Back in the PPC->x86 transition, the practical performance gains were significant. Especially if you were coming from a G4 machine, using an x86 Mac just felt fast. Compute operations actually did complete sometimes 2x faster, and you felt it. Combining that with the fact that you now had native Windows support, and the fact that virtual machine products could use native emulation, it was a pretty easy sell to convince customers to move. Developers also liked being able to target and optimize for one architecture. There were some very clear advantages to getting on board with x86 that went far beyond just performance or power usage.
I'm having a hard time seeing similar gains for a move to ARM. With ARM, we give up native Windows support (sorry, Windows on ARM is still very beta in my experience). ARM chips may be more power efficient, but they don't necessarily perform faster, especially not faster in a way the average user will be able to feel day to day. Based on some benchmarks I've read, some of the best ARM CPUs are on par with the high end Core M series processors - and while those processors are decent, they're nowhere near even a Core i9 laptop part, let alone a Xeon processor as seen in the Mac Pro. The only place I see any benefit would be in laptops, where you might be able to advertise a Macbook that can last 12 hours in real world use on one charge. But this won't matter much for desktops, and pro users will demand the performance over the longevity.
The best outcome I see here is a mixed arch ecosystem. There's little reason today to force people onto one architecture. We've come a long way with compiler technology and automated builds and such, and if I recall, Apple already is doing some sort of bytecode-based compilation for the App Store, leaving the door open for a multi-arch runtime. Most other major programming languages are multi-arch (.NET, Python, etc.). I believe that Apple would be making a mistake if they try to move the entire ecosystem over to ARM. ARM has its place, and so does x86. x86 for pro, ARM for portables. And like I said, there's almost no reason not to do that these days. As Steve once said: "let each element be true to itself." If an ARM CPU wants to run at low power, let it serve the low power market. If an Intel chip wants to provide killer raw performance at the expense of higher power consumption, let it serve the pro market.
Then again, I'm sure Apple would love to be able to finally end the Hackintosh community. Ulterior motive?
