Meta unveiled BOxCrete, an AI model designed to optimize concrete mixes using domestic materials, reducing reliance on imported cement. The $1 billion investment by Amrize, a North America cement giant, aims to boost U.S. manufacturing. BOxCrete leverages Bayesian optimization to predict concrete performance, cutting lab testing time by 43% while enhancing sustainability.

Amrize and Meta partnered with the University of Illinois to deploy BOxCrete in Rosemount, MN, where the AI-designed mix achieved full strength 43% faster than traditional methods. This innovation addressed challenges in using domestically sourced materials, ensuring compliance with U.S. standards. The collaboration also earned a 2025 Slag Cement Award for sustainable concrete projects, highlighting its industry impact.

Meta’s open-source AI framework, released under the MIT license, enables broader adoption. Quadrel, a Pennsylvania-based SaaS platform, integrated BOxCrete into its workflows, streamlining mix design and quality control. This adoption underscores AI’s role in modernizing ready-mix concrete production, which supports 600,000 jobs and contributes $130 billion annually to the U.S. economy.

By prioritizing U.S.-made materials, BOxCrete addresses supply chain vulnerabilities and fosters reshoring. The model’s success in data centers and highway infrastructure demonstrates its scalability, offering a blueprint for sustainable, cost-effective construction nationwide.