There's been quite a lot of talk over the past few days about an project MS is running called Midori that may, or may not, be the makings of a new legacy free OS.
The prospect of an entirely new OS is quite an exciting one, the world has moved a long way since the WinNT code base that XP and Vista are based around was first written. It's probably fair to say that to some extent today's hardware is held back by the legacy components that the OS are forced to include.
That said, as Ed Bott talks about on his blog, such a change is not something to be taken likely. There are A LOT of existing Windows users and applications that would need to be supported in any change. As I believe MS discovered with Vista, the cost of migrating applications to a new platform is a huge obstacle to business migration.
Or is it...
Virtualisation technology is advancing hugely, if not in overall functionality then in the ease with which it can be utilised. Microsoft's Hyper-V is part of the basic server OS, and as I understand it will be core to Windows 7. Application virtualisation is maturing nicely as a tool and is, I believe, about to become mainstream. Virtual Desktop Infrastructure (VDI) is now something that enterprises can realistically consider, again Windows 7 will drive this even further forward.
With all of these components in place, the release of a new OS should be much simpler than ever before. Application compatibility, the big killer of migrations, should slowly become less of a challenge. The toolset needed to run applications across multiple OS's on the one system is pretty much all there. Of course those apps won't benefit from whatever advances Midori might bring, but as cover for a transition period that's not so important. As long as they work, both independently and with each other, that's a huge problem out of the way.
I agree with Ed that any new OS is likely to live in parallel to a standard Windows product for some time, but I do think the migration will be quicker than we've seen before.