Monolithic 3D Integration of Oxide Semiconductor FETs and Memory Devices for AI Acceleration (Invited)

IoT edge devices are expected to have high-capacity and low-power storage memory, as well as smart data processing capability such as machine learning and AI, to fully utilize big-data by co-working with cloud computing system. Such edge intelligence requires energy-efficient data movement and computation, which opens new challenges and opportunities to integrated devices, circuits and systems. In this paper, we introduce monolithic 3D integration technologies of (1) high-density embedded memory and (2) in-memory computing, both of which are enabled by oxide semiconductor FETs and memory devices, for AI acceleration. We address key reliability issues regarding manufacturability for each technology.