People are buying AI PCs, but not for their AI abilities

[midian182]

In brief: Makers of AI PCs and processors will be happy with the news that people are buying these products, which feature NPUs designed to accelerate AI tasks. The less welcome caveat is that according to a new report, many people are purchasing them simply because they want or need a new PC – not because of the AI features.

The latest IDC Research report shows that shipments of personal computing devices are expected to grow 2.6% year over year in 2024 to 398.9 million units.

The report notes how pretty much every PC company these days pushes the benefits of their machines' NPUs and AI smarts, but most people buy them because they want a new or upgraded PC, as opposed to specifically requiring an AI PC.

"While AI has been a buzzword of late, it has yet to be a purchase driver among PC buyers," said Jitesh Ubrani, research manager, Worldwide Mobile Device Trackers. "Businesses certainly recognize the importance of AI though many struggle to see the immediate use case and instead are opting for AI PCs as a means to futureproofing."

Ubrani adds that manufacturers should explain better to consumers what an NPU is, why it's important, and why they might need one. The fact that all new processors have built-in NPUs means AI PC sales will continue to grow, even if people don't care about their headline capabilities.

Ubrani also points out that the inclusion of an NPU and the early hardware requirements for AI tasks mean premium laptops are coming with more storage and memory, which is increasing the average selling price.

In July, a survey of 22,000 people found that 84% were not willing to pay extra for a PC or hardware just because it offered enhanced AI capabilities.

Looking at PCs overall (excluding tablets), IDC writes that the traditional PC market will remain flat in 2024 with 261 million units shipped as growth stagnates at just 0.3%. That figure would be 2.8% when excluding China, which is experiencing several economic challenges that are impacting the PC market. Next year should be better, though, as the end of Windows 10 support is expected to spur commercial buyers to upgrade, pushing growth to 4.3%.

The tablet industry is a bright spot in the report. Slate sales are up 7.2% compared to last year as shipments reach 137.9 million, as manufacturers release more powerful models with larger screens. We're used to seeing Apple dominate this area, but growth has been driven by Android tablets in emerging regions. Unfortunately for OEMs, tablet growth is expected to hit 0% by 2028 as smartphones become larger and more powerful (and more foldables arrive) and tablets struggle in the face of touch-capable laptops.

Permalink to story:

People are buying AI PCs, but not for their AI abilities

 

[NumberNine]

If they don't catch on then I'm fairly confident Microsoft will pivot their strategy towards fabricating new system requirements to require NPU for all consumer editions of Windows.

That seems appropriate for Microsoft's business sensibilities. The soft move is making it the more affordable option, but it won't be a complete Microsoft strategy until it's the only option.

 

[Anton Longshot]

Lol "AI capabilities"? What the h*** for? I generate AI pictures but need not want to be able to do so locally - Remote servers render super fast and in excellent quality.
I've tried to come up with a different reason for local AI but came up empty.

Hardware companies are desperately trying to convince people that they need/want to BUY local AI stuff but seem to fail. I have many friends that work in IT and/or are nerds and NONE of them is interested in owning AI hardware.
Personally I don't even want it on my phone and prefer raw photos to AI-beautified/perfected shots.
What else can I say? A perfect AI PC would run ALL my tasks without my interference, and if that would be possible I'd let that happen elsewhere so I can read books and watch movies instead of computing for fun hahaha. It's one of those solutions looking for a problem if you ask me.[\edit-word]

 

[yRaz]

I still don't understand what an AI PC is. It has an NPU, but for what purpose?

 

[Dr Roboto]

Here is what ChatGPT has to say. It took several prompts to get it to describe what NPUs are useful for today and not "theoritically" useful for. Some of the things in the list seem helpful, but I would say that I don't have use for most of them. So, there is value, but I personally don't think it lives up to the hype and would be compelling to your average person to upgrade just for these features.

1. Facial Recognition Login (Windows Hello)
Windows Hello is widely available on laptops and desktops with AI accelerators or NPUs. This feature enables quick, secure login using facial recognition, and it's already optimized for faster performance thanks to local AI processing.

2. Real-Time Noise Cancellation in Calls (Zoom, Microsoft Teams, Google Meet)
Tools like NVIDIA Broadcast (which uses AI to cancel out background noise in real time) are widely available and rely on AI acceleration. In addition, Microsoft Teams and Zoom now offer built-in noise suppression features that are supported by AI-enhanced NPUs.

