This is one of the reasons I prefer proof-of-stake schemes, despite the complexity, low power consensus is preferable given roughly equivalent tamper resistance. State collusion with mining pools is a credible threat to proof-of-work schemes.
Miners will crank whatever handle they can crank to increase their payout, up to the point where they spend more marginal effort cranking the handle than the corresponding marginal payout is worth. In PoW (in bitcoin) that handle is SHA256. In Proof of Stake, it's more obscure, but they will find it, and they will crank it, and you'll just be back where you started.
I'm not convinced PoS is a practical improvement on PoW.
[1] C. Decker, Information Propagation in the Bitcoin Network: http://www.tik.ee.ethz.ch/file/49318d3f56c1d525aabf7fda78b23...
This would be fatal in case of a remote crash vulnerability, because an attacker could partition the network at will.
Or a nation/state could cut internet access.
Without real world energy, you lose byzantine fault tolerance.
Some people wonder if pure proof-of-stake is even possible. Ethereum is likely going with DPOS, yet to see a single solution to the nothing-at-stake problem.