Anyone have any rough numbers on what this would cost Joe Schmo to buy?
There's a huge variety. An individual subscription to Nature is $200/year[0], but they often charge $30 for electronic access to individual articles.
An institutional subscription to the Astrophysical Journal is $2000/year[1], though if you're a member of a professional society, you can buy an individual, electronic subscription for $25[e.g., 2]. As a point of comparison within astronomy, a similarly prestigious journal, the Monthly Notices of the Royal Astronomical Society institutional subscription is $9400/year[in the csv link in 3].
[0] http://www.nature.com/ecommerce/subscribe.action?productId=n...
[1] https://iopscience.iop.org/journal/0004-637X/page/Pricing%20...
[2] https://aas.org/individual-membership/dues-subscription-fees
[3] http://www.oxfordjournals.org/en/access-purchase/pricing-and...
That's necessarily true. It's true that universities pay for journal subscriptions, but I guess a significant portion (or even all) of that expense is covered by the overheads universities charge grants. Given that a significant number of grants are taxpayer-paid, it's likely that tax payers are already paying for the journal subscriptions. So making them open access will shift the costs around, but it isn't immediately obvious that it will cost the taxpayers more than it currently does.