3. On-Device Voice Dictation (Microsoft Word, Google Docs)
Windows 11’s Voice Typing feature lets you dictate text directly into documents or emails using on-device AI, powered by NPUs. This works without needing an internet connection and provides real-time transcription.

4. Faster Photo Search and Organization (Google Photos, Apple Photos)
Google Photos and Apple Photos both use AI to sort, tag, and allow you to search your photos by people, locations, and objects. These services use on-device AI acceleration on devices with NPUs, such as the Apple M1/M2 chips and certain Windows laptops.

5. Real-Time AI-Enhanced Camera Features (HP, Dell, Lenovo, Apple)
Laptops like HP’s Dragonfly G3 and Dell’s Latitude 9420 use AI-powered webcams for auto-framing, auto-background blur, and lighting enhancements during video calls. Apple’s MacBooks with M1/M2 chips also have AI-powered image processing to improve the webcam’s quality.

6. Smarter Email and Document Suggestions (Microsoft 365, Gmail)
Gmail’s Smart Compose and Microsoft Word/Outlook provide AI-powered text predictions and suggestions, saving you time when drafting emails or documents. These services work faster and more efficiently with NPUs.

7. Real-Time Malware Detection (Windows Defender)
Windows Defender uses AI-powered threat detection, and with NPUs, the system can run these AI models locally for real-time threat analysis without relying on cloud-based services.

8. Battery Life Optimization (Intel Evo, ARM-based Laptops)
Intel Evo-certified laptops and devices with Apple’s M1/M2 chips use AI to manage power consumption efficiently. AI accelerators dynamically adjust system performance and power usage to optimize battery life for tasks like web browsing, video playback, and light productivity.

9. Automatic Background Removal for Calls (Zoom, Teams, Google Meet)
Features like background blurring or replacement during video calls are already available in Zoom, Microsoft Teams, and Google Meet. These can be accelerated by NPUs for smoother performance, even on lower-end systems.

10. Voice Assistant Integration (Windows 11 Cortana, macOS Siri)
AI-powered voice assistants like Cortana and Siri use on-device AI to process voice commands and offer faster, more accurate results. AI accelerators help these assistants run more efficiently and respond quickly without requiring cloud processing.

 

[deathspeed]

I have been interested in tech my whole life, and building my own PCs for over 20 years, and I actively avoid any hardware "AI" features. I certainly will not pay more for them.

 

[ioliveirafarias]

Here is what ChatGPT has to say. It took several prompts to get it to describe what NPUs are useful for today and not "theoritically" useful for. Some of the things in the list seem helpful, but I would say that I don't have use for most of them. So, there is value, but I personally don't think it lives up to the hype and would be compelling to your average person to upgrade just for these features.

1. Facial Recognition Login (Windows Hello)
Windows Hello is widely available on laptops and desktops with AI accelerators or NPUs. This feature enables quick, secure login using facial recognition, and it's already optimized for faster performance thanks to local AI processing.

2. Real-Time Noise Cancellation in Calls (Zoom, Microsoft Teams, Google Meet)
Tools like NVIDIA Broadcast (which uses AI to cancel out background noise in real time) are widely available and rely on AI acceleration. In addition, Microsoft Teams and Zoom now offer built-in noise suppression features that are supported by AI-enhanced NPUs.

3. On-Device Voice Dictation (Microsoft Word, Google Docs)
Windows 11’s Voice Typing feature lets you dictate text directly into documents or emails using on-device AI, powered by NPUs. This works without needing an internet connection and provides real-time transcription.

4. Faster Photo Search and Organization (Google Photos, Apple Photos)
Google Photos and Apple Photos both use AI to sort, tag, and allow you to search your photos by people, locations, and objects. These services use on-device AI acceleration on devices with NPUs, such as the Apple M1/M2 chips and certain Windows laptops.

5. Real-Time AI-Enhanced Camera Features (HP, Dell, Lenovo, Apple)
Laptops like HP’s Dragonfly G3 and Dell’s Latitude 9420 use AI-powered webcams for auto-framing, auto-background blur, and lighting enhancements during video calls. Apple’s MacBooks with M1/M2 chips also have AI-powered image processing to improve the webcam’s quality.

6. Smarter Email and Document Suggestions (Microsoft 365, Gmail)
Gmail’s Smart Compose and Microsoft Word/Outlook provide AI-powered text predictions and suggestions, saving you time when drafting emails or documents. These services work faster and more efficiently with NPUs.

