K-Culture GlossaryNInfrastructure and chips
Neural Processing Unit (NPU)
Neural Processing Unit
A chip designed specifically for AI computation — the real substance behind "AI phone" and "AI PC" marketing.
In plain words
Neural Processing Unit — a chip designed for one job only: running neural network calculations. Unlike the CPU (does everything) or GPU (parallel computation), the NPU is optimized for AI inference, so it's fast while using little power.
Its defining trait is that it fits inside phones and laptops. The "AI phone" and "AI PC" marketing you see is really just marketing for having an NPU on board — on-device AI like translation, photo enhancement, and summarization that works without an internet connection runs on this chip. So when a smartphone launch article says "NPU performance up several times," it means "the AI the device can run on its own just got that much bigger."
GPU vs. NPU vs. TPU: What's the Difference?
Is it a low-power dedicated chip inside a device?, Is it a general-purpose processor?, Is it a chip Google built for its own use?
AI chip articles often mention all three in the same paragraph, making it hard to tell who's actually competing with whom. They're not fighting over the same turf — one dominates the data center, one sits inside your phone or laptop, and one was built by Google to run its own services.
| Aspect | Neural Processing Unit (NPU) | GPU | Tensor Processing Unit (TPU) |
|---|---|---|---|
| Where it lives | Inside your phone or laptop | Data centers. Cloud servers | Google data centers |
| Who makes it | Device makers like Apple, Qualcomm, Intel | NVIDIA, AMD | Google (manufactured with Broadcom's help) |
| Who uses it | Anyone using a device — usually without realizing it | Nearly everyone building and selling models | Google and Google Cloud customers |
| What it's good at | Running small models while saving power | The heaviest training and large-scale serving | Large-scale training and serving, cheaply |
| In the news | The reality behind "AI phone" and "AI PC" | "GPU shortage race," "NVIDIA earnings" | "Challenging NVIDIA's dominance" |
| Can you buy one? | It's already built into any device you buy | Yes, though supply falls short of demand | No — you rent it through Google Cloud |
Rule of thumbIf a model is built in a data center, it's a GPU. If it runs inside your own device, it's an NPU. If it's behind a Google service, it's a TPU.
See also
Stories using this term
No story has used this term yet. New ones attach here automatically.