Artificial
intelligence memory is expected to acquire, calculate,
and analyze a large amount of logical information and data in time
to dynamically respond to artificial neural networks. It is the most
promising candidate for realizing a new hardware artificial intelligence
architecture that mimics biological neural networks. However, the
research on artificial intelligence memory is still in the initial
stage, and there are some unresolved bottlenecks for the preparation
of artificial intelligence memory devices. Such as it require external
power supplements for the operating of memory devices, resulting in
high power consumption and difficulty in real-time neuromorphic computing.
Fortunately, self-powered memory devices can perfectly solve the above
problems. In this Review, we have systemically summarized the current
development on material, integration, and technology for the self-powered
memory application, as well as provide the prospect, suggestion, and
optimization method for neuromorphic computing and artificial intelligence
with self-powered memory.