You just have to look at the amount of money they sunk into R&D in the last 15 years, which is public data. Nobody spends that sort of cash to simply develop simple iterative versions of established products (which Windows and Office have been ever since XP, really) or one random new gizmo (Kinect).
Have a look at the list of projects here http://research.microsoft.com/en-us/ and see how many are probably going to be profitable in the next couple of decades. It's actually pretty amazing how much research Microsoft fund that seems unrelated to any of their current product lines.
I think MS Research would be a pretty amazing place to work.
MSR really is an incredible place to work, but also impossibly hard to get in. It's akin to getting a fully tenured professorship at a top-tier computer science program. Back when I was in graduate school, MSR dominated every other computer science research department in the number of SIGGRAPH papers they published each year. And that's just in graphics.
If i remember correctly, Haskell is being developed in MS RD. At least they have hired the main Haskell dev, and I think he was involved in the F# dev work, including ideas from Haskell. IThis was commented on in a talk with the Erlang and Haskell devs.
I've never worked there but I know people who did. And anyone who follows MS Research can read between the lines. No direct experience here though so I'm the wrong person to write it up.
Nadella's contribution is to make it possible to ship this sort of thing. Shipping this under Ballmer would have been a miracle.