State actors would need to aquire mining hardware in order to make a 51% attack. Several million dollars worth of it. Trying to do so would have obvious effects on the small bitcoin mining equipment market, and unless the government turned all their machines on at once, they would spend the time they ramp up to a 51% attack actually strengthening the network, and causing other miners to also buy more equipment to keep up with the new difficulty.

This is certainly a possibility, if someone in the government recognizes the potential threat bitcoin offers to the government money monopoly. However, anyone smart enough to recognize the threat is also incentiveized to not tell the government about the threat, or even downplay the threat, while investing in bitcoin themselves, as if the threat materialized they will become significantly wealthier than they would through a government job.

As for the inefficient argument, the operating costs of bitcoin (mining) should be compared to the operating costs of more traditional money transfer services of similar size. Think armored cars, vaults, double book accounting, and other such things that a 3-4 billion dollar firm engaged in money transfer would need, and compare those costs to the costs of blockchain mining.