NPU

What is an NPU?

An NPU — neural processing unit — is a chip built specifically for neural-network inference: running a trained model so it can make predictions. Its specialty is doing AI fast while sipping power, small enough to live inside phones, earbuds and cameras.

How is it different from a CPU or GPU?

CPU: does everything, but AI isn't its strong suit
GPU: plenty of compute, but power-hungry and bulky
NPU: only does AI, and draws so little power your phone lasts all day
Each has a job. The NPU fills the "on-device AI" gap the others leave open.

Where do NPUs live?

Your phone's AI photo editing, voice assistant, live translation and face unlock — much of that runs locally on an NPU, so the data never has to leave the device. That's faster and more private. Laptops, smart speakers, security cameras and car cockpits are piling on too.

Why does on-device AI need it?

Putting every bit of AI in the cloud means living with network lag and privacy trade-offs. An NPU lets a slice of AI run right on the device — works offline, saves data and battery, and keeps things private. Phone makers now treat this "edge AI" as one of their fiercest battlegrounds.

Bottom line: an NPU is the little AI engine inside your device, running AI offline on a trickle of power.