Out of the mouth of MS, on Windows ARM, for those thinking ARM will be displacing Intel 'any day now' for higher end computing:
http://armdevices.net/2011/01/06/microsoft-shows-full-windows-for-arm-powered-devices/
It's all about the mobile market, at least short term.
For those going back to why the move off of PPC, they were some great desktop/workstation and server chips, but there's wasn't much in the way of being able to move the iBooks into G5/970 territory as far as I recall.
Amusingly, the G5 based Apple servers performed pretty lousy under OSX 'Server,' but the same machines were pretty competitive vs current gen Xeons when running Linux. There was one benchmark in particular that was pretty clear on this in useful server situations - running an *AMP stack (it may have been just apache, couldn't quickly find the link, but effectively a real world was night and day difference between OSX and Linux on the same hardware..more than likely OSX's wrapping of BSD threads and mach threads wasn't doing any performance favors, as well as BSD threading being somewhat sub-standard at the time already (vs Linux, Solaris, etc.).
Burned once on 'nowhere to go' to bring updated mobile performance beyond the G4 powerbooks, I don't find it surprising that Apple doesn't want to get put in the same situation again, so 'reminding' Intel about power consumption and mobile-applicable CPUs isn't that big of a deal. And again, it's in Intels interests here, as there may be some future point at which ARMS do become performance competitive at the higher end, or if/when Apple decides they're ok sacrificing some performance on an Air for 15 hour battery life...Intel would prefer to have a viable embeddable CPU as they failed with XScale (ARM based) and Atom (including it's new gen) isn't too competitive with Tegra/ARM either...they don't want to start losing market share 'at the bottom' if they can avoid it.
In the short term, code execution on non-native platforms isn't anything new, really - besides Rosetta - BSD, Solaris and others had the ability to run Linux binaries, going back even to SCO (google 'lxrun') and the late 90s in that case..but I'm not seeing ARM at this time being capable of running x86 OSX code at anywhere approaching native speeds. As tablets move 'upward' meaning they can do more and more laptop/workstation tasks, that's where ARM is potentially going to keep growing...most likely with a future longer battery life Air or next gen 'super tablet', not on our MBPs or Mac Pros.
Some reading on ARM current (I believe not yet shipping) latest CPU architecture, A15:
http://www.anandtech.com/show/3905/...-headed-for-smartphones-notebooks-and-servers
The chip is aiming at the same market as Atom - tablets and netbook class, it's not too likely to displace current/next gen Mac Pro or SNB CPUs.
The low power server note is interesting, but I think Intel is doing reasonably well there with boost technologies along with much more lower voltage research - these things do matter in datacenters, and it's not uncommon for hundreds to thousands of servers to be sitting at relatively low CPU loads with occasional bursts, so their ability to scale back on power consumption (and heat, requiring additional cooling load) is certainly important. I can say that prior activities I've been involved in at reducing datacenter power/cooling consumption involved a lot of consolidation of servers and storage to blade chassis, less and less local disk storage vs SAN, virtualization, etc., and power was definitely taken into consideration but there weren't too many options for the performance needed for some applications. If options were available to further significantly reduce power and cooling were available while retaining 'enough performance,' even if it may mean a somewhat larger number of systems - it could be a compelling story at least in some cases.