Incoming Wave Of AI Is Making Buying PCs Riskier For Businesses

AI PCs due to land this year could increase businesses' risk of buying the wrong tech as there is still no current AI standard for software to work with and confusion remains over what makes up an AI PC.

A slew of new systems were promised at this week's mega mobile tech show MWC, but not all of these were necessarily AI PCs.

This is because Intel's definition of an AI PC is one with a CPU, GPU and an NPU (neural processing unit), and currently only the Core Ultra mobile chips fit this description. The 14th Gen Core desktop chips Intel unveiled last October don't have an NPU, for example.

The Santa Clara chipmaker made public an updated vPro platform for business systems at MWC, but the mobile version is Core Ultra, while the desktop chips are 14th Gen Core. Buy vPro laptops this year and they will be AI PCs; buy vPro desktops and they won't.

This was enough to fool even some tech news sites, which took Intel's declaration that more than 100 new vPro designs will be delivered from system builders this year to mean that more than 100 AI PCs were on the way, so buyers can be forgiven for being confused.

The risk of buying the wrong technology, however, has increased for several reasons, according to Directions on Microsoft analyst Michael Cherry.

It started with Windows 11, he says, when Microsoft changed the hardware requirements significantly so it could enforce a level of security, making a Trusted Platform Module (TPM) mandatory, for example.

"These changed requirements created risk by suddenly obsoleting many devices and introducing the possibility of buying a new or replacement device that was still capable of running Windows 10 but was incapable of running Windows 11," Cherry notes in a post.

Windows itself does not currently make use of AI PC hardware, although this is said to be coming this year. Cherry claims that for customers which might expect to perform some AI training locally on a client device, it isn't clear what NPU will be needed.

"Microsoft and others are particularly silent on this point, even though lots of machines demonstrated at CES in January are promoting their NPUs. But an NPU from Qualcomm is not the same as one from Intel or AMD or Nvidia," he says.

• Chinese 'connected' cars are a national security threat, says Biden
• Water worries flood in as chip industry and AI models grow thirstier
• Meta's pay-or-consent model hides 'massive illegal data processing ops': lawsuit
• Europe probes Microsoft's €15M stake in AI upstart Mistral

This points to the need for something like DirectX for AI, according to Cherry, referring to the API layer in Windows that standardized 3D graphics and audio across different hardware. A future version of the DirectML machine learning library could possibly provide this.

"Microsoft announced in early February a developer preview for DirectML 1.13.1, which supports Intel AI Boost NPUs. But until it is clear which vendors support APIs such as DirectML, or some other standard, there is a risk when buying an expensive NPU-enabled processor-based PC," he said.

On the other hand, there is also a risk in continuing to buy PCs without an NPU if you are looking to use large language models.

"The risk of buying a device that is obsolete before you get it in the door is higher than it has been in a long time," Cherry says.

This hasn't stopped the vendors from looking forward to a sales boost from AI PCs, expected in the second half of the year.

"In the near term, the PC market is still soft and we expect the recovery to push into the second half as enterprise and large customers remain cautious with their spend," Dell chief operating officer Jeff Clarke said on the company's latest earnings call.

He said the company is optimistic about the coming refresh cycle, because the PC installed base continues to age, Windows 10 is reaching the end of life later next year, and the industry is working to make advances in AI-enabled architectures and software applications.

"PCs will become even more essential as most day-to-day work with AI will be done on the PC," he added.

HP indicated there will be some benefit to its bottom line in 2024 from higher margin AI PCs, but is forecasting bigger and better things for 2025 and 2026.

"We remain extremely excited about the opportunity that AI PCs will bring in terms of both the customer value that they will deliver in terms of security, in terms of latency, in terms of cost and also the impact it will have over time in the company," HP CEO Enrique Lores said during its earnings call this week.

Lores said there were still some milestones to reach the AI PC goal, however. One is to deliver hardware to support AI processing, second is to ensure that applications support that hardware, and the third will be customer training.

"Our projections continue to be that three years after launch, the penetration of AI PCs will be somewhere between 40 percent and 60 percent of the total sales that we will be making. And that growth is going to be gradual," he said. ®