It was the always perceptive William Buiter who initially advocated a tax on such banks.
When size creates externalities, do what you would do with any negative externality: tax it. The other way to limit size is to tax size. This can be done through capital requirements that are progressive in the size of the business (as measured by value added, the size of the balance sheet or some other metric). Such measures for preventing the New Darwinism of the survival of the fattest and the politically best connected should be distinguished from regulatory interventions based on the narrow leverage ratio aimed at regulating risk (regardless of size, except for a de minimis lower limit).
And now the sheer chicanery associated with on the spectacular second quarter profits of Goldman Sachs and its massive executive compensation payments has left even the ultra-conservative Wall Street Journal proposing a "bailout tax",
"One policy response to the incentives created by last fall's bailout is simply to restrict the proprietary trading done by the subsidiaries of bank holding companies that enjoy both FDIC deposit insurance and an implicit government subsidy on their cost of capital. This is what Paul Volcker proposed, only to be overruled by Tim Geithner and Larry Summers. Another answer would be an FDIC-style bailout tax, perhaps tied to leverage ratios, for those in the too-big-to-fail camp. Developing a template to facilitate the seizure and orderly winding down of failing financial giants is also an essential element of whatever reform Congress cooks up."
Felix Salmon too supports the idea,
"There really is a public good to be served in taxing what you want less of — which is too-big-to-fail banks making outsize bets with other people’s money (backstopped by the taxpayer, of course) and then paying themselves billions of dollars in bonuses. The WSJ’s bailout tax idea is a good one — especially if it rose in line with a financial institution’s balance sheet, and gave those institutions a serious incentive to shrink. If you can’t legislate a hard cap on assets, then you can at least provide some gentle encouragement to get smaller rather than bigger."
In fact, in order to limit the "too big to fail" hazard, Buiter even proposes that only governments do certain banking activities
"Governments everywhere should be focusing on breaking up banks and keeping them small. If some banking activity (or indeed any other economic activity) is deemed to have a minimum optimum scale that is makes it too large to fail, it should be publicly owned. Small is beautiful for banks."
And about keeping banks small he writes,
"There is no economic reason for large banks. Therefore banks should be kept small. An obvious mechanism (apart from aggressive anti-trust policy) is to tax bank size. One way to do this is through making regulatory capital requirements increasing in the size of the bank’s activities. For instance, tier one capital as a share of (unweighted) assets could be made an increasing function of the value of the assets. Gary Becker has made a similar proposal."
James Surowiecki traces the reasons for bank size being persistently being big - account switching costs, government subsidies and guarantees (moral hazard which keeps consumers with the larger banks), and the banks’ market power (which increases the market credibility of the advice/service offered). See also James Kwak here.
Finally the Obama administration is apparently considering a tax on banks that pose systemic risks so as to recover the atleast $120 billion it spent to bail out the financial system. Under consideration is a tax based on the size and riskiness of an institution’s loans and other financial holdings, or a tax on profits. However, this would be different from a global transactions tax, as suggested by the EU, or a windfall tax on bonuses. See Simon Johnson and this debate on bank tax.