Windows 95 = Windows 4.0
Windows 98 = Windows 4.10
Windows ME = Windows 4.90 (almost 5.0, because they included some features from the NT line)
Windows 2000 = Windows (NT) 5.0 (this is where the NT line takes over)
Windows XP = Windows 5.1
Windows Vista = Windows 6.0
Windows 7 = Windows 6.1
Windows 8 = Windows 6.2 (I'm not even joking)
Windows 10 = presumably Windows 7.0
I completely expect Windows 10 to ship with NT 6.4. I think there is going to be simply a unification of current Windows API and other kernel factors throughout the rest of the Microsoft/Windows family, as opposed to any significant sweeping changes that would result in a new kernel version.
A new kernel version immediately after Windows 8 would also likely be a disaster. It was the reason the consumer-side launch of NT 5 (5.1, XP) was such a botched launch until significant improvements (service packs). It was the reason Vista had an uphill battle. They don't change the kernel version major number for the hell of it, it represents sweeping overhauls, and typically demands developers almost start over with overhaul their product.
Drivers are almost always the most obvious problem facing major kernel revisions, since new kernel stacks and the other API changes basically cause numerous stability issues when developers fail to create solid drivers early in the life of the new kernel. This gets addressed, and other unforeseen issues get patched up and improved upon, which is why the point releases have always been rather significant leaps in performance and stability over previous releases.
Windows 8 was a rather odd undertaking for Microsoft. They wanted to change the user interaction paradigm, but I think the goal was one step at a time, so that was the largest functional change (including additional APIs, though not rewriting the kernel stack or revising previous APIs) while they kept the major development side of the house in relative order for stability.
It was still botched, but in a relatively new way for Microsoft. It wasn't driver instability due to a new kernel, it was over-reaching and other missteps.
A new kernel now? They still have to save face from the previous OS release, which was, underneath it all, on a very stable platform. A new platform bound to introduce instabilities? Eeek.
Frankly, I don't expect an NT 7.0 for a long while. Considering Microsoft has apparently seen a way to unify development across mobile and integrated platforms into the current NT 6.x series (albeit entirely related to only the new APIs introduced in 6.2/6.3 - you won't see Photoshop suddenly appear on the Xbox platform), I'd say they have a very solid kernel that is keeping up with modern demands.
Microsoft's approach to kernel version numbers is an ancient modest one - only the most sweeping, significant changes cause a major version number. Almost all of the regular improvements are found in point releases.
This is very much different from Apple's approach with the Darwin kernel/system. Darwin started at 1.x for Mac OS X 10.0, and OS X 10.10 will ship with Darwin 14.0.
It's sort of like where Linux still stands, at v3.16.3, or where one of the base components of Apple's hybrid kernel, Mach, has remained for awhile, at v3.0.
