AI PCs aren’t the future: They’re here now and businesses need to take advantage

The time for AI cynicism is over, with businesses moving beyond the exploratory phase and finding tangible, real-world gains in efficiency and productivity. Recent research from Gallup found that 65% to 68% of US employees who use AI for research, writing or editing say it has improved their productivity. The figures are even higher for those who use it for coding or process automation. In the UK, figures from Bloomberg and Morgan Stanley suggest an average 11.5% increase in productivity for British businesses using AI.

Right now, those productivity lifts come from frontier AI: advanced, general purpose LLMs, running at massive scale in the cloud. These offer complex, multimodal reasoning capabilities but use vast compute resources. Think Anthropic Claude, Google Gemini, OpenAI ChatGPT.

However, this is changing, as more enterprises look to bring AI PCs into their fleets. According to a recent IDC whitepaper, 60% of organisations are piloting or have already deployed AI PCs, with a further 21% planning to do so within the next 12 months.

So what’s driving this change? And is it something you need to take action on?

The wind under AI PCs’ sails

There are two major factors behind this shift. Firstly, new AI-enabled laptops incorporate powerful technology that can run AI applications on-device. It’s no longer the general-purpose CPU cores that are shouldering the AI computer burden; Neural Processing Units (NPUs) built into the latest Intel, AMD and Qualcomm processors are designed specifically to handle the intensive parallel processing workloads involved. They can chew through AI tasks faster and do so without hitting the battery hard, as a CPU or GPU might.

These devices can also utilise integrated or dedicated graphics processing units (GPUs) to run more demanding models. With the new wave of AI PCs, AI isn’t constrained to distant servers or high-performance workstations. It can run on the laptop that’s sitting right in front of you.

Just as importantly, AI is evolving. The big LLMs – ChatGPT, Gemini, Claude – aren’t going anywhere, but there’s a significant growth in the use of streamlined LLMs and smaller language models that don’t require the same levels of processing power, or the same massive quantities of RAM. Where the major LLMs can feature hundreds of billions to trillions of parameters – the different values used to process data and recognize patterns – these smaller LLMs make do with under 10 billion parameters.

What’s more, the power of SLMs is only growing. We’ve seen SLMs perform just as well as an LLM from a year previous.

Power of the SLM

Frontier large language models vs AI PC small language models

While the table above summarises the differences between LLMs and SLMs, here’s the key takeaway: with optimisation and specialisation, SLMs can work more efficiently than LLMs with limited resources.

Models like Microsoft’s Phi 4 Mini or Google’s Gemma 3 4B might not be so hot a pondering the universe’s biggest questions, but they can summarise a document, dig up relevant info from a company knowledge base or pick out patterns in your business data. Especially if they’ve been trained with data specific to your industry or, even better, your actual business data.

Increasingly, they can do so at much the same speed on your device as on ChatGPT or Claude running in the cloud, or at least to the extent that there’s no serious difference. Gartner predicts that, by 2027, enterprises will be using task-specific SLMs three times more than the LLMs we use today.

This capable-device-meets-specialist-model combo brings tangible advantages. The models can be customised and fine-tuned for specific industries and tasks. With instant access to data on the device or the local network, you lose the lag and potential for disruption when you’re shifting prompts and data to and from the cloud.

This approach also has the edge on privacy and security. Sensitive customer or business data isn’t being uploaded to a third party, which may or may not be based in the same territory; it stays on devices under your direct control. Many businesses in many industries will appreciate the added privacy. Some in highly regulated sectors might find it essential.

Rising cost of Frontier AI

What’s more, there’s a cost implication. While LLMs used to offer “all you can eat” deals, they have now shifted to a usage-based or per-token model. This makes operating costs less predictable and, on the whole, far higher.

This is becoming another driver for AI PCs in larger enterprises, with Gartner predicting that around one-third could switch to AI PCs to reduce their cloud-based AI token costs. SMEs, running on tighter IT budgets, will surely follow down the same path.

This is the time to refresh your PC fleet and take advantage. Switching from cloud-based AI to on-device AI doesn’t happen overnight; it takes time to pilot initiatives, consider use cases and applications and train and integrate these smaller models into your day-to-day workflows. Building a solid technology platform right now can smooth the path to progress.

The devices are here right now, and they’re not expensive, high-performance laptops, but affordable, mainstream PCs.

Other advantages of modern AI PCs

AI PCs vs three-year old PCs

Nor is AI all that they to the table. A range of form factors, from slim-and-light ultraportables to big-screen, high-performance systems means there’s something built to fit the needs of every employee, whether desk-based or always on the move.

Even devices under 15mm thick and 1.4kg in weight have the power to run demanding apps. Spacious keyboard and high-resolution screens support intensive multi-tasking and all-day work, while Wi-Fi 7, USB4 and Thunderbolt 4 connectivity open up access to high-speed external storage and faster, more reliable wireless networks.

Perhaps most importantly, the latest technology brings dramatic boosts to battery life. Instead of talking about laptops that can get you through the average working day, we can talk about laptops that can power on through into the next afternoon and beyond. It’s possible to leave your charger at home for a day or two and still feel confident you won’t run out of charge.

Rising component costs make it tempting to sit your PC refresh out. But the earlier you investigate and implement what AI PCs can do, the earlier you can make that count within your business.