“The year of the AI PC” got off to a strange start.
All the “AI PCs” sold by manufacturers for the first half of the year are now effectively out of date. They won’t be able to run Windows Recall, the Windows Copilot Runtime, or all the other AI features Microsoft showed off for its new Copilot+ PCs.
Microsoft’s Copilot+ PC certification just taught us a valuable lesson in buying PC hardware: Never buy hardware based on the promise of what it might be able to do in the future. Only buy PC hardware because of what it can actually do today.
This isn’t the first time we’ve seen PC hardware that promised a glorious future end up not panning out and needing a new generation before really managing to deliver. It’s happened many times before.
Why the first “AI laptops” are already outdated
Intel, Microsoft, and PC makers have been promoting “AI laptops” since the start of 2024. PC makers have been talking a lot about how Intel’s Meteor Lake “Core Ultra” hardware has a built-in neural processing unit (NPU) that can accelerate AI tasks.
These systems certainly do have NPUs, and those NPUs aren’t totally worthless. You can use Windows Studio Effects for “AI-powered” webcam effects and you can hunt down a collection of third-party, mostly open-source utilities that take advantage of the NPU.
While that’s all true, it’s really just the beginning—or not.
If you thought these AI laptops were “future-proof” because of their neural processing units (NPUs), you were in for a rude awakening when Microsoft recently branded them “too slow” for new AI features.
Chris Hoffman
Intel’s NPUs deliver 10 TOPS (trillion operations per second) while AMD’s NPUs deliver up to 16 TOPS. Neither are fast enough to meet Microsoft’s Copilot+ PC baseline standard of at least 40 TOPS.
To be fair, no one promised that these first-generation AI PCs would be able to run future Windows AI features. But I feel like a lot of people in the PC industry wanted us to believe they would be ready for and capable of the AI features coming to Windows.
Or maybe we just wanted to believe.
I recommended not going out of your way to buy an AI PC near the start of 2024, yet even I’m surprised to see them left behind so quickly.
Again, here’s the lesson: You shouldn’t buy PC hardware based on its potential future. Wait until that future arrives, and then you can buy the right hardware to take advantage of it.
Nvidia’s GeForce RTX 20-series GPUs were early, too
When Nvidia announced their GeForce RTX 20-series GPUs back in 2018, they delivered transformative technology. These were the first graphics cards that could provide real-time ray tracing in games—something that was long considered the holy grail of gaming graphics.
They also included the first generation of Nvidia’s Deep Learning Super Sampling (DLSS) technology, which could run games at a lower resolution and then upscale them to a higher resolution, delivering similar graphical quality with much higher performance.
It all sounded great. And honestly, today, it is great! Ray tracing is actually here and functional. If you’re a gamer and you haven’t experienced ray tracing in the Metro series or Cyberpunk 2077, you really should give it a try—because it’s amazing. Nvidia’s DLSS technology is up to DLSS 3 now, and it works very well. Lots of games support these things.
CDProjekt Red
But things weren’t so great back in the first few years after the big launch. Gamers who bought those first-generation RTX 20-series GPUs based on the promise of transformative changes had to wait a while. Windows didn’t even support ray tracing when the GPUs were released, and game developers were slow to implement support for ray tracing and DLSS into their games.
By the time these technologies started becoming reasonably widespread, Nvidia was already launching GeForce RTX 30-series GPUs. Those 20-series GPUs weren’t immediately made obsolete, but if you wanted games that supported these technologies, you could’ve skipped those first-generation GPUs and gone straight to the faster 30-series hardware for an even better experience.
Our Nvidia RTX retrospective from a few years back gets into all the details, and I recommend you read it.
The curse of the early adopter repeats itself
This isn’t the first time “future-proof” hardware has proven anything but. Call it the curse of the early adopter. Don’t believe me? Let’s look back at just a few things that didn’t quite pan out in the PC space:
DirectX 12
DirectX 12 launched in 2015 and AMD talked up how its R7 and R9 300-series GPUs would be compatible, but many games were stuck with DirectX 11 for years.
There was no point in buying AMD’s GPUs to “future proof” your gaming PC. You were much better off waiting until DirectX 12 support became widespread in games and buying a current graphics card then.
3D monitors
Remember 3D monitors? Everyone was pushing those a few years ago. I hope you didn’t buy one based on the promise of an explosion in 3D content, because they didn’t take off and the media and software never really arrived for them.
Windows Mixed Reality
How about Windows Mixed Reality and Microsoft’s plan to run “universal” Windows apps on VR headsets? That didn’t pan out. In fact, Microsoft just removed Windows Mixed Reality from Windows.
Want Windows apps in VR headsets? This week, Microsoft started talking about how it’s now working with Meta on “Windows Volumetric Apps” for Meta Quest headsets.
You’ll want a Meta Quest headset for that sort of thing—not an out-of-date Windows Mixed Reality headset—but I wouldn’t buy a headset for this purpose until the software materializes!
Be an early adopter at your own risk
It happened before, it’ll happen again. Not everything pans out—and even when a technology does take off, the first-generation product may not be powerful enough to participate (which is what we’re now seeing with first-generation AI laptops and Microsoft Copilot+).
Now, if you’re all about bleeding-edge technology, I get it. So am I! I need to play with this stuff so I can share my hands-on experiences here on PCWorld and with the readers of my free Windows Intelligence newsletter. But you have to know what you’re getting into.
If you only care about the best and most reliable experiences, you’re better off avoiding the bleeding edge. Let other people get cut while the new products get polished.
Then, when it makes sense to upgrade, when there’s widespread support for the new hardware, you can swoop in and get an even more refined product—often for less money than you would’ve spent on that first-generation hardware. Score.
