Illinois and industry software providers like Quadrel represent the first wave of adoption, the goal is an industry-wide shift in how American producers approach mix design.

Meta plans to continue working with the construction industry to develop new AI tools over the next few years. As more platforms like Quadrel build on BOxCrete, AI-optimized mix design becomes accessible to producers without having to change their existing workflows. The team also plans further academic collaboration with the University of Illinois Urbana-Champaign to explore how AI can address not only substitution of domestic materials, but also broader challenges related to concrete sustainability and performance.

By reducing the barriers to adopting domestic materials, Meta helps American manufacturers compete on costs, reduce emissions and strengthen supply chain resilience – one blend at a time.

Join us

Discover Meta’s open source BOxCrete for sustainable concrete on GitHub.

Read our preprint: “BOxCrete: A Bayesian optimization open source AI model for concrete strength prediction and mix optimization.”