Even if Apple develops and releases the absolute fastest CPU on the market in every benchmark and can virtualize x86 as fast as concurent chips, x86 (Or, at least Windows) will still be the most popular OS. Apple can't (more likely won't?) price their products to dominate the market, they are happy being a boutique brand and having the fat margins that come with that.
It seems likely that ARM will scale out of the low performance and low power draw it's at now, but it probably won't be Apple dominating that market. Windows for ARM will likely be on top, running on Qualcomm etc etc silicon, because that's what will be in $200 walmart laptops, and will be what the average consumer will buy. The average person just won't buy a $1000 laptop, no matter how good it is. In contrast, a $200 laptop with ARM and running legacy windows programs is a killer preposition, especially because even with what is right now a middling-capacity laptop battery it would have great battery life. Apple needs to be able to dual-boot Windows properly (Windows on ARM with (fast!) emulated x86 for compatibility) on their laptops like their x86 models do now if they want to keep or grow their existing market share. It's entirely possible that Microsoft snubs them on this so they can't and Macs have to make do with some shaky emulation layer that won't convince people to switch to Mac hardware.
I'm sure you all know that x86 was not always speed king, high end RISCs were at the (expensive) top of the market for a very very long time, up until about the Athlons came out and could compete. A possible scenario is a return to that, though extremely optimistic about Apple Silicon's scalability, as well as ARM/RISC in general. The extreme high end might be ARM or other RISCs in the future, but the jury is out on that. It's possible that in 2025/2030 the high end of the market looks like 1995, with fast RISC workstations from a boutique vendor (Apple this time) at the extreme high-end, and x86 and low-end Apple Silicon (PowerPC stand-in) competing elsewhere, though in a few years it's going to look a lot better for ARM than it did for PowerPC, especially considering Microsoft is looking to hop over for real this time. Windows NT ran on RISCs before, but they started in the high end so it was doomed to never reach mass adoption.
It's very likely that in the very near future, ARM-running-Windows will have convincing control of the middle-low to low end of the overall computer market, and that x86 will be slugging it out in the middle and high end (of laptops). On the deskop, it's likely that low-end soldered ARM will creep into the bottom end of the market someday soon, the part of the market that is currently Atoms and the low-end Celerons and Pentiums. I don't see ARM winning in the higher desktop space unless somebody sockets it and it scales really well. An ARM win on the desktop also requires ARM domination in the laptop and server spaces first either way, to build up software compatibility. Server shouldn't be too hard, that's already underway because servers are so power-usage-sensitive and heavily multicore, so having a lot more, but weaker, cores is no problem.
It seems obvious that RISCs with fixed instruction length can scale to higher-issue per core than x86 because they don't, say, have to deal with converting wacky maximum-15-byte memory-unaligned CISC instructions to microOPs for execution. Whether that balances out that RISCs naturally do less per-instruction is the pinch. RISCs can also spend more silicon on performance-enhancing features because they don't have to deal with the aforementioned x86->uOp decoding.