Windows AI · Explainer
What Is an NPU? (And Why Windows PCs Have Them Now)
An NPU — neural processing unit — is a dedicated block of silicon built to run artificial-intelligence math efficiently, right on your device. It sits alongside the CPU and GPU inside modern processors and specialises in the repetitive matrix arithmetic behind AI models. Because it is purpose-built for that one job, an NPU can run AI tasks — live background blur, on-device translation, image generation — at far lower power and less heat than doing the same work on the CPU or GPU. That efficiency is exactly why so many new Windows laptops ship with one.
If you've seen "AI PC," "Copilot+ PC" or a "40 TOPS NPU" on a spec sheet and wondered what any of it means, this guide breaks it down: what an NPU actually does, what TOPS measures, how the NPU differs from the CPU and GPU, why Windows suddenly cares, and how to check whether your own PC has one.
An NPU is a specialised AI co-processor. It doesn't replace your CPU or GPU — it handles AI inference more efficiently than either, so features like camera effects can run all day without draining the battery. To see if you have one: Task Manager > Performance > NPU.
What an NPU is, in plain terms
Modern AI models — the kind that recognise a face in a webcam feed, translate speech, or generate an image from a prompt — are, under the hood, enormous piles of multiply-and-add operations arranged as matrices. An NPU is a chip block designed to do that specific kind of math, in parallel, extremely efficiently. Microsoft's Copilot+ PCs developer guide describes it as hardware built to "execute the deep learning math operations that make up AI models."
Two things make an NPU different from a general-purpose processor:
- It's specialised. A CPU is a jack-of-all-trades; an NPU does one family of tasks — AI inference (running a trained model) — and does it well. It is not designed to run your operating system, open apps or browse the web.
- It's efficient. NPUs often work in lower-precision number formats (such as INT8) rather than the full-precision math a CPU or GPU uses. Lower precision is good enough for most inference and uses dramatically less energy, which is why NPU-powered features are gentle on battery life. Microsoft's own primer, All about neural processing units (NPUs), makes the same point: an NPU uses less power and is far more efficient at AI tasks than a CPU or GPU, freeing both up for other work.
In practice, the NPU rarely works alone. On a Windows PC the CPU, GPU and NPU cooperate, and Windows routes each job to whichever is best suited — the CPU for general logic, the GPU for graphics and big parallel bursts, and the NPU for sustained, power-sensitive AI work.
TOPS: a rough measure, not a benchmark
You'll see NPUs rated in TOPS — "trillions of operations per second." It's a headline figure for how much AI math a chip can theoretically push through in a second. A 40 TOPS NPU can, on paper, perform 40 trillion operations per second. Bigger numbers suggest a faster NPU, and TOPS is a handy way to sort chips into rough classes.
TOPS is a peak, theoretical figure. Real-world AI performance also depends on memory bandwidth, driver quality, the model itself, the number format it runs in, and how well a given feature is optimised for that specific NPU. Treat TOPS as a ballpark class indicator — not a precise score to compare two machines down to the last percent.
NPU vs CPU vs GPU: who does what
It helps to think of the three as a team with different strengths:
| Processor | Best at | Trade-off |
|---|---|---|
| CPU | General-purpose work: the OS, apps, logic, one-thing-after-another tasks | Flexible but slow and power-hungry for heavy AI math |
| GPU | Massively parallel work: graphics, gaming, and fast AI training or bursts | Very fast, but draws a lot of power and generates heat |
| NPU | Sustained, efficient AI inference at low power | Not general-purpose; only accelerates AI-style workloads |
A useful mental model: if you need a single AI answer as fast as physically possible and don't care about battery, a big GPU may win. If you need an AI feature running quietly in the background for hours — think live camera effects during a long video call — the NPU is the right tool, because it sips power while the CPU and GPU stay free for everything else.
Why Windows PCs have NPUs now
NPUs aren't brand new — phones have had them for years, and some earlier laptops shipped small ones. What changed is that Windows itself now has AI features designed to run on the device, and those features need an efficient accelerator to feel instant without wrecking battery life. That demand pushed NPUs from a niche extra into a headline spec.
Microsoft formalised this with the Copilot+ PC category. To carry that label, a device must include a processor or SoC with an NPU capable of more than 40 trillion operations per second (40+ TOPS), along with at least 16 GB of RAM and a 256 GB SSD — the figures published on the Windows 11 specifications page. That 40+ TOPS bar is roughly the throughput Microsoft decided was needed to run its on-device AI models responsively. Chips that clear it in 2026 include Qualcomm's Arm-based Snapdragon X series and newer x86 parts from AMD (Ryzen AI) and Intel (Core Ultra) — though what matters is the specific model's NPU rating, not the brand. As vendor figures, Qualcomm rates the Snapdragon X Elite and X Plus NPU at around 45 TOPS, AMD rates the Ryzen AI 300 series NPU at up to about 50 TOPS, and Intel rates the Core Ultra 200V (Lunar Lake) NPU at roughly 47–48 TOPS.
