They need a so high quantity of BTC that it would result in a 51% attack before they reach in the necessary infinite money
What are you talking about? A 51% attack has to do with mining power, not with the amount of BTCs you have.
Niko reasoned in another thread that a 51% attack would probably cost 10-50 million to accomplish at the moment (I look up the link later), so in my mind, at the moment, it is not cheap but quite feasible for a goverment or a large corporation to perform such a task.
Gross overestimate. Would you buy shares in BFL at a company valuation of $10-50 million dollars? If not, then why would the government have to pay so much?
And no. The "txn fee attack" won't happen as long as there is a block size limit. That is what central planning rules are for. As in a PoS attack, you would attack by showering your enemies with money. It is not a sensible strategy.
What does the value of BFL have to do with how much it would cost a government to attempt a 51% attack? It would easily cost much more for them to do it first, they'd start soliciting bids from established vendors like Texas Instruments for ASIC hardware. They'd also solicit bids from the likes of Cray and begin going through surplus property to see what supercomputing resources they have that aren't already alllocated.
After going through the investigation process they will decide to assemble mining machines running some sweet yet horribly inefficient NVIDIA Tesla's. Those would run anywhere from a reasonable $2,200 for the C2075 on the market to $4,000 for the K20. Multiply that by the number of cards it would take to reach 51.5% of the network when the bid was originally proposed several months earlier, and then they take delivery of the system several years later when someone has finally shipped ASICs. Before long addendums to the original bid have the government ordering still more Teslas and trying to compete with ASICs (assuming ASICs shipp sometime before 2015).
Government could easily burn several hundred million trying to pull off a 51% attack without succeeding, as long as the task is assigned to the Pentagon.