Ångström AI

Ångström AI builds AI physics simulations for pharmaceutical research, with the goal of replacing laboratory experiments with computation. Its models compute properties such as solubility, lipophilicity, crystal structure stability, free energy differences, binding conformations and hydration sites, and the company states that the results match experimental error while running orders of magnitude faster than traditional simulation. The offering is aimed at drug discovery and solid form teams in pharma that need reliable physical property predictions earlier in a pipeline. Its work rests on machine learning interatomic potentials trained to reach the accuracy of density functional theory. Published research from the team includes a potential for crystal structure prediction that runs far faster than DFT, so many more candidate structures can be evaluated when derisking solid forms; an alchemical protocol that computes solvation free energies of small molecules with experimental accuracy, published in JACS in 2025; and generative methods for sampling Boltzmann distributions of molecules, presented at ICLR and NeurIPS in 2023. The crystal structure work was carried out in collaboration with industry scientists. Pricing is not published. Access is arranged directly with the company, which invites pharma teams to get in touch by email for a demonstration on their own discovery pipeline. The company was cofounded by Laurence Midgley as chief executive and Gábor Csányi as head of science, together with a head of AI, and it is advised by crystal structure prediction specialist Graeme Day. Buyers should expect an engagement led by a research team rather than a self serve software product.