I just spent $3,000 on a new gaming PC with a high-end AMD CPU and a screaming-fast Nvidia GeForce 5080 GPU. You’d think I’d be able to run the latest AI features in Windows with that, right? And yeah, I’d love AI-powered file search! But my PC can’t have it—no desktop PC can.
Right now, there’s no way to put together a desktop PC that meets Microsoft’s requirements for accessing Copilot+ AI features. You need a qualified NPU for Copilot+, period. (What’s an NPU?) It’s been this way ever since Microsoft’s Copilot+ PCs launched in 2024, and there still aren’t any qualified NPUs for desktops. Intel’s newer desktop NPUs aren’t fast enough and AMD’s desktop CPUs don’t have NPUs yet.
Long story short, if you want a proper desktop PC, then no Copilot+ AI features for you! This is one of the biggest problems with Microsoft’s Copilot+ movement, and there’s no good reason for it.
The dual meaning of “AI PC” in 2025
The term “AI PC” has gotten pretty confusing. In recent months, I’ve reviewed multiple laptops classified as “AI PCs” or “AI laptops,” but that has meant different things from different people:
- Some so-called AI PCs are lightweight laptops with NPUs but no GPUs. These are proper Copilot+ laptops with access to Copilot+ features and perhaps even preinstalled AI applications that connect to OpenAI’s GPT models or other cloud-based AI solutions.
- Other so-called AI PCs are heavier, pricier, performance-based gaming laptops without NPUs but with powerful GPUs. These are high-end laptops that can use their graphics processor for local AI applications, yet can’t run Copilot+ AI features.
These two terms are basically the opposites of each other—completely different hardware and completely different features. But manufacturers tend to call both “AI laptops,” which muddies the waters.
The PC industry dubbed 2024 “the year of the AI PC,” and yet those first-wave AI PCs were left behind by Copilot+ PCs. What’s worse, many new PCs being sold today in 2025 are still being left behind.
The solution here is simple: Microsoft should let PCs run Copilot+ AI features on GPUs. Then all the so-called AI PCs would be able to run AI features, not just the ones equipped with cutting-edge NPUs.
GPUs can handle it, so let ’em rip
Here’s the question you have to ask yourself if you’re interested in a desktop PC right now: Do you want to wait an unspecified amount of time for desktop NPUs to be announced and released some time in the future, or do you want to get a desktop PC right now?
As of mid-2025, Copilot+ AI features only run on laptops with qualified NPUs. That means Qualcomm Snapdragon X Elite/Plus laptops, Intel Lunar Lake laptops, and AMD Ryzen AI 300 series laptops. Technically, there are also a few lightweight mini PCs with mobile CPUs, like the Asus ROG NUC series, that also fit the bill. No desktop PCs.
For me, I decided that Copilot+ features didn’t matter enough to wait. I have a Surface Laptop 7 with Insider builds of Windows 11 and I never use the Copilot+ features on it. I have Recall activated, but I haven’t touched it in months. (Privacy concerns aside, the worst thing about Recall is that it’s just not very interesting or useful.) Other AI features—like generating images directly in the Photos app—feel like tech demos at best.
Chris Hoffman / Foundry
Still, looking at Microsoft’s feature announcements in the past few days, it’s clear that Microsoft is going full steam ahead with Copilot+, and some of those newer features are getting more interesting. Why yes, I would like AI to change PC settings more quickly! And guess what? My PC’s GeForce 5080 GPU could certainly run features like that… if Microsoft would just let it. (My blazing-fast AMD CPU probably could, too.)
When it comes to running AI tasks, NPUs sit somewhere between CPUs and GPUs—they’re faster at AI processing than CPUs, but slower than GPUs. NPUs are also way more power efficient than GPUs, and that’s a big deal when you’re handling AI tasks on a laptop. But with a desktop PC that’s always plugged in? Battery life doesn’t matter.
And while AI features might hog GPU resources, I don’t necessarily care if my GPU takes a hit for a second when I search the Settings app, for example. There’s so much potential to take advantage of in a GPU that isn’t doing anything else. Might as well use it for AI, right?
It’s not just about Windows AI features
All of the above aside, this whole problem encompasses so much more than just the Copilot+ AI features built into Windows.
Microsoft is encouraging third-party developers to integrate the “Copilot Runtime” in their apps for AI features, and said features will use the NPU in a PC to accelerate AI processing—but, again, only on Copilot+ PCs. For example, the popular DaVinci Resolve video editor can already use the Copilot Runtime to offload some tasks to the NPU.
Let’s assume most developers take Microsoft up on this and integrate the Copilot Runtime for demanding AI tasks. And let’s assume that one day every desktop PC will have a capable NPU. Even then, restricting the Copilot Runtime to only NPUs would still be a mistake. These apps won’t be able to take advantage of the powerful GPUs in desktop gaming PCs and laptop workstations. These apps will be much slower than they could otherwise be when processing AI tasks. (Remember, GPUs hog more power but beat NPUs on raw AI performance.)
Nvidia
Even worse, developers may have to implement multiple AI processing “backends” in their applications—one for NPUs on Copilot+ PCs and one to take advantage of powerful GPUs. That’s more work for developers, which makes Windows less appealing as an app platform. Right now, developers can offload the AI work to cloud servers and ensure it’ll work with their apps on any platform, whether that’s a Chromebook, iPad, Android phone, or Windows PC without an NPU.
If Copilot+ features ran on GPUs, they’d “just work” on a much wider variety of machines, and they’d run much faster on any system with a capable GPU. That would give Microsoft a head start, and that’s key when the PC industry has struggled to make a case for local AI.
Microsoft, it’s time to change course
I understand why Microsoft first launched Copilot+ AI features for NPUs only. It was a great way to push Arm laptops like the Surface Laptop 7 and their Qualcomm Snapdragon X Elite/Plus CPUs.
The company took many months to get those features working on Intel and AMD systems with the required NPUs. It took time to support more hardware, and Microsoft probably hoped that the PC industry would have lots of PCs with NPUs ready by now.
Chris Hoffman / Foundry
But nearly a year later, we still have no options for Copilot+ features on desktop PCs, no Copilot+ features on most laptops, and laptop manufacturers pushing some “AI PCs” and “AI laptops” that can’t run built-in Windows AI features at all—but could, if only Microsoft let them.
That’s why Microsoft should embrace GPUs and let Copilot+ PCs tap into GPU power, too. It’s not just good for users. It’s good for the future of Windows as an AI application platform.
Copilot+ for desktops? One can hope
Unfortunately, Microsoft may never extend Copilot+ features to GPUs. After all, the Copilot+ PC branding is all about selling more PC hardware, and it’s a big part of Microsoft’s current push to get people upgrading from Windows 10. If I were Microsoft, I’d be asking why I should freely give away all these AI features to current PCs.
That stance is understandable, but it’s a shame that PC hardware manufacturers haven’t caught up to Microsoft. Spending $3,000 on a powerful desktop rig with the latest hardware and feeling like it’s already out of date is not a good experience. If only you had bought a lightweight Arm laptop instead! Is that the future of Windows? Maybe. But I’d prefer a future that doesn’t leave behind desktop PCs with powerful hardware.
It doesn’t have to be this way, Microsoft! There’s still time to fix things. Honestly, I hope Microsoft announces Copilot+ features coming to GPUs at Build 2025 in late May. That’d be a win-win, if you ask me.
Further reading: I tried local AI chatbots on my laptop. They kinda suck