Wednesday, February 18th 2026

Intel Predicts Half of PCs Shipped This Year Will Be AI-Enabled

Intel's Japan president, Makoto Ohno, made interesting remarks that this year will be a make-or-break year for AI PCs. He also mentioned that by 2026, AI PCs will account for roughly half of the unit shipments for the year. This means that out of the total unit shipments, which IDC estimates at around 260 million, about 130 million are expected to be AI PCs. This loosely defined term means that the PC comes with an NPU or other AI processing unit, allowing the system to perform some local AI processing with enough power to meet basic user needs. However, Makoto Ohno also admits that the primary reason for purchasing an AI PC may not be its AI capabilities, but rather the better battery life due to newer and more optimized silicon.
Intel Makoto OhnoIt is also predicted that by 2026, this will exceed half, with one in two computers being an AI PC. However, given the current situation, reasons for choosing an AI PC include its high performance and the long battery life provided by the use of an NPU. In other words, it is important to reflect on the fact that people are not currently purchasing an AI PC in order to use its AI-related functions.
Additionally, the president of Intel Japan stated that the company wants to move quickly to make AI PCs the norm rather than the exception. Currently, the company acknowledges that these machines are largely seen as high-end products, and it aims to change that perception as quickly as possible. Intel also highlighted the need for more applications that truly leverage the capabilities of AI PCs, with the goal of reaching a point where people purchase an AI PC for a specific purpose, not just because it is the latest product.
Source: ITMedia Japan
Add your own comment

29 Comments on Intel Predicts Half of PCs Shipped This Year Will Be AI-Enabled

#1
Wirko
If Intel immediately starts producing DRAM and NAND again, at high volume and for PCs exclusively, then I don't see how he could be wrong.
Posted on Reply
#2
user556
Half of 5% of 2024 sales is still half.
Posted on Reply
#3
tommo1982
I still don't understand the benefit of having AI hardware. To modify pictures, audio? I don't use Gemini or other assistant, why would I need AI in my hardware then.
Posted on Reply
#4
ThomasK
AleksandarKThis loosely defined term means that the PC comes with an NPU or other AI processing unit, allowing the system to perform some local AI processing with enough power to meet basic user needs
Very loosely, really. Marketing loves to call it "AI enabled", but fails to mention a single workload that uses said NPU for local tasks. It all runs on the cloud, so you might as well call any device with access to a browser "AI enabled".
Posted on Reply
#5
mb194dc
What the hell does that even mean.... "AI" enabled. That they'll cost twice as much as last year due to the LLM investment bubble sucking up all the RAM and storage ?
Posted on Reply
#6
WonkoTheSaneUK
And most of them will be AI-disabled as soon as the new owners receive them.
Posted on Reply
#7
Hakker
kind of a moot argument as all top CPU's on Intel has some form of NPU on board. AI as useful as search on Bing.
Posted on Reply
#8
Bwaze
tommo1982I still don't understand the benefit of having AI hardware. To modify pictures, audio? I don't use Gemini or other assistant, why would I need AI in my hardware then.
All of this works on servers. There are very specific user cases where people drive LLM, AI locally, and most of them require much more "neural processing" capability than these feeble attempts marketing labels as "AI PCs".

This marketing mainly preys on people who think they'll need NPU for using ChatGPT.
Posted on Reply
#9
GenericUsername2001
tommo1982I still don't understand the benefit of having AI hardware. To modify pictures, audio? I don't use Gemini or other assistant, why would I need AI in my hardware then.
There are some minor stuff that is mildly useful; an NPU can efficiently do stuff like blur backgrounds and do noise reduction during video calls, and running some image & video editing filters. The efficiency bump is handy if you have to do that stuff away from a power outlet, I guess. That said, the NPUs going into current CPUs are massively overkill for those basic tasks; a NPU 1/5th the die size would be fine. And on the flip side, said NPU units are too weak for other uses of local AI, like local image generation or local chatbots. People who want to do that sort of thing get discrete graphics cards, the high end Macs with lots of shared memory, Strix Halo boxes, or some of those Nvida Sparks machines. Personally I think the average computer user would be better off if NPU units were chopped down, and the saved die space just used to make the iGPU bigger. GPUs, both discrete & integrated, can run all the same AI math you would run on NPU quite well, just with a bit more power usage for the same performance. So any actual useful AI stuff (like some media editing stuff) can just run on the iGPU instead; I mean seriously, who is doing picture editing for hours while not being near a power outlet?
Posted on Reply
#10
DavidC1
So their definition of getting to 50% marketshare is forcing "AI" enabled devices to every device regardless of whether they want it or not, and claim that it's a success, when you have no choice anyway.

One thing Intel might want to do if they want real success is if they fire all these people(or at least severely discipline) that's pulling absolute nonsense like this. AI is 2026 version of DEI.
Posted on Reply
#11
windwhirl
tommo1982I still don't understand the benefit of having AI hardware. To modify pictures, audio? I don't use Gemini or other assistant, why would I need AI in my hardware then.
Probably for the "Copilot+" stuff in Windows. That said, last time I went over what it offers, most of that felt ridiculously useless or unnecessary, with only live translation being something that could be called useful. And lastly, Recall earned itself a special place by being not only useless but also a privacy/security risk.
Posted on Reply
#12
ZoneDymo
Hakkerkind of a moot argument as all top CPU's on Intel has some form of NPU on board. AI as useful as search on Bing.
This, but I see it too many times, those that fully control it predict something....

Like Louis Rossman said he needed to use AI to rework his website in such a way that google shows it again, otherwise he was burried waaaayyy in the back.
Then Google can say: "see?!?! AI is super useful!".....Well yeah because you guys OWN google and made it so that AI is needed for it.