7. Real-Time Malware Detection (Windows Defender)
Windows Defender uses AI-powered threat detection, and with NPUs, the system can run these AI models locally for real-time threat analysis without relying on cloud-based services.

8. Battery Life Optimization (Intel Evo, ARM-based Laptops)
Intel Evo-certified laptops and devices with Apple’s M1/M2 chips use AI to manage power consumption efficiently. AI accelerators dynamically adjust system performance and power usage to optimize battery life for tasks like web browsing, video playback, and light productivity.

9. Automatic Background Removal for Calls (Zoom, Teams, Google Meet)
Features like background blurring or replacement during video calls are already available in Zoom, Microsoft Teams, and Google Meet. These can be accelerated by NPUs for smoother performance, even on lower-end systems.

10. Voice Assistant Integration (Windows 11 Cortana, macOS Siri)
AI-powered voice assistants like Cortana and Siri use on-device AI to process voice commands and offer faster, more accurate results. AI accelerators help these assistants run more efficiently and respond quickly without requiring cloud processing.

Seem's like most points already existed before the hype and a few people was urging for Npus.

 

[AlexIL]

I actually stay away from any "AI" marked PC. Technology is in alpha test stage and I am not interested in it anyway. Hype over it screams Sam Bankman-Fried style burning of billions. Considering there's another Sam in the game.

 

[Fastturtle]

Here is what ChatGPT has to say. It took several prompts to get it to describe what NPUs are useful for today and not "theoritically" useful for. Some of the things in the list seem helpful, but I would say that I don't have use for most of them. So, there is value, but I personally don't think it lives up to the hype and would be compelling to your average person to upgrade just for these features.

1. Facial Recognition Login (Windows Hello)
Windows Hello is widely available on laptops and desktops with AI accelerators or NPUs. This feature enables quick, secure login using facial recognition, and it's already optimized for faster performance thanks to local AI processing.

2. Real-Time Noise Cancellation in Calls (Zoom, Microsoft Teams, Google Meet)
Tools like NVIDIA Broadcast (which uses AI to cancel out background noise in real time) are widely available and rely on AI acceleration. In addition, Microsoft Teams and Zoom now offer built-in noise suppression features that are supported by AI-enhanced NPUs.

3. On-Device Voice Dictation (Microsoft Word, Google Docs)
Windows 11’s Voice Typing feature lets you dictate text directly into documents or emails using on-device AI, powered by NPUs. This works without needing an internet connection and provides real-time transcription.

4. Faster Photo Search and Organization (Google Photos, Apple Photos)
Google Photos and Apple Photos both use AI to sort, tag, and allow you to search your photos by people, locations, and objects. These services use on-device AI acceleration on devices with NPUs, such as the Apple M1/M2 chips and certain Windows laptops.

5. Real-Time AI-Enhanced Camera Features (HP, Dell, Lenovo, Apple)
Laptops like HP’s Dragonfly G3 and Dell’s Latitude 9420 use AI-powered webcams for auto-framing, auto-background blur, and lighting enhancements during video calls. Apple’s MacBooks with M1/M2 chips also have AI-powered image processing to improve the webcam’s quality.

6. Smarter Email and Document Suggestions (Microsoft 365, Gmail)
Gmail’s Smart Compose and Microsoft Word/Outlook provide AI-powered text predictions and suggestions, saving you time when drafting emails or documents. These services work faster and more efficiently with NPUs.

7. Real-Time Malware Detection (Windows Defender)
Windows Defender uses AI-powered threat detection, and with NPUs, the system can run these AI models locally for real-time threat analysis without relying on cloud-based services.

8. Battery Life Optimization (Intel Evo, ARM-based Laptops)
Intel Evo-certified laptops and devices with Apple’s M1/M2 chips use AI to manage power consumption efficiently. AI accelerators dynamically adjust system performance and power usage to optimize battery life for tasks like web browsing, video playback, and light productivity.

9. Automatic Background Removal for Calls (Zoom, Teams, Google Meet)
Features like background blurring or replacement during video calls are already available in Zoom, Microsoft Teams, and Google Meet. These can be accelerated by NPUs for smoother performance, even on lower-end systems.

10. Voice Assistant Integration (Windows 11 Cortana, macOS Siri)
AI-powered voice assistants like Cortana and Siri use on-device AI to process voice commands and offer faster, more accurate results. AI accelerators help these assistants run more efficiently and respond quickly without requiring cloud processing.

