All terms
Hardware & Systems
Neural Processing Unit
A specialized chip that runs neural network inference efficiently at low power on consumer devices.
Definition
A Neural Processing Unit (NPU) is a specialized processor designed to accelerate the tensor math of neural networks, particularly inference tasks like image recognition, speech processing, and on-device language models. NPUs are usually integrated into the system-on-chip of phones, tablets, and laptops alongside the CPU and GPU. By offering high throughput at low power, they enable real-time AI features locally without relying on cloud compute.