What you might be thinking of is the UMPC (Ultra Mobile PC) initiative but this has so far fallen flat on it's face.
That's what they said about Windows Mobile for the first few years...
From a production perspective, supporting a different codebase is expensive, and a logistics headache. Windows XP Embedded (and now Vista Embedded) is really beginning to come into its own, especially now that there are so many ultra-low voltage x86 chip variants.
Windows CE (the basis for Smartphone and PPC) made sense in the late-90s and early noughties because small-size hardware was low-powered and didn't have enough wellie to run XP. That's not true any more. Also, 2000's real-time thread handling sucked, whereas XP/Vista has a bunch of new modes that work pretty well and can accomplish a good imitation of a real-time OS.
In the long-run, this won't mean very much change for end-users. Because it's based on a subset of the older Win32 API, Windows CE applications can be easily re-compiled to run under XP/Vista. Also, the PPC/Mobile/SmartPhone functionality is enabled as a set of dependency templates, so as you mention, deploying an emulator layer that's simply a bag of interrupt handlers is pretty trivial. Most CE apps spend most of their time executing Win32 calls anyway, and these are abstracted from the hardware, so performance on a newer x86 clone chip would be as fast if not faster.
In the global scheme of things, MS has around 6% of the smartphone market (compared to Symbian with 73% and Palm at 2%). Therefore, switching over is not a huge headache, especially considering growth rates. Within a year or two, the older CE-based devices would be about as ubiquitous as Windows ME machines.
it may not make sense from a user perspective, but the extraordinary economies of scale available in the x86 market is driving this change. The same economies forced Apple to switch from Motorola/IBM to x86 - do you really think smaller companies with less budgets (such as HTC) would resist this for long?