On a Copilot+ PC, the NPU powers features that run locally rather than in the cloud — which tends to make them faster, more private and easier on battery. Examples include Windows Studio Effects (camera and mic enhancements for calls), live caption translation, Cocreator in Paint, and Recall. For the full picture of that hardware category and what it unlocks, see our guide to Copilot+ PCs and our overview of the Windows 11 AI features.
Plenty of laptops have an NPU without qualifying as Copilot+ PCs — their NPU simply doesn't hit the 40+ TOPS bar. Having an NPU is great for efficiency and some AI acceleration, but the specific Copilot+ on-device features require an NPU that clears that higher threshold. An NPU is the ingredient; Copilot+ is a recipe with a minimum spec.
How to check whether your PC has an NPU
The quickest way on Windows 11 is Task Manager, which now shows NPU activity alongside CPU, GPU, memory and disk on supported devices.
- Right-click the Start button and choose Task Manager (or press
Ctrl+Shift+Esc). - Select the Performance tab on the left. If Task Manager opened in its compact view, click More details first.
- Look down the left-hand list of components. If your device has a supported NPU, you'll see an NPU entry beneath CPU, GPU, Memory and Disk.
- Click the NPU entry to see live utilisation, plus details such as the driver version when an AI feature is using it.
No NPU entry? Then Windows either doesn't detect a supported NPU on your system, or your PC doesn't have one. You can cross-check your processor model in Settings > System > About against the manufacturer's spec sheet — but the absence of an NPU is nothing to worry about. Your PC works exactly the same; it just won't run the on-device AI features that depend on one.
Don't be alarmed if the NPU sits at 0% most of the time. It only lights up when an app or Windows feature actively runs an AI model on it — such as switching on a camera effect during a video call. Idle most of the day is exactly what you'd expect; it's a specialist, not the always-on workhorse the CPU is.
Do you need an NPU?
Honestly, for most everyday computing today, no. Browsing, Office, email, media and even cloud AI assistants like Copilot chat all run fine on PCs without an NPU. An NPU becomes worthwhile if you'll genuinely use on-device AI features — frequent video calls with Studio Effects, live translation, local creative tools, or Recall — and you value the battery efficiency that comes from running them on dedicated silicon instead of the CPU.
If you're buying a new laptop anyway, an NPU (and ideally a Copilot+-class one) is a sensible, future-facing bonus rather than something to overpay for. The set of Windows features that lean on the NPU keeps growing, so a machine with a capable one is better positioned for what ships next — but a good PC without an NPU remains a perfectly good PC.
An NPU is an efficient AI co-processor, TOPS is its rough speed rating, and Windows now uses it to run AI features locally. It's a capability, not a requirement — nice to have, essential for no one. Check for yours under Task Manager > Performance > NPU.
Frequently asked
Does my PC need an NPU to run Windows 11?
No. Windows 11 runs perfectly well on PCs that have no NPU at all — the NPU is an optional accelerator, not a system requirement. What an NPU adds is the ability to run certain on-device AI features efficiently. Without one, those specific features are either unavailable or fall back to the CPU, GPU or the cloud, but the rest of Windows works exactly the same.
What is the difference between an NPU and a GPU?
A GPU is a powerful, flexible parallel processor built for graphics and heavy compute, and it can run AI models very fast — but it draws a lot of power. An NPU is a smaller, specialised block tuned only for the low-precision matrix math behind AI inference, so it runs those sustained AI tasks at much lower power and heat. For a quick, one-off burst a GPU may be faster; for all-day background AI like camera effects, the NPU is far more efficient.
Is a higher TOPS number always better?
Not on its own. TOPS (trillions of operations per second) is a rough headline figure for how much AI math a chip can theoretically churn through. Real performance also depends on memory bandwidth, drivers, the model, and how well a feature is optimised for that specific NPU. Treat TOPS as a ballpark class indicator — for example, the 40+ TOPS Copilot+ bar — rather than a precise benchmark you shop by.
How do I check whether my Windows PC has an NPU?
Open Task Manager (right-click the Start button and choose Task Manager), then select the Performance tab. If your device has a supported NPU, an NPU entry appears in the left-hand list alongside CPU, GPU, Memory and Disk, and selecting it shows live utilisation. If no NPU entry is listed, your PC either has no NPU or one that Windows does not surface there.
Browse all how-to guides for more practical, jargon-free Windows walkthroughs, or dig into the windows-now.com archive of restored community posts. You might also like our companion explainers on what a Copilot+ PC is and the Windows 11 AI features.