Abstract: Computing In-Memory (CIM) is a hardware accelerating approach for AI algorithm, which overcomes the bottleneck of traditional Von Neumann architecture and realizes the cooperation of memory unit and logic unit. At present, CIM faces many key challenges, such as hardware resources reuse, cell design and analog computation.With the development of technology, CIM, with its excellent low power consumption and high response speed, will first be applied to the End-Inference tasks in IoT scenarios. In the future, the hardware carrier of CIM will develop fromNORFlash memory to novel non-volatile memory.

Key words: ComputingIn-Memory, AI, IoT