ARM is starting to encroach on x86 in the server space:
- lower data center power requirements
- they're coming out with a 64 bit version
- ARM has a much smaller die footprint.
Intel must do ARM to stay in that game.
ARM would not be Intel's first foray into alternate architectures for x86 [8080, 8085, 8086, 80186, 80286, 80386, 486, 586, 686]. Remember Itanium [;-)] but also the 432. The Itanium and 432 didn't pan out because [the market for] x86 was so strong, but this indicates that while Intel is wedded to x86, it isn't slavish to it either. They care about making chips at a profit more than they do a given processor architecture. x86 has been a great tool to allow them to do that. But, x86 is just a means to an end for them.
When x86 ceases to be the asset it currently is, Intel will adopt whatever the market demands. The trend for this is ARM (vs. sparc, mips, etc.). At this point, [even] Intel can't kill ARM. ARM has too much demand for it now [it's a better solution for mobile and embedded/hybrid systems and will surpass in the server space in the near future]. Intel is adapting/reacting now, while it has time to do so on its terms instead of waiting 10 years and being forced to do it in a panic.
Contrast this with MS and Windoze. MS lost the mobile space race because of its insistence on Windoze. Intel won't make the same mistake, if for no other reason, they saw what it did to MS.
As to MS, most likely, in 10 years, we'll see MS/Office running on OSX/iOS, Linux, and Android with Windows just a fond memory.
Long term, Intel must become a foundry because it will lose its process generation edge (e.g. 22nm->14nm, but after 6(?)nm there isn't much room left. Others will catch up).
Intel will make money on this. In the mid 80's, Intel was selling its first generation 386 chip for $750. An Intel engineer told me that the same chip was designed to be profitable even if it sold [had to be sold] for $35.