EETimes has a breakdown
and shows you the parts that it takes to create one and estimates the price of all of them to be about $56. The PrimeSense image processor is the major brains behind the device, as have been confirmed a while ago.
There are four microphones that help make Kinect
one of the better voice recognition devices out there. I was surprised at how well it picked up my voice and that's the reason why. You get some looks at the sensor board and main board in the article as well as breakdown of the parts.
Yes, it might only cost $56 in parts but there's all the other costs associated with the product that Microsoft has to make up for such as R&D, marketing, and the partnership cost with PrimeSense.
If you really want to see it broken down though, iFixIt is a great place
to see the camera being taken apart.