Microsoft MatterSim

Microsoft Research's Materials project develops AI-powered tools to design and simulate inorganic materials for in silico discovery and optimization. MatterGen is a diffusion-based generative model that can create novel inorganic materials across the periodic table and can be fine-tuned to target properties such as chemistry, symmetry, and mechanical, electronic, and magnetic characteristics, achieving state-of-the-art generation and outperforming traditional screening. MatterSim is a deep-learning material simulator that predicts properties over many elements, temperatures (0–5000 K) and extreme pressures, enabling accurate, efficient in silico materials design, with both models deployed on Azure AI Foundry.

AI named Microsoft MatterSim in September 2026.