Circuit Design Challenges in Computing-in-Memory for AI Edge Devices

Abstract

Computing-in-memory (CIM) structures are meant to overcome the memory bottleneck and improve energy efficiency for artificial intelligence (AI) edge devices. In this article, we review recent trends in the development of CIM macros for the Internet of Things and AI applications. We also look at recent advances in the development of CIMs based on SRAM and nonvolatile memory for AI edge devices as well as the challenges involved in circuit design.