Interesting. The article says that Apple is considering an ARM
co-processor, not a replacement for the Intel CPU.
I have no doubts at all that Apple have at least had extensive internal discussion and testing about this. In fact, I believe they've probably done extensive internal discussion and testing about replacing the Intel one altogether, but can't get there yet.
It'd give Apple more control over their device design and release schedule, potentially lower power consumption, and straight up more profit via vertical integration.
Like the PowerPC transition, they could come up with a new Rosetta - which was the emulator that ran old PowerPC apps on x86. Microsoft has recently demoed x86 apps on ARM, and I'm sure Apple could do the same.
As far as I know there's no fundamental reason ARM processors can't be as fast as x86 ones, they just haven't been targeted for those kind of devices yet - but I'd be happy to be corrected by any CPU experts. And with Apple's success in the mobile processor space, I have no doubts that if anyone could pull this off, it'd be them.
So what's holding them back? Thunderbolt. They've gone all-in on that already - touting as the future of high speed external devices, and giving the MacBook Pro nothing but four Thunderbolt enabled USB-C ports and a headphone port. But Thunderbolt is an Intel property, only available on their CPUs.
The 12" MacBook already doesn't have Thunderbolt (the MacBook Pro does), but it feels like with software compatibility, this would be an all-or-nothing thing. I don't see Apple continuing to sell both x86 and ARM machines on an ongoing basis. So how would they get Thunderbolt into the MacBook Pro?
Having a co processor to handle PowerNap would allow them to take a small step in that direction without losing Thunderbolt or having to develop a slow x86 emulator. They could even offload other parts of the OS to the ARM cpu, freeing up the main CPU for other software, and to go into low power mode more often.