Philip Johnson makes some back-of-the-envelope calculations, very conservatively assuming that all OA journals are author-pay (not true) and all author fees for publishing are born by the Universities (not true) and concludes that even with such harsh handicapping, universities that switched to OA-only policies would immediately save substantial amounts of money (also check the excellent comments on that post):
Subscription costs would obviously be nil for an open access journal: we are all free to access the content of an open access journal via the internet, with no restrictions on who can read the content. In contrast, the author would pay to publish the article. This is perhaps the biggest resistance from scientists (and I'm sure the situation would be similar in the arts, or law, or what have you) to the open access movement, many feeling they don't have enough funding for students or experimental equipment as is, and couldn't possibly afford to pay to publish as well. I can appreciate this argument, though some progress is being made as you can specifically request funding to cover open access publication charges from some of the granting agencies.
(Also, let's be honest, the current situation of paying for page charges and to have colour figures means the author is already paying to publish, and sometimes non-trivial amounts.)
Let's further assume that the economies of scale would kick in if universities around the world decided to embrace this philosophy. This should lead to an overall lowering of the publication costs, all the while bringing access to academic literature to everyone with an internet connection. It is also easy to imagine the costs being even lower, as the collaborative nature of academic work means many papers now have authors from multiple institutions, all of whom could share in the cost of publishing. (Determining the rules for who-pays-what would be tricky, but should be doable.)
There are probably some key issues I'm missing here (the most obvious one which I've even mentioned is that we need open access journals to publish in!), and my idea is prefaced on the assumption that universities are the significant driving force in the academic literature game, but I think the take-home message is reasonably clear, at least using the University of Toronto numbers: we could already afford going entirely open access.
I certainly wouldn't feel bad if Elsevier and their ilk went out of business given the exorbitant increase in subscription costs and the non-obvious reasoning why, and I'm sure the societies could come to embrace the open access movement, which would bring the majority of high quality journals into the fold.
Peter Suber adds some useful links and calculations, including:
# Since November 2006 several studies have give us newer data on the the ratio of no-fee OA journals to fee-based OA journals. The best current estimates are that 67% of the journals listed in the DOAJ charge no publication fees, and 83% of OA journals from society publishers charge no publication fees. Clearly if Johnson zeroed out the publication fees for two-thirds of the articles published by Toronto faculty, the projected saving would rise significantly.
# The bargain would be even more compelling if we could subtract the publication fees paid by funding agencies rather than universities. Unfortunately there are no studies yet estimating that number.