#3 we already have. All that an NPU will do is speed things up a bit. I've used the Win7/10 Voice to Text function and after training, it works pretty well
Here's what was missed that I feel is more important\
Improved Encryption Speeds for SSL/Drive Encryption - Possible to abuse this though with Ransomware.

As to Faster Image/Doc Search and such, I don't have any problem because I dont use the iTunes filing method. All of my stuff is organized. Email suggestions - why in hell would I even allow it to autocorrect my emails? LibGen becomes LinkedIn using that. Epp off on that. I'm already having issues with people that can't hold a conversation, track what's been said when it's right in front of them in text and too many others can't even link two damn thoughts together as Megacorps work to build the best ***** they can

 

[ActualIndians]

For all those people that don't understand what AI is for locally imagine future games require these processors to generate in-game NPCs that can chat with you like a normal person. Imagine in-game AI that can be tuned to make it more challenging and not cookie cutter responses. Imagine the people that like to keep their data local and don't want to keep exporting it to the cloud and like their Privacy. Imagine some people would like to train this on not suitable for work content. From my point of view it is all about freedom to use your machine in ways that big Tech will not allow.

 

[zamroni111]

I still don't understand what an AI PC is. It has an NPU, but for what purpose?

it's like gpu but specialized for mathematics operations of AI software.

ai software can run on gpu but will be much more power efficient on npu.
game 3d rendering is math 32 bit oriented so gpu is designed for that.
meanwhile ai is 16 bit oriented and doesnt need 3d rendering logic, ray tracing etc.

if using gpu, the cancelled microsoft ai recall feature might drain laptop battery too fast.

 

[zamroni111]

Lol "AI capabilities"? What the h*** for? I generate AI pictures but need not want to be able to do so locally - Remote servers render super fast and in excellent quality.
I've tried to come up with a different reason for local AI but came up empty.

Hardware companies are desperately trying to convince people that they need/want to BUY local AI stuff but seem to fail. I have many friends that work in IT and/or are nerds and NONE of them is interested in owning AI hardware.
Personally I don't even want it on my phone and prefer raw photos to AI-beautified/perfected shots.
What else can I say? A perfect AI PC would run ALL my tasks without my interference, and if that would be possible I'd let that happen elsewhere so I can read books and watch movies instead of computing for fun hahaha. It's one of those solutions looking for a problem if you ask me.[\edit-word]

it was because of the stupid microsoft recall feature.
it will use to much battery if it runs on gpu.

consumers indeed must bear "npu tax" due to this overkill npu.
in lunar lake, the npu circuit is almost as big as 4 P cores so surely it increases the cost of chip.

 

[yRaz]

it's like gpu but specialized for mathematics operations of AI software.

ai software can run on gpu but will be much more power efficient on npu.
game 3d rendering is math 32 bit oriented so gpu is designed for that.
meanwhile ai is 16 bit oriented and doesnt need 3d rendering logic, ray tracing etc.

if using gpu, the cancelled microsoft ai recall feature might drain laptop battery too fast.

I know what an NPU is, I'm asking what function does it serve. So it's more energy efficient when processing AI workloads than a GPU. So I ask again, what AI workloads am I going to be processing?

 

[zamroni111]

I know what an NPU is, I'm asking what function does it serve. So it's more energy efficient when processing AI workloads than a GPU. So I ask again, what AI workloads am I going to be processing?

like I mentioned before, ai software such as microsoft recall, llm, etc. are math 16 bit oriented.
they will use much less power in 16 bit math hardware like npu than 32 bit gpu.

bfloat16 floating-point format - Wikipedia

en.wikipedia.org

 

[yRaz]

like I mentioned before, ai software such as microsoft recall, llm, etc. are math 16 bit oriented.
they will use much less power in 16 bit math hardware like npu than 32 bit gpu.

bfloat16 floating-point format - Wikipedia

en.wikipedia.org

I will be doing none of that stuff and I think most people won't either. So, I will keep rewording this as best that I can. Why do I need one? What value will an AI PC provide to me? I do a lot of work with AI and, ironically, most of my job is filtering out AI nonsense submitted by people too lazy to do something theirs elf.

So what does an AI PC do aside from things that I specifically don't want a computer doing, such as MS Recall.

