The idea behind print-on-demand is to lower all those intermediary costs that keep nudging book prices higher and higher. Warehouse storage costs, for instance, are eliminated altogether. A horrible example: a few years ago I signed a contract with International Polygonics for a trade edition of all three of the opera mysteries. A Cadenza for Caruso was printed and stored in a warehouse -- when IP went bankrupt. And since IP couldn't pay the warehouse bill, the warehouse owner wouldn't release the books and so they were never shipped. The entire print run was destroyed. Something like that can't happen with print-on-demand.
But the biggest chunk of money goes to the distributor. Ingram takes 55% of the cover price for every book it places in a store. But since Ingram isn't doing the actual physical distributing of print-on-demand books, they're simply charging a flat fee for a listing in their catalogue -- much less than their percentage rate would come to. The ideal situation would be to eliminate the distributor altogether (since Ingram makes more from a book than the author and publisher combined), but all publishing would have to go to print-on-demand before that could happen.
The economics of this business are ridiculous.