The blog Self-Evident (via Felix Salmon--I'm basically outsourcing this blog to him today) has a really helpful example of how the Geithner plan might work. Basically, the key link here is the no-lose loan the FDIC provides to the private investor. Thanks to the loan, the investor can afford to substantially overpay for the bank's bad assets and still make a profit. The FDIC then ends up transfering a little bit of money to the investor and a lot of money to the bank ($100 and $3,400, respectively, in this example--on a purchase price of $8,400 for the bad assets).
This the reason the Geithner plan turns Paul Krugman's stomach (and makes people like Matt Yglesias a little queasy). And rightly so. Under this example, the Geithner plan ends up costing the taxpayers more than if the government just transferred money directly to the banks in exchange for assets of much lower value. (Because we have to pay off investors, too.)
But I actually find it slightly reassuring. My earlier concern was that the plan wouldn't work at all--because, even with the generous government loans, investors wouldn't be willing to pay a price the banks would accept. Now I see how it could work (at least on its own terms). We just have to deal with the fact that it's inefficient and rewards people who don't particularly deserve rewards. That's not great, but better than outright failure, I think.
Update: In fairness, the Geithner plan does require private investors to shoulder some risk. It's just not very much because they're putting up so little equity, and the equity is the only thing they can lose. (That's why I referred to the FDIC loan as "no-lose" above. The technical term is "non-recourse.")