Macs dominate 'AI-capable' PC shipments – for now

According to a report by technology market analyst firm Canalys, the last quarter — Q2 2024 — saw an estimated 8.8 million AI-capable PCs shipped, with Apple Mac systems accounting for 60% of those sales. Additionally, when looking at $800+ Windows PCs, AI-capable PC shipments jumped from 7% in Q1 2024 to 14% in Q2 2024. 

Does this mean that there’s growing interest in PCs that can handle emerging workloads, or is AI capability becoming a standard feature? Before looking at that, let’s clarify a few things.

Also: What is an AI PC? (And should you buy one?)

First, what is an “AI-capable PC”? Canalys says that an “AI-capable PC must be a desktop or notebook possessing a dedicated chipset or block to run on-device AI workloads,” and goes on to highlight AMD’s XDNA, Apple’s Neural Engine, Intel’s AI Boost, and Qualcomm’s Hexagon as examples of such dedicated chipsets. 

The next point that always comes up when discussing the whole concept of AI-capable PCs is whether “AI-capable” is marketing fluff or something definitive and useful. The argument can be made both ways – some will claim that AI workloads are nothing new and that it’s just the type of workloads that have changed, while the point can also be made that modern AI workloads go beyond what the CPU and GPU can comfortably handle without making too many compromises, and that dedicated NPUs (Neural Processing Units) are needed to back up existing silicon and keep everything moving along.

I’m old enough to have seen the “this is marketing fluff” argument made numerous times about new PC features. Making and selling PCs is a business, so there will always be marketing. However, workloads change, and progress must be made. 

Also: ‘SinkClose’ AMD CPU vulnerability explained: How dangerous is it really?

And AI workloads are the latest new thing.

But we can tease some additional data from the Canalys report that might shed light on consumer interest in AI workloads. A fact that’s hard to overlook from the data is that Apple Macs account for 60% of the 8.8 million AI-capable PC shipments. Since Apple made the switch to its M-series chips in November 2020 and began purging Intel CPUs from the Mac, dedicated neural engines have been built into those chips as standard, making NPUs the default. When most consumers buy a Mac, they’re not choosing an AI-capable system, they’re getting one by default.

The same can’t be said for x86 processors, where AI support has been lagging. Dedicated NPUs form part of chips such as the AMD Ryzen AI 300 and Intel Core Ultra, but these are premium chips and not mainstream silicon, and as such will only be offered in higher-end systems.

Also: If Intel can’t come up with a Qualcomm-killer soon, it’s game over for x86 PCs

So, what about that 7 to 14% jump between Q1 and Q2? Is it an increased demand for AI-capable PCs or a side-effect of more chips having NPUs built in? 

It’s a mixture of both. “AI-capable” is clearly a buzzword of this upgrade cycle, and enterprise decision-makers looking to upgrade old hardware are going to be wary of investing in hardware that might be obsolete before its time if AI workloads become more common in mainstream applications. NPUs are also going to become more prevalent in systems in that $800+ category, which means that high-end systems will also drive the adoption of neural engines, making it an expectation for higher-end systems.

Then let’s not forget that Microsoft is just as much a part of the PC upgrade cycle as the hardware manufacturers, and is pushing the idea of AI being the next wave that no one wants to miss out on with it’s Copilot AI chatbot, and this feature will push the adoption of NPUs further and faster.

Also: What CPU and motherboard do I have? Here are some fast and easy ways to find out

There’s no doubt that, as AI workloads become more commonplace, NPUs will make it into more and more PCs, and it’s only a matter of time until these chips are no longer seen as specialist.