Let us assume that google’s basic philosophy is to make information free (see for example http://www.google.com/corporate/tenthings.html) This way they can increase the value of the internet and the value of their search technology.
What can Google do to help the problem of the inflation of academic journals? The fundamental issue is the disparity between the ease of communicating information (scientific information which is, by community consensus, free) and the relatively difficult implementation of the nature human desire to make money by selling access to that information. The basic economic model is that taxpayers are the primary support for academic research at a US class I or a European university (tuition actually only fund a minority share of the cost of running an academic institution). On the other hand, taxpayers also pay for the cost of these journals to these academic institutions. If you, a typical taxpayer, enjoy paying twice or getting double-billed, I have two bridges in Brooklyn I’d like sell you.
In case the typical taxpayer thinks this is an issue they can hide from, I warn you, some of these journals are not cheap (eg, some are on the order of a few thousand dollars per issue). Moreover, some publishers, such as Reed Elsevier not only charge a couple of grand per issue but also publishes bogus journals (see for example (http://laikaspoetnik.wordpress.com/2009/05/08/mercks-ghostwriters-haunted-papers-and-fake-elsevier-journals/) and pays for fake positive reviews (http://golem.ph.utexas.edu/category/2009/07/elsevier_pays_for_favorable_bo.html).
The solution to this dilemma is rather simple, Taxpayers must insist that their research dollars only fund research which is available for free. In other words, the results of the their funded research much be publicly available. This goes for papers and software supported by their funding.
As a matter of complete disclosure, some institutions now require that their faculty post their research papers on free web servers (for example MIT), and some funding agencies require taht thier grantees make available their grantees research papers available on such servers (for example, NIH in the US).
Can Google play a role here? I think so.