There is an article in mentioning a curious research report from Gartner implying a redesign of the Windows Vista code base, around virtualization. Huh? To quote from the article:

Instead, the research firm predicts, Microsoft will be forced to migrate Windows to a modular architecture tied together through hardware-supported virtualisation. "The current, integrated architecture of Microsoft Windows is unsustainable - for enterprises and for Microsoft," wrote Gartner analysts Brian Gammage, Michael Silver and David Mitchell Smith.

The problem is that the operating system's increasing complexity is making it ever more difficult for enterprises to implement migrations, and impossible for Microsoft to release regular updates. This, in turn, stands in the way of Microsoft's efforts to push companies to subscription licensing.

There is a basic confusion going on here about architectural choices vs. release frequency.

On the last point (release frequency), Microsoft is committed to not repeat the mistake of delaying a Windows version that much (or any other Windows version) and instead, impose a sense of predictable rythm in releasing incremental, predictable and evolutionary improvements in future versions of Windows. Gaining this sense of rythm can be done even with an extremely complex code base such as Windows, and the proof is the predictable, stable releases of Windows Server versions already exist: Windows 2000 (beginning of 2000), then Windows Server 2003 RTM (april 2003), then Windows Server 2003 R2 (RTM November 2005). 

That said, if we are talking about Operating Systems in general, the architectural idea of achieving modularity through virtualization is a very interesting idea. I would like to point out to a cool research project called Singularity which achieves exactly that goal. If you want to see how operating systems will look like, say, 20 years from now, you can start with Singularity. One note, though: Singularity achieves isolation through pure software mechanisms. Hardware protection (like Hypervisor support from the AMD processors) is also possible to be used in conjunction with software isolation, although this is not a direct goal of the Singularity project.