In 2005, Hell froze over. After repeatedly doubling down on the PowerPC G4 and G5, with its laptops effectively stuck on older 32-bit PowerPC chips, Apple announced it would build future desktops and laptops around the Intel x86 standard. When the first laptops rolled out on Intel hardware, it was immediately clear Apple had made the right decision. Crippled by the older G4, even the fastest pre-Intel Mac laptops were no match for x86. Intel took over Apple’s CPU market and has maintained its lock ever since.
But now that’s changing, if Bloomberg is to be believed, and Apple will reportedly transition away from Intel silicon altogether and towards its own custom solutions for iOS and macOS both, beginning in 2020. As a move, this falls somewhere between “A long time coming” and “Hell just froze over, AGAIN.”
On the “long time coming” side of the equation there’s some big obvious points: Apple has been building best-in-class ARM processors for years now, with the highest single-threaded performance you can buy from any ARM core on the market, including those built by ARM itself. Android SoC vendors may win multi-threaded benchmarks and Apple’s own relentless pursuit of high-end performance may have exacerbated their battery woes, but the performance of Apple’s CPU cores in single-threaded code is excellent. It makes sense that the company would double down on putting its own silicon in its entire product lineup. Why give Intel market space it doesn’t have to?
Two other pieces of news belong in this bucket as well. Apple’s decision to support eGPUs and its development of its own internal GPU silicon means that it can build an Apple laptop with Apple silicon and shove any question of a third-party GPU into a separate accessory, should it choose to do so. Again, pushing Imagination Technologies out of the GPU block won Apple a larger piece of its own pie; pushing dGPUs into external chassis allows it to control its own product SKUs more precisely.
But it’s genuinely surprising that Apple would choose to abandon CPU compatibility given the significant impact x86 had on its Mac product lines. Mac adoption rates shot upwards once people knew their hardware would be seamlessly compatible with Windows. Walking away from that same compatibility now seems foolish, at least as far as good customer support is concerned.
Windows on ARM theoretically presents a solution to this problem, but the WoA OS is limited to 32-bit applications, with no support for x86 drivers, Hyper-V, and limited API compatibility. Supposedly this transition won’t take place before 2020, which gives MS and Apple another 20 months to get their ducks in a row, but 20 months isn’t actually all that much time to perfect cross-OS compatibility, especially not if the goal is to add better and more robust support for 64-bit applications and various types of system drivers.
There’s also the question of what a plan like this would mean for the Mac Pro. We’ve already seen plenty of evidence that there’s a major untapped market for faster Mac Pro hardware based on Intel CPUs; it’s unlikely that Apple could get away with keeping the 2013-era Mac Pro around for another 20 months without losing some of its professional customer base, who aren’t going to be excited about the idea of buying seven year-old graphics cards and CPUs.
At the same time, there’s not going to be much stomach for buying new x86 solutions if Apple is teasing a shift so massive it’s willing to launch an all-new line of high-end desktop CPUs. Serving the Mac Pro market with an equivalent lineup of ARM cores would require Apple to throw its hat into the high-end ARM silicon ring in a way that few companies have managed. Building an 18-core ARM CPU isn’t just a matter of slapping dual-cores down next to each other — not if you want appropriate and competitive performance scaling, anyway.
Economies of scale are another factor to consider. The tremendous success of Intel’s high-end Xeon products has always been partly underwritten by the overall success of Intel’s CPU division as a whole. Presumably Apple feels it can strike a similar balance between selling tens of millions of iPhone SoCs, thereby covering the cost of developing the much more expensive CPU cores that’ll power machines like the Mac Pro.
has always maintained that Apple could take this step, but have questioned the timing that made such a move out to be a near-term eventuality. Neither Apple nor Intel have commented on the Bloomberg story, but Intel shares took a significant hit, falling 9.2 percent. That’s despite the fact that Apple represents just 5 percent of Intel revenue, compared with vastly higher shares for companies like HPE and Dell.