Yes they were flawed, it was only a back-of-the-envelope calculation. I didn't think the maintenance cost and building cost would be so expensive compared to the GPUs themselves, but you probably know better.
I think you may have misunderstood my line of attack though. Once somebody has more computing power than everybody else put together, they should go about 1,000 blocks back and try to build a new chain building off that, in secret. (This would take a week.) However their chain will not include any of the transactions from the last 1,000 legitimate blocks. Once it becomes longer than the existing chain, they publish it and instantaneously a whole week of transactions un-happen, i.e. they are reversed and the money (including mining fees) returns to the hands of the original owner, whereas the 50,000 BTC legitimately mined disappear, and the new 50,000 BTC generated are owned by one bitcoin address, the attacker's.
This would cause a massive panic and the price of BTC would crash on the exchanges.
They are then pretty much done in my opinion. They can turn off their datacentre and as long as everybody knows that the datacentre is there and can be powered up again at any time, people will not want to use bitcoins or any similar system.
Of course ideally all the miners would rally together and increase the network power, so much so that our attacker is forced to order another batch of GPUs or give up. But that's a little too idealistic for me.
So in this scenario, they will have spent a massive amount of money on computing power. And it has to be FAR greater than the current combined network (to create fake week-old transactions, and to catch up to present, and then finally to surpass the present). And then, according to your scenario, they will create their own 50,000 BTC (which, at current market prices is itself worth almost $1M), and then throw away that $1M, by driving the price of BTC to zero.
Who would do this? A government, I suppose. But you wouldn't do it to make money!