Engineers at the University of California San Diego have developed a new brain-inspired hardware platform designed to overcome energy and speed bottlenecks in artificial intelligence. Published on March 9 in Nature Nanotechnology, the research led by Professor Duygu Kuzum introduces a neuromorphic architecture that combines memory and computation on a single chip. Unlike conventional computers where data must shuttle between separate processors and memory, this system stores and processes information within the same material, mimicking the collective interactions of neurons in the brain. The platform utilizes a hydrogen-doped perovskite nickelate material called neodymium nickelate. When voltage pulses are applied, hydrogen ions move within the material to alter its electrical resistance, creating memory properties that allow nodes to retain recent signals. Crucially, all nodes share a common substrate that functions similarly to the ionic fluid surrounding biological neurons. This design enables signals to spread and interact across the network, meaning the output of any single node depends on the activity of the entire system. This collective behavior allows for spatiotemporal computing, which analyzes signals through both time and spatial dynamics. In two simulated tasks, the device demonstrated superior performance compared to traditional time-based processing. It successfully recognized spoken digits with high accuracy and detected early signs of epileptic seizures from brain-wave recordings. For the seizure detection test, the system identified warning signals using only a few seconds of data, as early signals from one node influenced others to help identify anomalies sooner. The hardware operates with remarkable efficiency, processing data in hundreds of nanoseconds and consuming approximately 0.2 nanojoules per operation. This technology is categorized as brain-inspired rather than brain-like, drawing conceptual ideas from neural networks without attempting to replicate the brain's biological structure. The researchers emphasize that learning in their system emerges from the rich, dynamic interactions of the network rather than isolated components. The extreme energy efficiency and speed make the platform particularly suitable for edge AI applications, where compact devices like wearable health monitors, smart sensors, and autonomous machines must process data locally without relying on large data centers. Currently, the technology remains in an early stage of development. While hardware demonstrations have confirmed small-scale capabilities, larger tasks such as speech recognition and seizure detection were validated through simulations based on experimental measurements. Future work will focus on scaling up the system, integrating it with conventional semiconductor electronics, and exploring additional practical applications. This advancement marks a significant step toward developing compact, energy-efficient hardware capable of keeping pace with the growing demands of artificial intelligence.