The original report from Bloomberg focussed its attention on tablets and phones - in today's tablet-crazy tech media landscape, that's hardly surprising. Smartphones and tablets are very hip right now, and as such, it's what people tend to focus on.
However, it makes zero sense for Windows NT to be ported to ARM so it can power phones and tablets. Microsoft already has the very modern and capable Windows CE 6.0 to do so; it's the operating system, powering the Zune OS and Windows Phone 7. The latter, of course, is what most likely will become Microsoft's tablet operating system - it's designed for touch, after all.
It's an additional story about this subject which gives a very important clue as to why Windows on ARM would make sense. The Wall Street Journal states that the release of such an ARM-powered Windows NT is about two years away - which left consumer-flashy-gadgets-zomg-sparkles oriented folk like John Gruber confused.
It doesn't leave us OSNewsians - the more nitty-gritty type of geeks - confused, of course. About two years from now? That would coincide with the release of Windows 8, which makes sense when you think about it; that way, the developers can take ARM into account when developing Windows 8. If they were to port the current Windows 7 codebase alongside the development of Windows 8, the ARM team would have to do the porting all over again when Windows 8 hits the streets. This could be avoided by making the porting process to ARM a major bullet point in the Windows 8 development process.
But why would you want Windows NT on ARM? Simple: servers. ARM is currently developing very cool multicore chips for servers (virtualisation built-in), which could, potentially, revolutionise the server world. I'd much rather have a few passively cooled and very efficient ARM processors in my server than overpowered, energy-sucking, incredibly hot x86 chips which need complicated cooling systems to operate.
On top of that, Windows servers run a much narrower array of software than desktops, most of which is either owned by Microsoft (and thus easily ported by Microsoft itself), or open source (and thus most likely already available for ARM anyway). As such, the application problem doesn't exist.
I'm fairly sure that if this ARM-Windows rumour is indeed true - and according to additional sources from Ars, it is - then we're most likely talking about Windows Server for ARM, most likely as complete packages. I can also imagine Windows Home Server using such an ARM version of Windows Server, since it could mean cheaper, smaller, more efficient devices.