Meta has launched an open-source Artificial Intelligence tool designed to create concrete mixes that are stronger, more sustainable, and cure faster, representing a major advancement for the construction industry write Julius Kusuma, Sebastian Ament, Eytan Bakshy, Rebeca Ayala for Engineering at Meta. Developed in collaboration with the cement manufacturer Amrize and the University of Illinois Urbana-Champaign (U of I), the tool utilizes Bayesian optimization—powered by Meta’s BoTorch and Ax frameworks—to rapidly discover high-performance, low-carbon concrete formulations. This innovation is critical for advancing sustainability goals, as concrete production globally accounts for approximately 8% of all CO2 emissions, and is a significant source of embodied carbon in data center construction. By making this AI freely available, Meta aims to accelerate the adoption and optimization of sustainable concrete across the construction sector.
Conventionally, concrete formulation focuses mainly on cost and 28-day compressive strength, but modern projects like data centers also require optimization for sustainability, rapid curing speed, and excellent surface quality. Traditional methods for innovating low-carbon concrete are notoriously slow and often present challenges like longer curing times and supply chain complications when introducing novel materials. The complexity lies in optimizing a mix of ingredients—including cement, water, aggregates, and supplementary cementitious materials (SCMs) like fly ash and slag—whose performance can vary significantly by source and condition. The new AI model tackles this multi-objective problem by predicting compressive strength curves, allowing designers to efficiently optimize the trade-off between short- and long-term strength properties and sustainability.
This collaborative research has already produced tangible results, leading to the successful design and deployment of an AI-optimized green concrete mix at Meta’s new Rosemount, Minnesota data center. The AI pipeline was able to quickly exceed the performance of standard low-carbon industry formulas in terms of strength, speed, and sustainability metrics. Specifically for data center slabs, which require exceptionally flat and durable surfaces, the AI algorithms incorporated finishability constraints and identified formulas with faster curing and lower global warming potential that met the strictest quality requirements. The process involved a continuous loop of the AI suggesting promising new mixes, the U of I lab validating them with testing, and the results then refining the AI for subsequent, more efficient iterations.
Looking ahead, Meta and Amrize plan to continue their partnership to scale the use of AI in the concrete industry, while the basic AI solution will remain open source to foster R&D and commercial application across the globe. By providing this technology openly, Meta intends to de-risk the integration of sustainable materials and encourage the industry to adopt performance-based requirements. Meta is also committed to engaging with other major technology and construction organizations, such as iMasons and the Open Compute Project, to publish reference designs, AI-informed formulas, case studies, and best practices, further driving down carbon emissions in large-scale construction.
Read more here: https://engineering.fb.com/2025/07/16/data-center-engineering/ai-make-lower-carbon-faster-curing-concrete/