Yes the problem is not the lower profits of the average miners, it's the high centralication of hashing power. And if some universities or whoevers fpga cluster can put this much hashing power to the network, some 3 letter agencies can put bitcoin down with the snap of a finger (ok, not REALLY putting it down but performing the >50% hashing power attack, thats not much better).
What bitcoin needs to get strong enough is that the average miner is using the same maximum effective technology which any entity could bring up. Therefore miners must switch to fpga/asic mining or the algorithm needs to be extended to be no more highly parallel computable (but I don't know if that's even possible as every software basically is faster if implemented in hardware).
It would be trivial for them to do as well. Once you get >%50 of the computational power you can basically counterfeit transactions and as long as > 50% of the network validates it, it's valid to the protocol.
No, no, no. This is completely 100% WRONG. You don't seem to understand how any of this works at all.
With more than 50% of the hashing power, you can hide a double spend
for a few minutes, and only from funds that you have legitimately. You can not counterfeit anything.
I stand corrected then, I haven't really put time into researching attack vectors yet.