What Is an NPU? AI PCs and Neural Processors Explained
AI PCs come with an NPU, a chip built for running AI efficiently. Here's what it does, how it differs from a CPU and GPU, what TOPS means and whether you need one.

In this article
- What is an NPU?
- CPU vs GPU vs NPU: what each one does
- What does TOPS mean?
- What is an AI PC, and what is a Copilot+ PC?
- What NPUs actually do for you today
- Do Macs and phones have NPUs?
- How to check if your PC has an NPU
- Do you need an NPU in your next laptop?
- Frequently asked questions
- Does an NPU make my computer faster overall?
- Is 40 TOPS a lot?
- Can I add an NPU to my existing PC?
- Do I need an NPU to use ChatGPT, Copilot or Gemini?
- Is an AI PC worth it?
- The bottom line
An NPU, or neural processing unit, is a chip (or a section of a chip) designed specifically to run artificial intelligence tasks efficiently. Laptops marketed as "AI PCs" include an NPU alongside the usual CPU and GPU so they can handle features like background blur, live captions and on-device image tools without draining the battery. Whether you actually need one depends on which AI features you plan to use, and where those features run.
In this explainer we cover what an NPU does, how it differs from a CPU and GPU, what "TOPS" means, what makes a PC a Copilot+ PC, and how to decide if an NPU should influence your next laptop purchase.
What is an NPU?
Modern AI models, from speech recognition to image generation, mostly boil down to enormous numbers of simple math operations, especially multiplying and adding large grids of numbers (matrices). A neural processing unit is hardware built to do exactly that kind of math in parallel, using as little power as possible.
Microsoft describes the NPU as "a specialized computer chip for AI-intensive processes like real-time translations and image generation" in its Copilot+ PC developer guide. In practice, the NPU is usually built into the same processor package as the CPU and graphics, rather than being a separate card you install.
NPUs are not new. Phones have had similar AI accelerators for years, and Apple has included a "Neural Engine" in its iPhone and Mac chips. What changed is that Windows laptops started shipping with NPUs powerful enough for Microsoft to build operating-system features around them.
CPU vs GPU vs NPU: what each one does
Your computer now has three different engines that can run AI work. Windows and apps try to send each job to the most suitable one.
| Processor | Best at | AI strengths | AI weaknesses |
|---|---|---|---|
| CPU (central processing unit) | General tasks, running the OS and apps, quick decisions | Flexible; runs any AI code | Slow and power-hungry for large AI workloads |
| GPU (graphics processing unit) | Graphics, games, video, heavy parallel math | Very fast for big AI models, especially discrete GPUs | Uses a lot of power; drains laptop batteries quickly |
| NPU (neural processing unit) | Sustained AI tasks at low power | Efficient for always-on features such as camera effects and captions | Less flexible; apps must be built to use it |
The key word for NPUs is efficiency. A powerful gaming GPU can often run a large AI model faster than a laptop NPU, but it does so while consuming far more energy. The NPU's job is to run steady, background AI work quietly so your battery lasts and your CPU stays free for everything else.
What does TOPS mean?
NPU performance is usually quoted in TOPS, short for trillions of operations per second. A higher number generally means the NPU can process more AI work at once.
A few caveats are worth knowing before you compare spec sheets:
- TOPS figures are peak numbers. Real performance depends on the model, the software and how well an app is optimized for a particular chip.
- Precision matters. TOPS are typically measured with low-precision integer math (such as INT8). AI models often need to be converted, or "quantized", to run on an NPU at all.
- Some brands quote "platform TOPS". This adds up the CPU, GPU and NPU together, which produces a bigger number than the NPU alone. Check which figure you are looking at.
What is an AI PC, and what is a Copilot+ PC?
"AI PC" is a broad marketing term. Chipmakers and laptop brands use it for almost any computer with an NPU. There is no single official definition.
"Copilot+ PC" is narrower. It is Microsoft's label for Windows 11 computers that meet specific hardware requirements so they can run its on-device AI features. Microsoft says these features need an NPU capable of more than 40 TOPS, and its Windows 11 specifications page adds that Copilot+ PCs need at least 16GB of RAM and 256GB of storage. Qualifying processors include the Qualcomm Snapdragon X series and selected AMD Ryzen AI and Intel Core Ultra chips, such as the Ryzen AI 300 series and Core Ultra 200V series, and Microsoft updates the list as new generations arrive.
