Short answer
An AI PC is a computer with a neural processing unit (NPU), a low-power chip designed to run AI tasks on the device instead of in the cloud. Microsoft's Copilot+ PC label requires an NPU of at least 40 TOPS, 16 GB of RAM and 256 GB of storage, which unlocks Windows AI features such as Recall, live caption translation and image generation. NPUs are great for always-on, battery-friendly AI, but for running large language models or serious AI work, a GPU is still much more powerful.
What makes a PC an "AI PC"?
Every modern computer can run AI software. The term AI PC usually means a computer whose processor includes three kinds of engines:
- CPU: general-purpose tasks, plus small AI jobs.
- GPU: fast parallel maths for graphics and heavy AI workloads.
- NPU: a dedicated, very power-efficient engine for sustained AI tasks such as background blur, transcription and on-device assistants.
Running AI locally means faster responses, better privacy because data stays on the device, and features that work offline. See NPU vs GPU vs TPU for how these chips differ.
Copilot+ PC requirements
Microsoft introduced the Copilot+ PC category in May 2024. To qualify, a Windows PC needs:
| Requirement | Minimum |
|---|---|
| NPU performance | 40 TOPS or more |
| Memory | 16 GB RAM |
| Storage | 256 GB SSD |
Copilot+ features include Recall (searchable snapshots of your activity, opt-in), Live Captions with translation, Click to Do, image generation tools in Paint and Photos, and Windows Studio Effects for webcams.
NPU performance by chip
TOPS means trillions of operations per second, usually measured at INT8 precision. Higher is faster, but figures are not always comparable across vendors; see TOPS vs TFLOPS.
| Chip family | NPU performance | Notes |
|---|---|---|
| Qualcomm Snapdragon X Elite / X Plus | 45 TOPS | First Copilot+ PCs (2024), Arm-based |
| Qualcomm Snapdragon X2 Elite | 80 TOPS | Announced 2025 |
| Intel Core Ultra 200V (Lunar Lake) | 48 TOPS | x86, 2024 |
| AMD Ryzen AI 300 series | 50 TOPS and up | x86, 2024 |
| Apple M4 | 38 TOPS (Neural Engine) | Macs are not Copilot+ PCs but run on-device AI |
| Apple M5 | Neural Engine plus Neural Accelerators in the GPU | Launched October 2025 |
Newer chips launched in 2025 and 2026 continue to raise NPU performance, and PC makers often quote "platform TOPS" that add the CPU, GPU and NPU together. Check the NPU figure specifically when comparing Copilot+ eligibility.
NPU vs GPU: which matters for AI?
| NPU | GPU | |
|---|---|---|
| Strength | Very power-efficient, always-on AI | Raw speed for large models |
| Typical AI performance | 40 to 80 TOPS | Hundreds to thousands of AI TOPS on RTX graphics cards |
| Memory | Shares system RAM | Dedicated VRAM (or unified memory on Macs) |
| Best for | Video calls, captions, assistants, light local models | Local LLMs, image generation, training |
| Software support | Growing, varies by app | Very broad (CUDA, PyTorch, Ollama, LM Studio) |
For running local chatbots and image generators, a discrete NVIDIA GPU or a Mac with lots of unified memory is far more capable than an NPU today. See the best GPU for local LLMs, RTX 5090 for AI and how to run an LLM locally.
Can an AI PC run ChatGPT-style models locally?
Yes, small ones. Tools such as LM Studio, Ollama and Microsoft's Foundry Local can run models with a few billion parameters on laptop chips, using the NPU, GPU or both. The main limit is memory: a 7 to 8 billion parameter model needs roughly 5 to 6 GB at 4-bit, so 16 GB of RAM is a practical minimum and 32 GB is better. See how much VRAM you need for an LLM.
Should you buy an AI PC?
- Yes, if you want long battery life, better video calls and Windows AI features, or you are buying a new laptop anyway: most new mid-range and premium laptops now include a capable NPU.
- Look at the GPU instead, if you want to run larger local models, generate images or learn AI development.
- Memory matters most: choose at least 16 GB, ideally 32 GB, because AI models live in memory.
Frequently asked questions
What is the difference between an AI PC and a normal PC?
An AI PC has a dedicated NPU for running AI features efficiently on the device. A normal PC can still run AI, but on the CPU or GPU, often using more power.
Do I need an AI PC to use ChatGPT?
No. ChatGPT, Claude and Gemini run in the cloud and work on any computer with a browser.
Is a MacBook an AI PC?
Apple does not use the AI PC or Copilot+ labels, but Apple silicon Macs include a Neural Engine and unified memory, and run many on-device AI features and local models well.
How many TOPS do I need?
For Copilot+ features, at least 40 TOPS on the NPU. For local LLMs, memory size and GPU power matter more than NPU TOPS.
Sources and further reading
- Microsoft, Introducing Copilot+ PCs (May 2024)
- Microsoft, Copilot+ PCs developer guide
- Qualcomm, Snapdragon X2 Elite
- Apple, Apple unveils M5 (October 2025)




Leave a Reply