AI processing used to mean sending your request to a distant server and waiting for an answer. More of that work is now happening right on the machine in front of you. Here's what the specs actually mean, what's new on the market this fall, and why timing your purchase matters right now.
Why on-device processing is worth caring about
When a computer can handle AI tasks locally, without sending anything over the internet, it changes a few things in your favor. Your documents, photos, and files never have to leave your desk. There's no monthly subscription for cloud processing, and no dependency on a fast connection. It also tends to just feel faster, since there's no round trip to a remote server involved.
The three specs that actually matter: NPU, RAM, and VRAM
If you're shopping for a new PC and want it to handle on-device AI features well, these are the numbers worth checking before anything else:
- NPU (Neural Processing Unit): a dedicated chip built specifically to handle AI tasks efficiently, separate from the CPU and GPU. Microsoft's "Copilot+ PC" standard requires an NPU rated at 40 TOPS (trillions of operations per second) or higher.
- System RAM: 8GB is no longer enough for a modern machine running AI features alongside everyday use. 16GB is the realistic minimum, and 32GB is worth the extra cost if you want things to stay smooth.
- VRAM (video memory): if you're doing anything creative — generating or editing images and video with AI tools — you'll want a dedicated graphics card. 8GB of VRAM is a starting point; 12–16GB gives you more headroom.
What's actually on the market this fall
A few notable releases this season are built specifically around on-device AI performance:
- Apple's M5 Mac Studio: the M5 Max and M5 Ultra chips can be configured with up to 512GB of unified memory, letting creative professionals run large AI models directly on the machine.
- AMD Ryzen AI Max PRO 400: AMD's latest chips support up to 192GB of unified memory for Windows users who want top-tier local AI performance.
- Intel Core Ultra Series 3 ("Panther Lake"): focused on running AI tasks efficiently in the background without draining laptop battery life.
- NVIDIA DGX Spark: pairs NVIDIA's Grace CPU with Blackwell graphics architecture in a compact desktop unit, aimed at developers and enthusiasts who want workstation-level AI power without a full workstation footprint.
Component prices are rising — timing matters
These AI-capable machines lean heavily on high-capacity memory and fast SSD storage, and suppliers are currently seeing a shortage of both. That's pushing RAM and storage prices up industry-wide. If you've been putting off a new build or an upgrade, it's a reasonable time to move sooner rather than later before component costs climb further.
If you're weighing a new AI-capable build against upgrading what you already have, a custom PC build consultation or an on-site diagnostic is the fastest way to get a straight answer on which makes more sense for your setup and budget.