Well, according to the article's "rumors," it'd take a machine being used 25,000-32,000 or so times before it manages to pay for itself, at $3 per book. And at an estimated 5 minutes per book, that means that the machine would have to be constantly printing for around a quarter-year non-stop, without any delays, break-downs, or merely people not using it (such as the building being closed) in order to make that.
As a result, I have my doubts if it'd actually turn a profit; it's a very nice idea to think of, though; many of us simply don't like reading off of a screen, or reading off of e-paper; physical books are more comfortable and easier to read.
Of course, I'd also question their definition of "library quality;" given the brief time to print it all, I wonder what form of inks are being used, as well as the quality of the paper involved. Perhaps mostly, I question what durability the binding will be; since all the books are "printed on-demand," that means the machine is going to have to automate and adjust to binding a different number of pages each time, which could yield problems. (normal book printing involves precise, human-adjusted setting of the binding machines, that won't vary since it'll print a LOT of just one book)