1) They would retain a 27% margin -> they would cost more. The margin sustains R&D, which they could drop and survive. riiiggghhht.
2) Hardware design would become more 'ordinary', as power dissipation requirements would dictate. Not just the cases, but the motherboard chipset too. Custom northbridge with 4x ATA buses? Buzzz. Apple can do that _now_ because they drive the motherboard chipset on Macs. They wouldn't be able to on x86 - without the price climbing even higher.
3) The few niches where Macs really _do_ (or at least did) shine (single precision floats come to mind) would be lost. When the processor is _identical_, it _IS_ all about the MHz. When MacOSX-for-x86 is introed and is 20% slower at benchmark x, y, z. - now what?
4) The actual move is _possible_, but the difference between 'running', 'running well', and 'blazing speed' is the difference between DP1, 10.1, and 10.2 -> years. Darwin/x86 is apparently _VERY_ slow. Feel free to start optimizing if you really want Apple to go that way.
5) Developers haven't caught their feet after the _last_ major shift -> bad plan. The number of '1.0' style releases (bugs + slow) for X is quite high. Some developers aren't 'switching' to Mac OS X so much as providing their app for the XWindows environment that is available. Would anyone _bother_ to provide a non-Doze version if there was any way a Mac could run Doze?
After 6 more pages of arguing we'll get to 'But of course I meant on Apple Branded hardware locked x86s where they'd still be in complete control'... Locking it sufficiently to prevent the OS from _ever_ escaping the hardware into the at-large x86 base is near impossible without first un-opensourceing Darwin. If you tell me the BSA will help Apple enforce their licensing, I may not stop laughing. Ever.
Or maybe the 'How about a compatibility card?' Fine. But it will _also_ cost more & come out later. Apple did compatibility cards. They were pretty cool - but they were 9 months slow relative to the CPU of the same capability, and the card alone was the same price as a newer faster x86 box. Slamming everything on a single card costs. More expensive chips, hotter, more design work. If they got really good at it, it might not cost quite so much - but it is only for the people that _need_ that ability for some reason, not ever going to be good enough for it to be a big big market.
The one other argument is 'What if they did it just for the extreme high end, like, say, Avid.' There I might see it - the added price of a CPU is cheap at that point if it helps whatever the specific goal of the niche is. But that doesn't help _us_, and I'd see that more as a compatibility card sort of situation also.
The question isn't 'what are the downsides', it is 'name a single upside'?
2.53 GHz, 2.8GHz... it would be years before the full speed would be realized. And no, a 1.4 GHz ppc doesn't match a 2.8 GHz x86... but what about a 1.8, a 2.0, or a 2.2? Do they? Most of the lagging should even out if Moore's observation holds, so then where are we? We switched to the 'faster' chip just in time for their process to hit the same sort of slowdowns that hit the ppc side.
'Lower prices' are commonly cited as a plus. Lower prices on _what_? Drives? Um, same drives. Optical drives? Um, also, pretty much the same drives. RAM? Slots? What? You only get 'cheaper motherboards' if you follow the herd. You only get cheaper cases if you follow the herd. If you follow the herd... I hope our cow boxes are at least designed to resemble a cool cow like the Black Angus. We could at least save the $10 on the no-floppy design. That'll clearly be enough to draw flocks of new adopters.