In my experience, what happens is that a scanner is bought for a hospital and/or research dept, costing around 3million (c. $5 million). To claw this money back, the number of hours of function that the scanner will likely achieve in its lifetime is estimated, and this is how much you charge per hour for its use.

Certainly where I was doing my research, the scanner sat empty and unused for at least one quarter of the time. It seems that it's harder to get money to do the research you want (I was looking at about 200 [$300?) an hour) than it is to find a free slot.

Incidentally, lots of potentially really useful medical work could be done using techniques like this, but the cost is just plain prohibitive: a large scale study = 30+ participants probably taking up the scanner/resources for an hour each + staff time (that's for a radiologist to oversee it all + radiography staff, and the latter may have to spend quite a lot of time downloading and backing up images from the scanner) + image processing + post-processing ... you need a lot of people to make fMRI research happen.