What if power saving is no longer a necessity (or much of one)? Once you take away that constraint, how powerful can an ARM chip be? Do we know?
The original ARM2 was first used in
powerful personal computers back in 1987 and, at the time, could leave its contemporaries - the Intel 286 and the Motorola 68k choking on its dust, despite running off very modest power. However, it couldn't run DOS/Windows*, and a non-PC-compatible British computer wasn't going anywhere outside a local niche... so when ARM Ltd was spun off they concentrated on the ultra-low-power embedded systems - and later mobile - market, while Intel threw the kitchen sink at the x86 to make it into a super powerful space heater.
No reason ARM can't play catch-up on performance: In the last few years ARM and others have started working on 64-bit chips, server chips etc. I think the ARM's trump card is that the cores are smaller and lower-power than Intel so you can fit more cores on a chip. Also, we're at the stage where the chips in the iPad, iPhone X are
fast enough for tablets, convertibles and ultraportable laptops - and that's where ARM will get in, because it has the huge advantage of pick'n'mix licensing, allowing the likes of Apple to build their own systems-on-a-chip to exactly match the needs of each device, and they're likely to be running lightweight, new-ish software which - if it isn't a platform-independent webapp - can be rebuilt for ARM just by re-compiling.
However, Mac Pros, iMac Pro etc. that need
real grunt, and must support huge Pro applications with hundreds of third-party plugins will be the hardest targets for ARM.
*Actually, it could - using software emulation - which worked pretty impressively but, with the best will in the world, wasn't fast enough to be viable if you were mainly using PC software. Modern software emulation/code translation is rather more sophisticated, and modern software relies far more on operating system-provided functionality that can be run natively...