Why can't my $3000 gaming PC run Copilot+ features? Come on, Microsoft!

The twin that means of “AI PC” in 2025

The time period “AI PC” has gotten fairly complicated. In current months, I’ve reviewed a number of laptops categorized as “AI PCs” or “AI laptops,” however that has meant various things from completely different individuals:

• Some so-called AI PCs are light-weight laptops with NPUs however no GPUs. These are correct Copilot+ laptops with entry to Copilot+ options and maybe even preinstalled AI functions that hook up with OpenAI’s GPT fashions or different cloud-based AI options.
• Other so-called AI PCs are heavier, pricier, performance-based gaming laptops with out NPUs however with highly effective GPUs. These are high-end laptops that may use their graphics processor for native AI functions, but can’t run Copilot+ AI options.

These two phrases are mainly the opposites of one another—fully completely different {hardware} and fully completely different options. But producers are likely to name each “AI laptops,” which muddies the waters.

The PC business dubbed 2024 “the year of the AI PC,” and but those first-wave AI PCs were left behind by Copilot+ PCs. What’s worse, many new PCs being bought right this moment in 2025 are nonetheless being left behind.

The resolution right here is straightforward: Microsoft ought to let PCs run Copilot+ AI options on GPUs. Then all the so-called AI PCs would be capable of run AI options, not simply those geared up with cutting-edge NPUs.

GPUs can deal with it, so let ’em rip

Here’s the query you need to ask your self in the event you’re curious about a desktop PC proper now: Do you need to wait an unspecified period of time for desktop NPUs to be introduced and launched a while sooner or later, or do you need to get a desktop PC proper now?

As of mid-2025, Copilot+ AI options solely run on laptops with certified NPUs. That means Qualcomm Snapdragon X Elite/Plus laptops, Intel Lunar Lake laptops, and AMD Ryzen AI 300 sequence laptops. Technically, there are additionally just a few light-weight mini PCs with cell CPUs, just like the Asus ROG NUC sequence, that additionally match the invoice. No desktop PCs.

For me, I made a decision that Copilot+ options didn’t matter sufficient to attend. I’ve a Surface Laptop 7 with Insider builds of Windows 11 and I by no means use the Copilot+ options on it. I’ve Recall activated, however I haven’t touched it in months. (Privacy considerations apart, the worst thing about Recall is that it’s simply not very attention-grabbing or helpful.) Other AI options—like producing pictures instantly within the Photos app—feel like tech demos at best.

Still, Microsoft’s characteristic bulletins up to now few days, it’s clear that Microsoft goes full steam forward with Copilot+, and a few of these newer options are getting extra attention-grabbing. Why sure, I would love AI to change PC settings more quickly! And guess what? My PC’s GeForce 5080 GPU may definitely run options like that… if Microsoft would simply let it. (My blazing-fast AMD CPU in all probability may, too.)

When it involves working AI duties, NPUs sit somewhere between CPUs and GPUs—they’re sooner at AI processing than CPUs, however slower than GPUs. NPUs are additionally far more energy environment friendly than GPUs, and that’s a giant deal if you’re dealing with AI duties on a laptop computer. But with a desktop PC that’s at all times plugged in? Battery life doesn’t matter.

And whereas AI options would possibly hog GPU sources, I don’t essentially care if my GPU takes a success for a second once I search the Settings app, for instance. There’s a lot potential to make the most of in a GPU that isn’t doing the rest. Might as properly use it for AI, proper?

It’s not nearly Windows AI options

All of the above apart, this complete downside encompasses a lot extra than simply the Copilot+ AI options constructed into Windows.

Microsoft is encouraging third-party builders to combine the “Copilot Runtime” of their apps for AI options, and stated options will use the NPU in a PC to speed up AI processing—however, once more, solely on Copilot+ PCs. For instance, the favored DaVinci Resolve video editor can already use the Copilot Runtime to dump some duties to the NPU.

Let’s assume most builders take Microsoft up on this and combine the Copilot Runtime for demanding AI duties. And let’s assume that someday each desktop PC could have a succesful NPU. Even then, limiting the Copilot Runtime to solely NPUs would nonetheless be a mistake. These apps received’t be capable of make the most of the highly effective GPUs in desktop gaming PCs and laptop computer workstations. These apps might be a lot slower than they might in any other case be when processing AI duties. (Remember, GPUs hog extra energy however beat NPUs on uncooked AI efficiency.)

Nvidia

Even worse, builders could should implement a number of AI processing “backends” of their functions—one for NPUs on Copilot+ PCs and one to make the most of highly effective GPUs. That’s extra work for builders, which makes Windows much less interesting as an app platform. Right now, builders can offload the AI work to cloud servers and guarantee it’ll work with their apps on any platform, whether or not that’s a Chromebook, iPad, Android cellphone, or Windows PC with out an NPU.

If Copilot+ options ran on GPUs, they’d “just work” on a a lot wider number of machines, and so they’d run a lot sooner on any system with a succesful GPU. That would give Microsoft a head begin, and that’s key when the PC industry has struggled to make a case for local AI.

Microsoft, it’s time to vary course

I perceive why Microsoft first launched Copilot+ AI options for NPUs solely. It was an effective way to push Arm laptops just like the Surface Laptop 7 and their Qualcomm Snapdragon X Elite/Plus CPUs.

The firm took many months to get those features working on Intel and AMD systems with the required NPUs. It took time to assist extra {hardware}, and Microsoft in all probability hoped that the PC business would have a lot of PCs with NPUs prepared by now.

Chris Hoffman / Foundry

But practically a yr later, we nonetheless don’t have any choices for Copilot+ options on desktop PCs, no Copilot+ options on most laptops, and laptop computer producers pushing some “AI PCs” and “AI laptops” that may’t run built-in Windows AI options in any respect—however may, if solely Microsoft allow them to.

That’s why Microsoft ought to embrace GPUs and let Copilot+ PCs faucet into GPU energy, too. It’s not simply good for customers. It’s good for the way forward for Windows as an AI utility platform.

Copilot+ for desktops? One can hope

Unfortunately, Microsoft could by no means lengthen Copilot+ options to GPUs. After all, the Copilot+ PC branding is all about promoting extra PC {hardware}, and it’s a giant a part of Microsoft’s current push to get people upgrading from Windows 10. If I had been Microsoft, I’d be asking why I ought to freely give away all these AI options to present PCs.

That stance is comprehensible, but it surely’s a disgrace that PC {hardware} producers haven’t caught as much as Microsoft. Spending $3,000 on a strong desktop rig with the newest {hardware} and feeling prefer it’s already old-fashioned is not a very good expertise. If solely you had purchased a light-weight Arm laptop computer as a substitute! Is that the way forward for Windows? Maybe. But I’d desire a future that doesn’t go away behind desktop PCs with highly effective {hardware}.

It doesn’t should be this fashion, Microsoft! There’s nonetheless time to sort things. Honestly, I hope Microsoft pronounces Copilot+ options coming to GPUs at Build 2025 in late May. That’d be a win-win, in the event you ask me.

Further studying: I tried local AI chatbots on my laptop. They kinda suck