For many years, people who generally lean pro-free-market and small-government have argued that when the government does things, it should usually do them through private entities. Don't have the government run utilities; have them be run by private companies. Don't fund solely government-run schools; even given a public commitment to government-funded education, it's better to have much (in the view of some, all) of that education be conducted by private schools. If you're going to pay for housing for the poor, do it through vouchers, not through government-run housing projects.
Moreover, you generally shouldn't burden the private organizations with common-carrier-like "you must serve everyone" requirements, at least unless they're monopolies. Better leave landlords and educators, for instance, free (or mostly free) to decide whom to let in, and on balance their economic self-interest will generally yield optimal results. (I set aside here the debates about bans on discrimination based on race, religion, sex, and so on; the common carrier model would ban all discrimination, or all discrimination found to be "unreasonable" by some government decisionmaker, and not just discrimination based on certain proscribed categories.)
My sense is that this is indeed generally a very good policy: Once you decide that the government should be funding or assisting some services, you should nonetheless prefer that they be privately provided (better yet, competitively provided), of course with some exceptions. There are costs to this approach, but they are less than the costs of inefficient government operations. "The public end may be as well or better served through an agency of private enterprise than through a department of government." Berman v. Parker, 348 U.S. 26, 33-34 (1954).
The funny thing is that, in Kelo v. City of New London, it is the (mostly liberal) majority's test that would give the government flexibility to serve public goals by taking property and selling it to private parties, when the government thinks the private parties will be better positioned to provide the public benefit. And it is the conservative dissenters' test that would give the government a strong incentive to own and operate various enterprises itself, or insist that whoever owns and operates them labor under the burdens of being a "common carrier."
Under the dissenters' view, if the City of New London wants to take property to run a shopping mall, which would presumably provide more jobs and government revenue, it's free to do so. But if it wants to take property and resell it to a private shopping mall owner, it may not. True, the latter solution isn't the pure free market: The mall owner would be getting a government benefit in the form of property taken from the original owners (albeit with compensation), just as school choice programs get a government benefit in the form of money taken from taxpayers. Still, it seems better than the City running retail stores — yet the dissenters' approach would give the City an incentive to do that, rather than lining up more efficient private businesses to do it.
Now there are certainly other arguments in the dissents' favor. Perhaps Justice Thomas is right that the original meaning of the "public use" requirement was to mandate that the property be owned by the public or by a common carrier, and that we should therefore insist on this meaning. Or perhaps Justice O'Connor is right that without some such requirement, the government would have too much power to merely take property from one person to another, with no real public benefit. Or one could argue that as between (1) no takings, (2) takings of property to be owned and operated by the government, and (3) takings of property to be owned and operated by private parties, but for a public benefit, #3 is indeed more efficient than #2, but #1 is most efficient of all — and that we therefore should allow only #2, precisely because its inefficiencies will provide a natural deterrence to takings, and push the government more often into the best position, #1. I'm certainly sympathetic to finding some constitutional limit to government power here.
Nonetheless, if we're evaluating the practical benefits of either approach, we might want to be skeptical of an approach that favors government ownership or operation of publicly useful enterprises over private (albeit government-assisted) ownership or operation.
UPDATE: I should stress, by the way, that my point isn't just that a "government ownership and operation is OK, private ownership and operation is not" rule creates perverse incentives. The question is also whether a rule that disfavors privatized but government-assisted enterprises compared to purely government-run enterprises makes sense, given that we generally think privatized enterprises are better.