Features Microsoft has built for Copilot+ PCs include Windows Studio Effects for webcams (such as background blur and eye contact correction), Live Captions with translation, image tools in Paint and Photos, and Recall, an optional feature that lets you search your past activity. Availability varies by region, device and Windows version.
What NPUs actually do for you today
Here is where an NPU tends to make a real difference in everyday use:
- Video calls: Background blur, noise suppression and framing effects can run on the NPU instead of taxing the CPU, which helps battery life during long meetings.
- Captions and translation: Turning speech into text locally, in real time, is a natural fit for an NPU.
- Photo and image features: Upscaling, object removal and some generative image tools can run on the device.
- On-device AI models: Developers can run smaller language and vision models locally, which keeps data on your PC and works offline.
And here is what an NPU does not do: it does not speed up web browsing, office apps or most games. It also does not affect cloud-based AI tools. When you use a chatbot in a browser, the heavy lifting happens on the company's servers, not on your laptop. If you mainly use AI through websites, an NPU is a nice-to-have, not a must-have. (If you are getting started with those tools, our guide to writing better AI prompts will help more than any chip upgrade.)
Do Macs and phones have NPUs?
Yes, under different names. Apple's M-series and A-series chips include a Neural Engine, which powers features such as on-device dictation, photo analysis and Apple Intelligence. Most recent flagship Android phones also include dedicated AI hardware in their chips. The concept is the same: a specialized, power-efficient block for AI math.
How to check if your PC has an NPU
On Windows 11, open Task Manager (press Ctrl + Shift + Esc) and select the Performance tab. If your computer has a supported NPU with up-to-date drivers, it appears in the list alongside CPU, Memory, Disk and GPU, with its own usage graph. Microsoft's support article All about neural processing units (NPUs) explains what NPUs are and how they handle AI workloads. You can also check your processor model in Settings > System > About and look it up on the manufacturer's website.
Do you need an NPU in your next laptop?
Use this quick guide:
- Get one if you want Microsoft's Copilot+ features, take lots of video calls on battery, care about privacy-focused on-device AI, or want a laptop that stays relevant as more apps add NPU support.
- Do not pay a big premium for one if you mostly use cloud AI tools, play games, or do traditional work where the CPU, RAM and screen matter more.
- Consider a strong GPU instead if you want to run large AI models locally or do serious AI development. Discrete GPUs remain the faster option for heavy workloads.
Because many current laptop processors include an NPU anyway, you will often get one without trying. Focus first on the fundamentals: enough memory (see how much RAM you need), a good display, battery life and build quality. Our best laptops for students guide applies the same priorities.
Frequently asked questions
Does an NPU make my computer faster overall?
Not for general tasks. It speeds up and improves the efficiency of AI features that are programmed to use it. Everyday speed still depends mostly on your CPU, RAM and storage.
Is 40 TOPS a lot?
It is the threshold Microsoft set for its Copilot+ PC features, and it is plenty for current on-device features like captions and camera effects. It is far below what large data-center AI chips deliver, which is why big chatbots still run in the cloud.
Can I add an NPU to my existing PC?
Generally, no. Laptop NPUs are built into the processor. Some desktop add-in AI accelerators exist, but they will not make an older PC eligible for Copilot+ features.
Do I need an NPU to use ChatGPT, Copilot or Gemini?
No. Those services run in the cloud and work on almost any modern computer or phone with an internet connection.
Is an AI PC worth it?
It can be, but buy it for the overall laptop, not the label. An NPU is a useful extra that will become more valuable as software catches up.
The bottom line
An NPU is a specialized, power-efficient processor for AI tasks. It makes features like webcam effects, live captions and on-device image tools run smoothly without hammering your battery, and it is required for Microsoft's Copilot+ PC features. It will not speed up everyday work or cloud chatbots, so treat it as one factor among many when choosing a laptop. For more buying advice, visit our Laptops & PCs section.




0 Comments
Join the discussion. Comments are moderated and your email is never published.