Open Electrolyte Databank: Unlocking Molecular-Mixture Intelligence for Battery Electrolyte Discovery through a Standardized, Multimodal Foundation

Meeting AI's growing data-center energy demands will require higher-energy-density storage, and at the heart of those systems are the electrolyte chemistries that govern performance, stability, and safety. Yet the data foundation for electrolyte discovery remains misaligned with the real problem: benchmarks like MoleculeNet standardized AI on single molecules, while electrolyte performance is governed by molecular mixtures, compositions, and experimental context. Current open battery datasets, meanwhile, remain largely centered on full-cell aging and performance with limited chemical diversity. We therefore propose Open Electrolyte Databank, a standardized, multimodal resource centered on molecular-mixture chemistry for battery electrolytes, initially focused on metal-anode systems where electrolyte design is especially critical for achieving very high energy density. The databank will transform fragmented literature into machine-readable, formulation-native records linking standardized compositions to protocols, outcomes, and multimodal data, including spectroscopy, molecular dynamics, and DFT, while growing through contributions from the broader electrolyte community, including labs developing automated battery workflows. Our early work already establishes proof of concept, with a Coulombic Efficiency (CE) dataset about an order of magnitude larger than previous public efforts. By moving battery AI beyond single-molecule prediction and toward formulation-aware learning, Open Electrolyte Databank would provide shared infrastructure for mixture-aware benchmarking, retrieval, and closed-loop electrolyte discovery.