Apple ditching Intel to make its own processors for Mac products has obviously gone pretty well. Its ARM-based M2 chips pack both powerful CPU and GPU cores into a single package, offering a great blend of midrange to highish-end performance for the whole Mac portfolio. This has especially benefited products like the MacBook and Mac Mini, which don't have the room for a powerful graphics card but can accommodate the higher-end M1 and M2 chips (like the M2 Max) with powerful integrated graphics just fine. The M2 Pro, M2 Max, and M2 Ultra are pretty amazing.
So why don't Intel and AMD do the same thing and make their own big chips with lots of CPU and GPU cores? After all, they both offer good CPU and GPU technology themselves, they've been doing PC stuff for as long as Apple (and processor design for much longer), and they have far more control over the wider processor market. But ultimately, Intel and AMD have several good reasons to not copy the bigger variants of Apple's M2.
Intel and AMD do have the technological capability to make an M2-like chip
But first, let's get one thing out of the way: Intel and AMD can make a chip like the M2 Max or even M2 Ultra if they wanted. While the M2 Pro, Max, and Ultra are fairly different from what we're used to in PC, all Apple has done is make a big chip that includes both CPU and GPU cores. Intel and AMD have been making CPUs with integrated graphics for years, and the only substantial difference between the Core i9-13900K or Ryzen 9 7950X and the Apple M2 series is that the bigger M2 chips pack a much larger and more powerful GPU.
It's also not like Intel and AMD haven't already made chips similar to the bigger M2 processors, even before the M1 came out. Intel's Kaby Lake G series combined a quad-core Kaby Lake CPU with a midrange Vega GPU from AMD, and although Kaby Lake G was a commercial failure and didn't work super well, it did in the end offer what was at the time a powerful CPU and GPU combo.
One of AMD's main products of course is its APU, which is really just a marketing term to describe AMD's CPUs with integrated graphics intended for gaming. AMD's most successful line of APUs are the ones that go into consoles, ever since the Xbox One and PS4 came out a decade ago. The latest console APUs in the Series X and PS5 combine an 8-core Zen 2 CPU with a midrange RDNA2 GPU, putting it on a similar level to at least the M1 Pro and M2 Pro, and maybe even some of the higher-end chips too.
The M2 has some downsides that Intel and AMD wouldn't like
While Apple's M2 series is really cool and great for Apple, it's by no means the perfect processor, and it has some significant drawbacks compared to chips from Intel and AMD. These drawbacks primarily have to do with memory, core counts, and manufacturing, which have a big impact on both performance and price.
The weirdest thing about Apple's M processors (at least in my opinion) is the memory. Normally, CPUs and lightweight integrated graphics don't require that much bandwidth, so Intel and AMD tend to pair their mainstream chips with small 128-bit wide memory buses. However, GPUs require lots of memory bandwidth, which is why discrete GPUs are paired with GDDR VRAM and have more memory buses, with 128-bit usually being the bare minimum and reserved only for very small GPUs.
However, Apple puts very powerful integrated graphics on its M chips, and this requires lots more memory bandwidth than usual. While the M2 is normal and has 128-bit wide buses, the M2 Pro and M2 Max have 256-bit and 512-bit wide buses, and since the M2 Ultra is two M2 Maxes combined, that means the Ultra has a massive 1024-bit wide bus. These buses take up a ton of space and account for about 13% of the size of the M2 Max and Ultra, which is a ton of space to dedicate just to memory buses.
All that area reserved for the memory buses has a domino effect. It's less room for more CPU and GPU cores, which leaves the M2 Pro and M2 Max a bit unimpressive when it comes to performance per mm2. For instance, the M2 Ultra in the Mac Studio is well behind both the Core i9-13900K and Ryzen 9 7950X (which are two of the best CPUs) in benchmarks like Cinebench, especially in multi-threaded performance. These are much, much smaller CPUs thanks to the fact that they aren't hobbled by a massive memory system required by a large integrated GPU.
Trying to combine a big CPU, big GPU, and big memory system into a single chip has consequences for manufacturing. At an estimated die size of 550mm2 (assuming Apple's side-by-side comparison above is to scale as no actual measurements seem to exist), the M2 Max is super big, and the M2 Ultra is the largest consumer chip ever made at over 1,000mm2. The cost of producing these things on TSMC's 5nm node must be astronomical.
It's just not Intel's or AMD's style to offer highly specialized silicon for the mainstream
But aside from all the hardware problems that chips like the big versions of M2 introduce, there's also a fundamental difference in business models here. Apple is very different from Intel and AMD: it makes its processors for itself and its own special products. Intel and AMD meanwhile are making chips that power nearly every PC that isn't a Mac in the entire world, and pursuing that large of a market means specialization is a disadvantage rather than an advantage. That creates a very different set of incentives for each company when it comes to designing hardware.
If Intel and AMD were to try and make a big M2-like processor of their own, a major issue would be motherboards. What's nice about traditional, mainstream x86 CPUs is that they tend to be pretty small and don't require anything crazy. But to service an M2 Max/Ultra-like chip, Intel and AMD would have to launch new motherboards with massive sockets, tons of VRM stages, and probably eight memory slots that you would have to fully populate with fast modules to get good GPU performance. That's obviously very expensive and very cumbersome.
Apple can get away with all that because the M2 is purpose-made for what kind of computers Apple wants to make and what its customers want to buy. Intel and AMD can't do that because we like having a wide variety of CPUs that we can put into personalized but ultimately similar PCs, where it's just plug and play when it comes to RAM, storage, and GPUs. Imagine having to upgrade your 1,000mm2 processor if you just wanted faster graphics; if you thought the RTX 4090 was expensive, you'd probably go bankrupt from a single upgrade.
But it works both ways, and the M2 has some distinct advantages. For instance, while Apple's CPUs and GPUs are pretty weak compared to the high-end stuff Intel, AMD, and Nvidia make, the M2 product stack from top to bottom has top-end encoding performance. Apple can justify adding these encoders to its chips because lots of people already use Apple devices to edit videos. But since Intel and AMD CPUs are essentially one-size-fits-all chips for a wide variety of use cases, such high-end encoders wouldn't make sense, especially when high-end GPUs can pick up the slack.
Both the hardware and differing business models are inseparable from each other when it comes to processor design for these three companies. If it were just down to hardware and technology and specifications, then we would probably see Intel and AMD try to make a processor like the M2 lineup as well. But Apple is a computer-designing company that makes its own processors, while Intel and AMD are processor-designing companies that sell their processors to other companies.
There are some cases where M2-like chips from Intel and AMD could prosper
While the traditional PC is very unlikely to be a space where Intel and AMD can introduce a big processor with lots of CPU and GPU cores, I think such a chip could find success in other areas. One of the most recent innovations from Intel and AMD are chiplets (or "tiles" as Intel calls them), which would make it easier to build a processor like the M2 since Intel and AMD could just take chiplets that are generalized on their own and combine them in such a way that it creates this big, specialized processor (which would also be cheaper to make with chiplets).
Consoles are an obvious place where these kinds of chips have worked for x86, but NUCs or OEM desktops or even laptops could be more suitable for an M2-like processor. After all, these devices are already pretty locked down and customized to start with, so the benefits of using hardware that's designed to be compatible with lots of other components isn't such a big deal. I don't think the demand is there to make this a reality, but I could be wrong, and honestly, I hope I am because I would love to have what is basically a gaming version of the Mac Mini or Mac Studio.