New AI simulates 4 billion atoms to build carbon-neutral concrete

Researchers at USC Viterbi School of Engineering have developed Allegro-FM, an AI-powered simulation tool capable of modeling the behavior of over 4 billion atoms simultaneously with 97.5% efficiency. This represents a nearly 1,000-fold increase in scale compared to traditional models. Using the Aurora supercomputer at Argonne National Laboratory, Allegro-FM enables scientists to virtually test molecular compositions of concrete, accelerating the discovery of carbon-neutral, stronger, and fire-resistant formulations without costly lab experiments. The AI system can simulate complex atomic interactions across 89 chemical elements, making it versatile for applications beyond concrete, including carbon storage, battery chemistry, and biomedical devices. The key breakthrough is Allegro-FM’s ability to predict atomic interactions without relying on element-specific quantum mechanical equations, instead using machine learning trained on extensive datasets. This allows for highly accurate and efficient simulations of concrete’s mechanical and structural properties. The researchers found that incorporating CO₂ into concrete can trap carbon dioxide released during cement production,