This obviously assumes the attacker is interested in profits that can be extracted from the system. An attacker who is already wealthy, and has a greater interest in undermining the system than extracting profit from it, can trivially overwhelm the network by assembling processing power - especially if the attacker already has a stockpile of processing power.
Usually, the cost of destroying something is much cheaper than creating it. That's why terrorism can work. The cost of attacking is not that large compared to fear, destruction and cost of guarding. 9/11 proved it very well.
In case of Bitcoins, obtaining 50% of the network compute speed required for completely disrupting the Bitcoin network grows with Bitcoin size. When Bitcoin network compute speed crossed the fastest supercomputer Tianhe-1A, it is no longer that easy. Destroying Bitcoin network is roughly as costly as the Bitcoin economy size and if it grows, it will become even more costly. This is not asymmetric as in case of terrorism. It is still possible for national governments but it's no longer a matter switching on a small cluster.