In general a professor is an entrepreneur. They are running a small business where they sell their product ("knowledge") to customers (business and industry) by means of sales and marketing (conferences and papers). Grad students are employees of this business, and of course the prof's incentive is to keep the good ones and fire (graduate) the bad ones, thus the reward for the student working hard is more hard work.
1) I am evaluated on the progress of my students. While bad ones will actually be fired, good ones are more useful to me if they go off and get postdocs etc. Some of the most powerful "businesses" in the field are created by the network of people who all trained with the same PI and then went on to found their own labs.
2) A burnt out, trapped graduate student is not actually one that produces good work, in my experience.
3) Several grants, which are how I fund my lab, will look very poorly on trainees making no progress. There's very little exploitation that's worth a program officer at the NSF wondering if I'm worth there time.
Are there bad, exploitative PIs? Absolutely. But these people are assholes, not the only logical outcome of the system.
But no good student who approaches the grad school process in a sane manner is going to work for a professor whose students stay there for 12 years and don't graduate. A common thing for deciding if you want to work with someone is seeing how long their students took to graduate, where they ended up, emailing them, etc.