 

[ioliveirafarias]

For all those people that don't understand what AI is for locally imagine future games require these processors to generate in-game NPCs that can chat with you like a normal person. Imagine in-game AI that can be tuned to make it more challenging and not cookie cutter responses. Imagine the people that like to keep their data local and don't want to keep exporting it to the cloud and like their Privacy. Imagine some people would like to train this on not suitable for work content. From my point of view it is all about freedom to use your machine in ways that big Tech will not allow.

Ok, please gimme a list of 3 games with this feature that I can play today.

 

[ioliveirafarias]

I know what an NPU is, I'm asking what function does it serve. So it's more energy efficient when processing AI workloads than a GPU. So I ask again, what AI workloads am I going to be processing?

By the hype, I'm thinking that maybe Windows will do background data spying using AI for even more personalized ADS.

 

[ActualIndians]

Ok, please gimme a list of 3 games with this feature that I can play today.

Research Unity Muse. Games created with this platform will have these capabilities.

 

[ioliveirafarias]

Research Unity Muse. Games created with this platform will have these capabilities.

I asked for a list, not a premonition.

 

[mbrowne5061]

Here is what ChatGPT has to say. It took several prompts to get it to describe what NPUs are useful for today and not "theoritically" useful for. Some of the things in the list seem helpful, but I would say that I don't have use for most of them. So, there is value, but I personally don't think it lives up to the hype and would be compelling to your average person to upgrade just for these features.

1. Facial Recognition Login (Windows Hello)
Windows Hello is widely available on laptops and desktops with AI accelerators or NPUs. This feature enables quick, secure login using facial recognition, and it's already optimized for faster performance thanks to local AI processing.

2. Real-Time Noise Cancellation in Calls (Zoom, Microsoft Teams, Google Meet)
Tools like NVIDIA Broadcast (which uses AI to cancel out background noise in real time) are widely available and rely on AI acceleration. In addition, Microsoft Teams and Zoom now offer built-in noise suppression features that are supported by AI-enhanced NPUs.

3. On-Device Voice Dictation (Microsoft Word, Google Docs)
Windows 11’s Voice Typing feature lets you dictate text directly into documents or emails using on-device AI, powered by NPUs. This works without needing an internet connection and provides real-time transcription.

4. Faster Photo Search and Organization (Google Photos, Apple Photos)
Google Photos and Apple Photos both use AI to sort, tag, and allow you to search your photos by people, locations, and objects. These services use on-device AI acceleration on devices with NPUs, such as the Apple M1/M2 chips and certain Windows laptops.

5. Real-Time AI-Enhanced Camera Features (HP, Dell, Lenovo, Apple)
Laptops like HP’s Dragonfly G3 and Dell’s Latitude 9420 use AI-powered webcams for auto-framing, auto-background blur, and lighting enhancements during video calls. Apple’s MacBooks with M1/M2 chips also have AI-powered image processing to improve the webcam’s quality.

6. Smarter Email and Document Suggestions (Microsoft 365, Gmail)
Gmail’s Smart Compose and Microsoft Word/Outlook provide AI-powered text predictions and suggestions, saving you time when drafting emails or documents. These services work faster and more efficiently with NPUs.

7. Real-Time Malware Detection (Windows Defender)
Windows Defender uses AI-powered threat detection, and with NPUs, the system can run these AI models locally for real-time threat analysis without relying on cloud-based services.

8. Battery Life Optimization (Intel Evo, ARM-based Laptops)
Intel Evo-certified laptops and devices with Apple’s M1/M2 chips use AI to manage power consumption efficiently. AI accelerators dynamically adjust system performance and power usage to optimize battery life for tasks like web browsing, video playback, and light productivity.

9. Automatic Background Removal for Calls (Zoom, Teams, Google Meet)
Features like background blurring or replacement during video calls are already available in Zoom, Microsoft Teams, and Google Meet. These can be accelerated by NPUs for smoother performance, even on lower-end systems.

10. Voice Assistant Integration (Windows 11 Cortana, macOS Siri)
AI-powered voice assistants like Cortana and Siri use on-device AI to process voice commands and offer faster, more accurate results. AI accelerators help these assistants run more efficiently and respond quickly without requiring cloud processing.

Except for maybe #4, none of these require any kind of neural net to be performed, though.

 

[LimyG]

Hopefully there will be a kind of decrappifier app freely available so people who don't want this can easily de-activate it.
Fed up with these faceless corporations telling people what they want.