AI for PCs – Is it all hype or should you switch?

Beyond the hype

It’s fair to say that AI PCs are fast becoming a major talking point for vendors and organisations alike. Yet, for businesses who are looking to refreshing their devices, making the right choice has never been more important.

In a recent webinar, Elliott Jones spoke with AI expert Rob May to discuss whether the talk around AI PCs is all hype or if there is more to it. In this article, we offer the key takeaways – with video clips (with subtitles) – from the webinar to help you make informed decisions as we enter a new era of computing.

What is an AI PC?

AI PCs contain a dedicated chipset designed to handle artificial intelligence applications and to utilise AI to improve performance, boost security and provide greater personalisation.

Programmes powered by AI are nothing new, but PCs featuring neural processing units (NPUs) designed to enhance machine learning tasks – are a relatively new frontier. With the rise of AI-powered chatbots such as ChatGPT, we’ve heard plenty about Large Language Models (LLMs) – algorithms that through machine learning and large datasets have the power to understand and reproduce human language.

AI PCs work in a similar way on a smaller, more localised, scale, using Small Language Models (SLMs), which, though more limited, are better suited to optimising individual devices and performing more specific and focused tasks. One key benefit of SLMs is the ability to selectively move data between a machine’s physical storage and cloud storage networks, which AI expert Rob May thinks offers the best of both worlds.

“That mixed approach contains the strength of the cloud for intensive tasks and data storage with the speed and the privacy advantages of local processing,” he explains. This combination of advantages, Rob adds, “lowers latency, saves bandwidth, improves data security and reduces the amount of sensitive data that’s sent to the cloud.”

Where are we now with AI PCs?

2024 has been tipped by many as the “year of the AI PC” – but does that stand up to examination? Market analysts are consistent in their predictions, expecting around 50 million AI PC units to be shipped in 2024 – making up roughly one in five PC shipments. Researchers from Gartner expect 43% of all PCs shipped to be AI-ready next year, while the International Data Corporation expects the number to hit 60% before the end of 2027.

This all points towards AI PCs soon becoming the norm, but what are we using them for? Many applications optimised for AI PCs are currently used for video conferencing and project management – such as Zoom, Slack, or Webex – or creative design tools like Adobe, Audacity or GIMP.

This might suggest a lack of breadth in this technology’s use, but Rob suggests these apps are an example of the “early stages of AI integration” and that there is still plenty of room for growth. “It’s targeting areas with high demand and clear benefits from AI enhancement, and whilst the focus might seem narrow, I think it’s an obvious starting point,” he says. “So, I don’t think that apparent lack of breadth should be a worry, but rather, I see it as a developing landscape. “As AI tech becomes more advanced and more accessible, its applications are expected to diversify massively, and the current focus areas serve as a testing ground. “Also, I think these applications drive the improvement of hardware and software ecosystems, which then create the foundations for broader AI integration.”

When to refresh?

AI PCs will change how we work, which is why IT managers may be tempted to buy straight away and take advantage of improved productivity, better security and privacy, and lower system latency. However, some businesses may be opting to hold out for longer because emerging technology is often more expensive at first, combined with the fact that applications are still relatively limited.

Faced with such a fast-moving market, you could buy yourself time by upgrading your existing systems with the latest generation of DDR5 memory and NVMe SSDs to get the most out of emerging AI applications. Much of this will depend on budget, requirements and where you are in your workplace’s technology refresh cycle. If managers opt to buy new AI PCs instead, they should be wary of models with parts that are soldered to the motherboard instead of socketed, which could result in being locked-in with inadequate storage, memory, and functionality.

IT managers should be thinking carefully about the cost of replacement against their company’s needs – not only now, but also in a few years. As AI applications build up data on users’ preferences and how they use their systems, this in turn will increase a workplace’s storage and memory requirements. It can be all too easy to under-spec when adopting AI PCs, which is a mistake for which you could be paying for up to five years, meaning it really does pay to think ahead.

Watch the webinar