The problem is, it all depends on the market segment.

For your smartphones and <600 vertical px. devices, an ARM/non-x86-compatible processor is perfectly fine. Battery life is a -big- issue, and being able to run desktop applications isn't - As we found on the N900, even when you -can- run full programs like Gimp or OO.org, the limited screen space makes it painful at best. And if you are redesigning the GUI already, compiling for a different chip is no big deal.

For tablets and <768 vertical PX devices(6-8"), the same thing may apply, but you can probably put in a bigger battery so power usage isn't quite so much of an issue. Still, an ARM chip makes sense in these devices - You want portability, and you don't need all the features of a desktop chip.

For laptops, especially desktop-replacement ones, things change. You need power at low heat, the ability to run Windows(realistically; only people like me use Linux exclusively), and battery life isn't expected to be great - four hours of light usage is typical, one or so under full usage conditions.

For full Desktops, Power is important, especially in gaming, graphics workstations, or CAD applications. You want a heavy-duty processor, powerful graphics, and who cares about power usage(So long as you can deal with the heat generated). Intel's the best, provided you have $$$ to spend. AMD comes in a close second, with the best power/cost, at the expense of a little top-end power. Still, as more applications become multi-threaded, AMD will be a better and better choice.

For Servers, most everything is massively parallel(Look at Apache & web-servers for instance). It looks to me like the requirements are lower heat and cost, with performance/watt being king. AMD's pretty good at this, though Intel again can provide more performance/chip.

So, provided I haven't screwed up -too- badly with my representation, there's room for multiple companies and multiple segments. Intel's not going to lose any of its main market shares to anyone other than AMD, unless we see a sea-change in the market.

For example, what would happen to the market if - instead of using a CPU for most all tasks - the new concept is using a small and fast, featureless CPU that ends up doing on-the-fly reconfiguration of a PLC or DSP to optimize it for whatever tasks being done. Provided the applications could cope, it could certainly shake things up in the computer market.

Just my thoughts...
-Rob