Recent Progress and Opportunities in Oxide Semiconductor Devices for In‐Memory and Neuromorphic Computing

ABSTRACT The increasing complexity of artificial intelligence has exposed critical limitations of conventional von Neumann architectures, particularly in terms of data transfer bottlenecks and high energy consumption. Consequently, alternative paradigms such as in‐memory and neuromorphic computing have attracted significant attention. Oxide semiconductors, which have achieved commercial success in the display industry, have recently garnered significant attention for neuromorphic computing applications due to their unique properties, including extremely low leakage current, low processing temperatures, and excellent compatibility with back‐end‐of‐line integration with conventional silicon circuits. This review discusses recent advancements and challenges in oxide semiconductor‐based devices for in‐memory and neuromorphic computing. It explicitly addresses multilevel memory devices optimized for analog multiply‐accumulate operations, highlighting key trade‐offs among retention, endurance, operational speed, and energy efficiency. Neuromorphic synaptic devices utilizing oxide semiconductors are highlighted for their effective emulation of synaptic behaviors for spiking neural networks. Additionally, recent developments in optoelectronic neuromorphic systems and reservoir computing using oxide semiconductors are presented, along with insights into emerging device structures and future opportunities for 3D integration to maximize computing efficiency.