Well, if you haven't been paying attention to the tech demos, you can do a lot more than just arm gestures.
You could start off with it as a webcam, or potentially a depth-enhanced webcam (2.5D or 3D panel capable, not holographic, but enough so you can sense a scene). The hardware on it, especially when coupled with the computing power of a PC, can build wireframes of not only limbs and appendages, but also facial recognition. They have done tech demos where it can (with the older lower-res version at a few meters away) detect most of the facial features and map them talking onto a virtual staged xbox avatar, coupled with the speech and expression. This could provide for lower-bandwidth remote conferencing versus a group video-call. People have used basic extrusion modeling to use the depth sensors to rotate around objects to build 3D-models. Can you think of a game where you could properly map/model yourself into the game? Or maybe a Little-Big-Planet style thing where you could add your own custom objects into the maps? Surely not quite hardcore there as a controller, but certainly a valid use.
Also there is all of the fun to be had with this for things like robotics, or even remote RC-car type things. Have a huge Tonka truck, a small laptop or tablet with data connection, and one of these? 3D (well 2.5D) remote viewing RC car (a bit heavy for a plane). The hackery that could be done with this is simply astound, especially with the increased resolution of depth field and shorter possible depth sense accuracy.
Lastly there is also the standard possible uses for this that already come up: Voice controls (these things have multiple microphones for positioning and noise cancelling), and simple touch-esque gestures. Yeah, those don't work as well when you have a mouse and keyboard in front of you, but when using them on a HTPC, it could see using that versus a full remote.
I agree with many others it doesn't warrant a full extra US$100 on top of the current price of US$150, that smells of new-device lust gouging. However, the device itself is pretty nifty, and there are lots of use-case scenarios. Maybe just not so much for _HARDCORE_ gamers (unless you want face-mapped smack-talk when you pwn noobs).