It can be fun to engage in computing alternate-history fanfiction from time to time. Truth be told, I do it a lot. My personal predilections tend toward integration of computing and telecommunications equipment (and their vendors), slightly different products, and other things in the past. From time to time, it's interesting to engage in some armchair CEOing about the present and/or near future.
That in mind: one of the things that comes up from time to time is whether or not Apple will ever build an ARM based Mac. There are variations, but the core concept is the same: that Apple will at some point build one or more ARM based Macintoshes.
I'll start by saying that here in the real world, I'm pretty firmly in the camp that believes this isn't possible or reasonable. Here in the real world, there's one or two products that could hypothetically get the treatment, but they'd probably mess it up and it would make Intel very angry.
My personal variation on the story is that there are one or two products at the "bottom" of the stack that could use the treatment. An ARM-based Retina MacBook makes a lot of sense, as it's a product that's focused on being small above all else. The ARM chips should allow it to stay the same size while possibly getting a small increase in battery life. In something like that, you're at the situation where the display is almost certainly the largest user of battery. The Mac mini could also gain an ARM variant, and there is room for an ARM based iMac, but the question of how far you take it depends a lot on who you think will be using systems of any given form factor.
The iMac is complicated especially as the traditional expectation is iMacs are more powerful than the MacBook, Air, and mini. The reality is that the iMac is available on a horsepower gradient, going all the way from systems that use the same i5-4260U CPU that the MacBook Air does, to systems with 4.0GHz Quad-Core chips that should outperform the base Mac Pro. It's one physical form factor, but different price and CPU options dictate different tasks each specific model is good at.
I personally believe the most likely scenario for the near to mid-range future is that Apple would release one ARM computer to sit alongside its Intel-based brethren, either for limited availability, or for relatively normal availability, but with the caveat that it was an "experiment" a lot like the Apple TV was. It would have to cost a bit less than the Intel-based model, and be faster in one or more ways. That will be the problem facing all potential desktop experience computers using ARM chips. The vendor must prove that it's either a better experience or cheaper. Ideally it'll be both.
Today, the only system Apple can do that with is the MacBook, and possibly the Mac mini. The question is, how much will that improve over the next few years? Apple already claims that the A9X CPU in use in the iPad Pro is faster than most laptops that get sold today, but it clearly doesn't count its own in this figure. So what it means is that it's faster than most laptops that cost less than $600 or so, which is where you tend to find Silvermont Atom cores, or i3s.
Apple could put an ARM chip in a Mac Pro. They'd need to figure out how to get PCI Express lanes in and out of the chip, for graphics, but even though you "can" put such a chip in doesn't mean it's the right one for the job. There aren't currently any ARM chips (certainly none that Apple makes) suitable for the workload of that system.
The most commonly suggested "Big Problem" seems to be in how you handle the aspects of maintaining an OS for two CPU architectures at once, as well as how you manage porting software to the new architecture, or creating an environment that incentivizes writing software that runs on both Intel and ARM.
I think that Apple has the solutions to these problems a lot closer than people suspect, but the solution introduces another problem. The solution I propose is a combination of a return to Universal Binary executables, and a heavy emphasis on the use of the Mac App Store for software distribution, both for the ARM MacBook specifically, and for the Mac platform moving forward.
The Mac App Store is kind of an odd duck. Apple could easily use it to distribute universal binaries of programs, or the correct binary for whatever system you're on. The App Store has its own problems, however. Although few people are willing to really pester Apple about it, the App Store (both on the Mac and on iOS) has a few things that are really missing and preventing it from being a good distribution platform for high end applications. For starters, it could use the ability for developers to offer a trial version of their app, as well as upgrade pricing to advance from one version to the next major version.
There are some other problems with the app store, primarily pertaining to visibility of applications, searching algorithms, what ends up on the front page, and so on. It's not always a very good environment for developers or even for consumers, and there's even the question of whether or not the 70/30 profit split (and the rules about in-app purchases and using the app to initiate subscriptions) is reasonable.
By all appearances, Apple plans to keep everything the way it is and this is probably one of the biggest problems with Apple's culture and the culture surrounding Apple today. Developers don't like it, but there's no other way to get your programs running on the iPhone and iPad, and those platforms are the ones where people who are willing to toss a few dollars into something on their phone are.
That brings me back to the ARM-based MacBook. Back when Apple transitioned to the Intel-based Macs, their recommendation was to write everything in XCode. Everything is still written and compiled in XCode, and Apple can easily add the ability to build universal binaries in XCode again. Many Mac developers are now a lot more responsive to changes in Mac OS X and the environment than they were ten years ago when the announcement that Apple was going to switch to Intel processors first landed, and it took many small developers several months to release updates to their applications. I like to think that if such an announcement dropped today, most developers would be preparing their applications almost immediately, and that with Office and the Adobe Creative Cloud now on continuous release patterns, we'd see versions of those large applications much faster than we did in 2006-2008 when the PowerPC to Intel transition was on.
