<p>The well-known von Neumann bottleneck has emerged as a significant obstacle in computationally demanding AI tasks, as frequent data transfers between logic and memory impair system performance and reduce energy efficiency, thereby motivating the exploration of in-memory and neuromorphic computing as potential solutions. Among emerging material options, halide perovskites (HPs) exhibit a tunable bandgap, rapid ion migration, mechanical pliability, and facile low-temperature processing, making them particularly advantageous for next-generation computing. This article surveys HP-based memristors and synaptic transistors, emphasizing their underlying physical mechanisms, capacity to replicate synaptic behavior, and multilevel memory functionality within novel computing architectures. Additionally, we address integration challenges, particularly regarding stability, variability, and scalability. By examining recent advances, we highlight HPs as a promising materials platform for next-generation AI hardware that can overcome the limitations of traditional von Neumann architectures.</p> Graphical Abstract <p></p>