It's worth noting that, until June 2006, Intel made their own ARM processors before selling the business to Marvell. They saw the conflict of interest between their ARM lines and their forthcoming Atom lines, and threw their weight behind the architecture that they controlled. ARM chips are also much less expensive than Intel is used to charging for it's silicon -- they wanted to keep their costly fabs churning on high margins products.
This would effectively be a retreat from those ideals, and reduce them back to the role of another commoditized supplier of ARM processors -- something they had hoped to avoid by pushing x86 in the low power market.
Porting processors between fabs is not easy. When Marvell made this transition in reverse, by buying the XScale ARM business from Intel, it took a long time to move from Intel's fab process to TSMC.
All the generations of XScale are 32-bit ARMv5TE processors (ARMv5TE ISA without the FP instructions) manufactured with a 0.18 µm or 0.13 µm (as in IXP43x parts) process.
I think the reason for Intel selling off most of the Xscale business to Marvell is that Marvell has an architecture license, and ARM, Ltd probably didn't want to sell one to Intel.
Apple helped develop ARM, so it likely has a less expensive license than anyone else.
Given that Intel knows how to get an ARM5 into 0.13, getting a theoretic A6 into its 0.22 process is likely straight-forward.
It would, however, be pretty big shift in strategy for Intel. They've been the best at silicon process for a long time, and they could have chosen long ago to be a fab house for everyone. That, however, doesn't deliver high enough margins for them.
They want a product that people _want_ -- something that pushes them up the value chain. That's hard, since normal consumers don't understand or see processors. With stickers and ads, they've created brands like Pentium, Core 2, and i7. (My Mom couldn't tell you that she has a Snapdragon processor in her phone, but she knows she has an Intel processor in her computer.) Considering Intel's margins are _much_ higher than most silicon vendors, they've largely succeeded.
This makes them give that up. Intel's fabs are more expensive but superior. Samsung and many others offer slightly inferior chips at lower costs. Winning this business may be good for Intel, but it will hurt their cherished margins. Apple pushes their vendors very hard.
Intel recently let Achronix use their fabs for 22nm FPGAs , so this is not really a new move by them either (that said, Achronix was a /tiny/ deal. This one may not be.)
Intel has a 22nm process coming online this year. I doubt the A4 or A5 are ported to this, but an A6 could be designed for the Intel 22nm process.
Apple could probably run the A5 @ 1.25GHz in intel's 34nm process (were it to port same.)
So, more features, less expensive parts, less power consumption per GHz. I see Intel as a definite possibility, especially if Apple can exclude other competitive ARM SoCs from Intel's process.
Intel: "You can't have ARM, but would you like an Atom, sir?"
Intel/Micron also have 20nm flash now. Densities are about to scale in mobile storage as well.
This would effectively be a retreat from those ideals, and reduce them back to the role of another commoditized supplier of ARM processors -- something they had hoped to avoid by pushing x86 in the low power market.
Porting processors between fabs is not easy. When Marvell made this transition in reverse, by buying the XScale ARM business from Intel, it took a long time to move from Intel's fab process to TSMC.