AI PCs: do you need one and what to know before buying

A number of new laptops are offering AI capabilities that may benefit some professionals and companies. We examine what to consider before determining if an AI PC is the right buy.

A few weeks ago, some of the biggest laptop manufacturers, such as Microsoft, Dell, HP, and Samsung, started shipping their Copilot + PCs, also called AI PCs.

While the most prominent and controversial one, Recall, was postponed due to security concerns, these laptops have other AI improvements.

Thanks to the latest Qualcomm System on a Chip (SoC), they will enable AI-powered image generation, editing, real-time translations with captions, and more.

Soon, offerings of AI-capable Windows devices will expand as manufacturers start shipping laptops with chips made by companies like Intel and AMD.

For those who prefer Mac, there are models powered by Apple's silicon capable of handling AI workloads.

While AI PCs are currently the new buzzword, they will soon become the new norm.

Canalys predicts that in 2027, 60% of all shipped PCs will be capable of handling AI tasks.

Earlier, we covered the basics of AI PCs. Now that the first models are here, let's take a look at the more practical side: is it a good idea to buy an AI PC ahead of everyone else? What is there to know before buying one? And who would benefit the most from having an AI-capable laptop?

The new architecture

The latest wave of Copilot + PCs coincides with the rise of the new architecture, Arm-based computers. Knowing which devices come with Arm may be useful, as it brings several significant benefits to the hardware.

For many years, Intel's x86 architecture-based computers dominated the laptop market. However, Apple, after ditching Intel and its x86 architecture in 2020, showed that it’s possible to create fast and efficient laptops on Arm architecture, which was typically used for mobile devices.

These Arm-based Macs, also made with Apple's silicon, significantly outperformed those with Windows in terms of speed and battery life.

Microsoft has also tried to run Windows PCs on Arm a few times, but it ran into software compatibility issues, as there were no native apps such as Microsoft Office or Edge.

But this time, things seem to be different. Thanks to the Prism emulator, much more software is now capable of running on Arm.

According to The Verge's Tom Warren, who tested the latest Microsoft Surface devices running on Arm, there are still issues with some apps, like Adobe Premiere Pro and Blender, and some games are unavailable. We can expect them to be solved in the future.

Perhaps the most important thing here is that Surface and other Windows devices made on Arm can finally have performance and battery life comparable to – or even better than – Apple.

Qualcomm and Apple aren't the only companies offering Arm-based laptops. According to a report by Reuters, Mediatek is also planning to launch an SoC on Arm, as well as AMD and Nvidia.

Soon, there will be more AI laptops with chips from Intel and AMD that are powered by x86 architecture. These companies also promise some improvements over the previous generation of devices.

NPUs and TOPS – what are they?

One distinguishing feature of all the new AI PCs is that they have neural processing units (NPUs), also called neural networks, specifically dedicated to handling AI tasks.

They handle computations, allowing the CPU and GPU to concentrate on other tasks. NPU efficiency is measured by how many operations it can perform per second.

For example, Microsoft says that an AI PC should be capable of achieving at least 40 trillion operations per second (TOPS). For comparison, Qualcomm's devices can achieve 45 TOPS, Apple's Macs with the M3 processor 38 TOPS, AMD's upcoming chips will be capable of handling 50 TOPS, and Intel's upcoming Lunar Lake – 48 TOPS.

Generally, the more TOPS, the better. However, other factors must also be considered.

Qualcomm CEO Christiano Amon noted in his keynote speech at Computex last month that people should also pay attention to how much energy these chips use.

What is there to know before buying an AI PC?

All these NPUs are designed to complete all these tasks locally on your device. This provides substantial benefits in processing speed, data privacy, and real-time analytics, says Elliott Jones at Kingston Technology, one of the largest memory and SSD manufacturers in the world.

According to the expert, several professionals would benefit from using a PC with a dedicated NPU.

These include financial analysts needing to run complex models and simulations, healthcare providers processing high-resolution images and diagnostic data, and scientific researchers handling massive datasets.

Before buying an AI PC, he recommends evaluating several factors:

Compatibility with specific AI software tools and frameworks

Hardware capabilities and requirements, such as specific types of processors (GPUs or NPUs)

Memory (RAM) and storage: "Do they match the work you plan to do? Can you upgrade in the future, or is it "soldered" to the motherboard?"

Software dependencies, scalability, performance needs, vendor support, and community resources are also important considerations

Dedicated gaming PCs may be better for gamers

According to Adhip Ray, founder of Digital Marketing Consultancy WinSavvy, the first step in deciding whether an AI PC is necessary is to decide what AI tasks will be run on the device.

Data scientists, AI researchers, and experts in finance and healthcare who must swiftly process massive datasets would be those who would benefit from these machines.

He points out that AI-driven tasks can perform better with Arm-architecture SoCs, both for Windows and Mac devices, as they integrate AI capabilities directly into the chip.

"Choosing between Macs and Windows for AI activities depends on software and use cases. Macs' solid build quality and Apple ecosystem integration make them ideal for media-related AI activities like image and video processing. Due to its software support and customizable hardware, Windows PCs, especially those with NVIDIA GPUs, are popular for AI research and gaming," he says.

However, Ray points out that while AI PCs can benefit gamers if they use AI capabilities like real-time ray tracing and AI-driven graphical upgrades, pure gaming performance may be better on dedicated gaming PCs.

Is an AI PC worth buying?

Jones says that despite AI PCs' significant privacy advantages, a hybrid approach that combines local AI PCs with cloud solutions might be more advantageous for current applications.

"Cloud environments provide unparalleled scalability and access to constantly updated and more powerful AI tools, which might not yet be fully available on AI PCs. The concern about the limited software and the first-generation nature of these AI PCs is valid, so waiting for a more mature technology and software ecosystem could be wise," he says.

The expert points out that the ability for local AI experiences to work "on-device" will be of huge benefit, particularly for organizations wanting to take advantage of AI while protecting their sensitive information and intellectual property and not have it transmitted to public AI models for training.

"But for many organizations, there are real risks of "under-specifying" the hardware that will be required to even run the AI models/applications available today – let alone what will be coming to market in the coming months/years. The lack of clarity on minimum specifications means that organizations could "buy twice" – as they integrate AI PCs within their typical 3 to 5-year refresh cycles," the expert concludes.

Nikola Baldikov, the founder of digital marketing agency InBound Blogging, shares a similar opinion. According to him, those who decide to buy have to be prepared to pay a premium and understand that technology is rapidly advancing.

"What you buy for full price today might become outdated in just a few years. Holding off lets you see which features will become must-haves in the future. Personally, I would wait for the market to mature and offer exclusive AI-powered features at a more competitive price," the expert says.

Ray thinks that AI PCs may not be the best value due to limited software optimization and greater expenses, but they point toward the future of computing.

"It's worth considering if you need the cutting-edge today and can use early adoption for competitive advantage. As the technology evolves and the software ecosystem grows, budget-conscious purchasers may want to wait for future editions," Ray concludes.