I should've put something up about this earlier, since I've known about it for a while now, and am still considering using it as a platform for my project.
Its currently free, though its in "Beta" (they call it Community Technology Preview, or CTP). I've heard it will be free for the forseeable future. Jesse is right, it is Microsoft's answer to all of the Linux bots, however its mission is much greater than that. Microsoft is also making a push into the embedded industries, and this is part of it. Further, industry analysts claim robotics is on the move (both hobbyist and industrial), and just as Mr. Gates was in the right place at the right time with MS DOS, he's hoping to jump in on the ground floor of this opportunity.
The question you should be asking is what does it offer you, the hobbyist (or more appopriately will offer you).
First of all its targeted at applications in which a full computer is tied to the hardware. If this doesn't describe you (e.g. you build autonomous microcontroller based bots), then it won't help you much in implementation. However, any sort of device may be controlled with it, so if you are building a complex array of bots, some with full computers on board and one or two without, you can, through a little work, tie them all into the system. But for those bots that are carrying your old laptop arround, Robotics Studio might be for you. It uses a service oriented architecture, which, coincidentally enough, uses web services for inter-process communication. This is extremely scalable and easy to use, though there is a slight performance penalty. Further, you can develop teleoperated bots with ease, and most of the communications plumbing is already there for you to use.
You might now be thinking this is useless to you, but read on. What I consider the best part of Robotics Studio is its simulation system, complete with a physics engine and graphics. Thats right, you can build up your bot, drop in pre-made sensor objects (including things you probably can't afford, like a SICK laser rangefinder, Sonar, encoders, etc.), write your code, create some obstacles and terrain, and see how it runs. This is EXTREMELY promising, as you can design with almost NO capital investment. While the simulation side of it works, its still a work in progress, so once again keep an eye on it but you may want to wait a few more months before really spending you time on it (unless you have nothing better to do). One of the main things still missing is its abilty to simulate noise in the sensors.
Microsoft's vision is really wonderful, and what they can do with this is astounding. If they were to add in a bunch of computer vision processing libraries, drivers for all kinds of devices (and manufacturers may even be developing drivers for their products, or better yet designing their products for Robotics Studio), a REALLY realistic simulation system, trajectors calculators, etc. this could become an EXTREMELY powerful system. But for now its still really early to tell where they will take it. Who knows they might end up dropping it like they have many other niche projects.