I think the real question is whether the ARM MacBook gets sold as a fully-fledged Mac, or as some kind of appliance, something to be sold alongside the iPad, as an integrated system that's supposed to be better than the iPad if you have typing-intensive work to be done, or something that's a limited, low cost system, comparable to a Chromebook or Windows RT type of device. Essentially, it would be the computer for that particular type of Mac user who never installs anything other than Pages, Photos, and Safari on their system.
Although I don't know if Apple or any of its executives has said anything on the subject, I don't think they really like the Chromebook concept, nor do I think they thought very highly of the Surface RT, which means I think that the main option left is for the ARM MacBook to be a fully-fledged Mac, and an open system for whatever developers want to code and sign, as with Mac OS X on Intel.
I think the way it would play out is that Apple would release the system, perhaps with intent to release more, perhaps as an experiment. I think that we'd find that modern Mac developers are fairly responsive to what Apple does, a lot more responsive than they were ten years ago, but software availability will still be a problem. They'll sell a few (I'd buy one) but I don't think that ARM chips can emulate Intel binaries fast enough to do something like Rosetta, the way that Intel systems were able to do with PowerPC binaries.
The real problem with this suggestion is that I suspect (and I think Apple knows this) that Intel would retaliate or otherwise be very angry with Apple for doing so. I can't say for sure whether or not that's true. I doubt that Intel exactly retaliated against, say, Dell or Lenovo for building Windows RT devices. However, it's been pointed out that Apple has significant input into Intel's product design.
Some people have said that Apple would have a hard time maintaining the OS for two different hardware platforms, but they already maintain the base code on ARM for iPads, and everything else, they'll literally be able to click some checkboxes to cross-compile higher level components such as the desktop applications, so I'm not exactly worried that it'll take a really strong engineering push to do it. For third party applications: One of the recent changes for the iOS app store is that developers submit intermediate files, rather than fully compiled binaries. It's possible for Apple to request that the source of intermediate files (ready to make binaries for both ARM and Intel) be submitted to the store. Developers that do this are ready to have their applications ported anywhere else at some
Even in a world where Apple resolved some of the problems with the app store, Intel didn't retaliate against the product, and developers got their applications ported quickly, I worry about the long term success of a product. The MacBook is already probably one of the least popular Macs Apple has released in a while. It's got good but unimpressive battery life, a "weird" keyboard, and relatively low performance. (Core M benches well compared to normal 25-35W Sandy Bridge chips, but it's a lot slower than its Haswell and Broadwell contemporaries.) Pricing it lower or giving it a different processor probably isn't going to help the reasons it's unpopular, but since it's unpopular, it could be a good proving ground for that different processor.
I don't think that such a product is impossible, but I think it would prove to be unpopular. In my experience, Intel chips are longest-lived in terms of getting software updates and in terms of being able to keep up with tasks. I suppose the question is when you give it eight gigs of RAM and put full Mac OS X on it, how much faster is the A9X or a potential A10X going to be at those types of workloads. It's doomed from the start, of course, because Apple's not going to build a variant of a laptop with a new type of processor without some kind of long term commitment to that platform, and I don't think Apple will add or change one of its laptop families to a new CPU platform without a commitment to transition the whole product line. And, this wraps back around to the fact that that won't really be reasonable for several years, and I don't think Apple is control enthusiast enough to lop off several of their top performance models to get rid of Intel in their product line, especially given how good Intel has been to them over the years.
But, given that it's all fan-fiction anyway, my particular version of the story is that Apple builds one model of the machine, it is a complete Mac, but isn't very well received, and then gets discontinued. OS support for it may or may not be killed. Far earlier than it "should" be, especially if the machine gains an enthusiast type of following. The real problem is that I highly doubt that binary translation or emulation would work anywhere near as well as it did in 2006-2007.
It's interesting talking about ARM-based Macs from the context of having used ARM-based desktop experience computers (the Microsoft Surface RT in particular) and the newest and best ARM chips being reasonably competitive with the lowest end Intel chips that Apple uses. It's a different world from a few years ago when I dismissed the idea outrightly, if it was mentioned.
In 2005, leading up to Apple's WWDC conference, John Gruber outrightly dismissed the possibility of Intel-based Macs. I wasn't interested in the possibility at the time, either. If the A9 and A9X are as real as early GeekBench numbers suggest, then there may be a possibility. It'll be a matter of other factors (such as Apple's control enthusiasm, or costs) taking precedence over performance. Its fanfiction today, but it could be a reality, in some form, this year, next year, or in another ten years.