Concrete results: Meta uses AI to create 35% lower-carbon concrete for data centers

Meta and Amrize have developed an AI-optimized concrete mix specifically designed for Meta’s data center construction in Rosemount, Minnesota. The collaboration, which includes the University of Illinois, produced a concrete formulation that reduces carbon emissions by 35% while meeting the structural requirements for data center infrastructure.

What you should know: The partnership combines Amrize’s materials science expertise with Meta’s open-source AI models to create high-performance, low-carbon concrete.

• The AI-designed ECOPact mix delivers high strength, controlled set time, and a reduced carbon footprint compared to traditional concrete formulations.
• University of Illinois researchers contributed critical lab-generated data to train the AI models used in the concrete optimization process.
• The new concrete has already proven effective for critical slab-on-grade applications at Meta’s Rosemount facility.

The big picture: This initiative represents Meta’s broader strategy to decarbonize its massive data center infrastructure as the company prepares to invest “hundreds of billions of dollars” in AI-focused facilities.

• Meta has embedded low-carbon requirements into its data center construction specifications, targeting up to 20% reduction in concrete’s carbon intensity below regional baselines.
• The company is actively redesigning infrastructure to reduce concrete use where possible and prioritizing low-carbon alternatives in essential applications.

Why this matters: With Meta accelerating its AI infrastructure spending and building a dedicated artificial super intelligence team, sustainable construction materials become critical for meeting net-zero goals at scale.

• Traditional concrete production is a significant source of carbon emissions, making AI-optimized alternatives essential for sustainable data center expansion.
• The approach demonstrates how AI can simultaneously optimize for performance, cost, and environmental impact in construction materials.

What they’re saying: Industry leaders emphasize the potential for AI to transform concrete formulation for specialized applications.

• “Using AI, we can optimize our specialized concrete formulations for data center requirements, from performance needs like strength and durability to thermal regulation and energy-efficiency,” said Jaime Hill, president of Amrize Building Materials.
• “AI-driven mix design lets us optimize concrete for performance, cost and carbon in one step,” added Nishant Garg of the University of Illinois, who led the data-generation effort.
• “We work to design our data centers as efficiently and sustainably as possible, while driving our AI ambitions forward,” said Julius Kusuma, research scientist at Meta.

Competitive landscape: Meta is engaging with multiple industry initiatives to accelerate low-carbon concrete adoption across the digital infrastructure sector.

• The company participates in coalitions like the iMasons Climate Accord and partners with the American Concrete Institute and Open Compute Project to test and validate new concrete technologies.
• Meta has backed startups including CarbonBuilt and CarbiCrete to accelerate commercialization of cement alternatives that capture or avoid CO2 during production.
• The company has piloted alternative concrete mixes using fly ash, slag, and other supplementary materials across its data center portfolio.