The biggest problem that people do not understand is ARM is fast ONLY in lightly threaded workloads. When you get something that has to scale, something heavily threaded, it chokes performance. Latest Graviton v2 ARM Chip, from Amazon is faster than anything in Single Threaded workloads, or lightly threaded, but those 32 ARM cores, clocked at 2.5 GHz completely choke when they have to scale.
Could you technically explain exactly what in the ARM architecture prevents it from scaling? Beause stating that without any argumentation or logic behind is like saying x86 scales horribly because Intel +8core CPUs are monolithic toasters that make its power consumption unbearable for laptops (that'd be cherripicking to prove something, so you get my point). One specific case doesn't make a law. AMD has overcome the scaling problem with Infinity Fabric, it's not like there aren't solutions/alternatives for architecture bottlenecks.
You also seem to forget 2 important things:
1. If you design your own chip you can add specific hardware modules (like NPU for AI, or VPU/IPUs for video and image processing) that are way more powerful at that specific task than general purpose cores. So you don't need to have a big amount of cores, specially in laptops/tablets.
2. We are far yet to see ARM desktops, ARM will be used in laptops wich aren't intended to substitute powerful workstations where high core and threat counts matter. The scaling problem becomes negligible on laptops the moment an iPad pro is already equally or more powerful than a MBP.
You cannot have BOTH worlds on Mobile.
Sorry but what comes after that phrase is just wording and opinions, not facts. Ironically the new AMD laptop chips (the x86 that you defend) proves you wrong on your quote: the new 4900HS performs like a desktop (among 3700X and 9900K) having a power envelope of a mobile chip, that's why it's put on a laptop.
For what you say next: all your points are that you need lots of money to redesign an ARM arch, that it would add more power consumption in computers, and that it makes no sense. Those are really easy to answer: Apple has all the money they want to put into R&D. Not only that, they already spent it on a full ARM design which gave great results and that (responding your 2nd point) performs like a Mac consuming way less. Taking the main workload of designing a chip that already does like laptop x86 competitors and improving it is a breeze compared of what Apple CPU designers have already done. The main point of doing it is not depending on any 3rd party vendor, and specially having more benefit margins.
You also seem to be implying x86 is the only general purpose architecture beause it "has to fit ever, single glove, out there". Well there's news for you: almost any actual type of processor has all the units needed to compute any task, but that doesn't mean it will be good at it. That's why GPUs, NPUs, and specific hardware in SoCs exists even you can play Crysis only using the CPU (in the AMD 64c Rome for example). People seem not to realize that ARM and x86 are basically the same, the main difference is one is full RISC and the other just has a CISC decoder that translates complex instructions into reduced ones, but the rest of the pipeline and operations are RISC. And all the "ARM isn't good enough for heavy loads" myth comes because until now it was only designed and used in mobile stuff, but there's nothing in it's architecture preventing an ARM chip to be designed for high performance computation.