If 2023 was the year of artificial intelligence, 2024 might end up being the year of the AI PC. Plenty of so-called AI PCs are on the market today, and more will come throughout the year. While cloud-based AI tools don't require much, if any, new hardware, the same can't be said for AI that uses on-device processing. The recent advancements in AI and their corresponding performance demands for on-device processing could mean Windows system requirements could finally be pushed forward.

It might seem wild to believe, but it's been a while since minimum and recommended system requirements have meaningfully increased. You can daily-drive Macs and PCs from over a decade ago with legacy OS patchers or simply install a Linux distro on your PC. Brand-new Macs with 8GB memory and some Windows laptops are still sold in 2024. For both types of computers, 16GB has been the standard for quite a long time. AI might be the thing that breaks decades-old notions of what is good enough for the average PC.

What is an AI PC, really?

It's the latest computing buzzword without a clear definition

For a term that's thrown around as often as "AI PC" is nowadays, it lacks a concrete definition. The logical answer is that systems with a Neural Processing Unit (NPU), or Neural Engines on Apple Silicon computers, are AI PCs. However, it gets more complicated than that because plenty of AI-based tasks can be completed with off-device processing. It stands to reason that true AI PCs should be capable of on-device AI processing, though.

Microsoft itself doesn't clearly define what an AI PC is right now. That said, the company explains what artificially intelligent computer systems are on the resource page for Microsoft Azure. Here's what the company says:

Using math and logic, a computer system simulates the reasoning that humans use to learn from new information and make decisions. An artificially intelligent computer system makes predictions or takes actions based on patterns in existing data and can then learn from its errors to increase its accuracy.

Without explicitly stating it, Microsoft is describing an AI PC as a computer with an NPU. The entire purpose of an NPU is to optimize parallel computing and more flexibility, thus creating a computing process that reflects human reasoning.

Why modern AI tasks require better specs

Running local AI models requires an NPU and a lot of memory

Considering that having an AI PC at all means having an NPU, system requirements for a so-called NPU are bound to include a new processor. Intel's new Core Ultra chipset has one, and it has already appeared on laptops such as the Acer Swift Go 14 and the HP Spectre x360 14. Qualcomm's upcoming Snapdragon X Elite computing platform also features an NPU, and AMD's new Ryzen 8040 series processors have NPUs as well. As such, the baseline for having an AI PC is having a chip with an NPU, which most Windows PCs currently out in the wild don't have.

Aside from the NPU, on-device AI processing requires a significant amount of memory. With a software tool like LM Studio, it's possible to run LLMs locally on your computer. There are a few advantages to running an AI model on your PC instead of using cloud processing, including cost. It's quite simple to start with something like LM Studio as long as you have the right specs. For example, LM Studio, in particular, recommends 16GB of RAM or higher, with more being better. Video memory is also key to running AI models locally, and LM Studio says 6GB or higher is recommended.

The performance demands of on-device AI processing have led to talk of Microsoft officially increasing the minimum memory requirements for AI PCs. A report from TrendForce says that Microsoft will mandate a few things for AI PCs, such as Copilot, an NPU, and 16GB minimum RAM. This would be a big shift for Microsoft because it hasn't yet hinted at separate requirements for regular Windows PCs and AI PCs. However, it would make sense, given the recommended and minimum thresholds for running local AI models.

That report should be taken with a grain of salt because Dell's new XPS 13 will have a Copilot key, an Intel Core Ultra chipset with an NPU, and just 8GB of memory. This was first pointed out by Windows Central and is very easy to see from Dell's press materials. AI is all over Dell's press release for the XPS 13, and features like Copilot are obviously present on the company's newest laptop. It's perhaps the strongest evidence that 16GB will not be mandated by Microsoft on AI PCs.

System requirements are finally climbing

Microsoft might not demand AI PCs have higher specs, though

But even if Microsoft doesn't outright ban AI PCs with 8GB of memory, system requirements for Windows machines will start to climb for the first time in a while. Microsoft gives minimum system requirements for PCs, and then it gives the recommended system requirements that people should use. The company may increase the latter system requirements for AI PCs, but not the former.

If that doesn't happen, the narrative around Windows system requirements will certainly begin to change. Reviewers like myself will start recommending systems with NPUs, more memory, and more VRAM for buyers who want to use AI. Manufacturers will start to phase out 8GB machines as consumers gravitate to more powerful systems. Whether it's required or happens naturally, Windows system requirements are going to change in a meaningful way for the first time in a while.

Just a few years ago, Windows 11 introduced new hardware requirements, like a TPM, a 64-bit processor, and 4GB of RAM. However, there is a breadth of users who are still happy with using Windows 10. With all the new AI features coming to Windows 11, more people will become inclined to upgrade. Aside from the required changes to Windows systems, users will likely need to upgrade things like RAM and vRAM to keep up with the performance demands for AI processes.