Data-driven inorganic material design for future all-solid-state batteries: Innovations and perspectives

Data-driven approaches have emerged as a transformative force in accelerating the discovery and optimization of inorganic materials for all-solid-state batteries (ASSBs), leading to substantial progress in recent years. Earlier reviews often focus on isolated methodologies or specific material classes. Here, this study systematically highlights the pivotal role of machine learning in advancing solid electrolytes and cathode materials in ASSBs, targeting critical properties including high ionic conductivity, wide electrochemical stability windows, and interfacial compatibility. We chart the methodological evolution from conventional descriptor-based predictive models to advanced deep learning architectures that autonomously extract features from composition and crystal structure, and further to integrated active-learning platforms that close the loop between prediction and experimental validation. Beyond summarizing these technological strides, we critically assess current limitations, particularly the gap between idealized models and the complexity of real devices. Finally, a forward-looking research agenda is outlined, advancing physics-guided and multi-fidelity learning frameworks, developing generative models for inverse design under realistic constraints, extending modeling to interface and cell-level complexity, and establishing reliable closed-loop discovery systems. This work will consolidate current knowledge and chart a path toward more predictive, actionable, and integrative artificial intelligence tools that can accelerate the realization of high-performance, commercially viable solid-state batteries.