Here intel saying X amount will be AI capable...well yeah because you (forcibly) shove AI crap in your products, if we could buy a cheaper part because it would not have an NPU I can guaranthee that would sell a lot better ><
Posted on Reply
#13
chrcoluk
tommo1982I still don't understand the benefit of having AI hardware. To modify pictures, audio? I don't use Gemini or other assistant, why would I need AI in my hardware then.
A useless invention can still be considered a success if consumers are convinced its worth buying, "AI" right now is some serious mental gymnastics on its usefulness claims.
Posted on Reply
#14
Dirt Chip
Just another PR bullshit to keep inflate the bubble a little more.
Those NPU are good for AI workload even less than the onboard iGPU of CPU to handle games.
It is physically possible but practically irrelevant.
Nothing more than hot air balloon that stinks as it gets higher and higher.
Posted on Reply
#16
Hecate91
When most of their newer CPU's include an NPU, of course Intel can sell "AI enabled" PC's.
All of this pointless AI marketing, and the use cases are limited. IMO, I'd rather have a few extra cores or a larger iGPU than an NPU.
Posted on Reply
#17
Dirt Chip
WirkoIf Intel immediately starts producing DRAM and NAND again, at high volume and for PCs exclusively, then I don't see how he could be wrong.
This will be awesome and actually requested if we are to conquer the last corner that AI still didn't corrupt - CPU price.

If Intel is to repousse cpu manufacture lines to make more dram and nand than we will be officially royaly screwed.
Posted on Reply
#18
windwhirl
RedBlackAkaI will never buy a CPU with an NPU...
All of them are gonna have a NPU so have fun never upgrading I guess.
Posted on Reply
#19
Dirt Chip
RedBlackAkaI will never buy a CPU with an NPU...
"640K ought to be enough for anyone"
Posted on Reply
#20
Dr. Dro
Dirt Chip"640K ought to be enough for anyone"
Don't think that applies here, exactly. Short of LLMs, what is the use for an NPU? They aren't powerful enough to do anything else right now, and potentially, ever.
Posted on Reply
#21
Dirt Chip
Dr. DroDon't think that applies here, exactly. Short of LLMs, what is the use for an NPU? They aren't powerful enough to do anything else right now, and potentially, ever.
You will have NPU in your CPU if you want it or not. You will probably not use it as it is really too weak but still, NPU inside as a PR bullshit will live on.

But I get your point and agree, it's not a 1:1 compare :toast:
Posted on Reply
#22
tommo1982
GenericUsername2001There are some minor stuff that is mildly useful; an NPU can efficiently do stuff like blur backgrounds and do noise reduction during video calls, and running some image & video editing filters. The efficiency bump is handy if you have to do that stuff away from a power outlet, I guess. That said, the NPUs going into current CPUs are massively overkill for those basic tasks; a NPU 1/5th the die size would be fine. And on the flip side, said NPU units are too weak for other uses of local AI, like local image generation or local chatbots. People who want to do that sort of thing get discrete graphics cards, the high end Macs with lots of shared memory, Strix Halo boxes, or some of those Nvida Sparks machines. Personally I think the average computer user would be better off if NPU units were chopped down, and the saved die space just used to make the iGPU bigger. GPUs, both discrete & integrated, can run all the same AI math you would run on NPU quite well, just with a bit more power usage for the same performance. So any actual useful AI stuff (like some media editing stuff) can just run on the iGPU instead; I mean seriously, who is doing picture editing for hours while not being near a power outlet?
I thought one of the use cases would be, the NPU used to help game developers with AI enhanced NPC's. Unfortunately, it seems dev's are not interested in that at the moment. Perhaps it's more of a challenge or we're not there with the current state of LLM's.
Dirt ChipYou will have NPU in your CPU if you want it or not. You will probably not use it as it is really too weak but still, NPU inside as a PR bullshit will live on.

But I get your point and agree, it's not a 1:1 compare :toast:
It's true and sad. I can think of one case where the NPU is used and it's definitely not for the user benefit.
Posted on Reply
#23
Mack4285
They still intend to ship consumer PCs? Prices are going up due to AI, so not sure what they are expecting.
Posted on Reply
#24
Dirt Chip
tommo1982I thought one of the use cases would be, the NPU used to help game developers with AI enhanced NPC's. Unfortunately, it seems dev's are not interested in that at the moment. Perhaps it's more of a challenge or we're not there with the current state of LLM's.


It's true and sad. I can think of one case where the NPU is used and it's definitely not for the user benefit.
The NPU is there to facilitate more AI demand just as a buzz word agent ,which in return make the (basically true, practical irrelevant for NPU capabilities ) need of more NPU/ token power.

A very short, closed loop, additive interactions alas without proper any foundation.
Posted on Reply
#25
AntiX3DGlazer
WirkoIf Intel immediately starts producing DRAM and NAND again, at high volume and for PCs exclusively, then I don't see how he could be wrong.
I would like whatever you are smoking, it seems quite potent.
Dr. DroDon't think that applies here, exactly. Short of LLMs, what is the use for an NPU? They aren't powerful enough to do anything else right now, and potentially, ever.
They're actually really good at low-power CNN networks (computer vision). Think of person detection on a security camera running locally on the NAS box that stores all the recorded video footage. Not a huge use case, but NPUs aren't entirely useless. Should they be everywhere? No, I don't think so. But they do have their uses.
Posted on Reply
Add your own comment
Nov 17th, 2026 14:39 CST change timezone

New Forum Posts

Popular Reviews

TPU on YouTube

Controversial News Posts