Ah, but one crucial fact you neg nellies are missing is that it depends how many computers you throw at it. If it would take a 1000 years on one PC, then it would take 1 year on a 1000. Especailly if using the earlier distributed idea where ranges are assigned. Yes, it may not be economically feasible to pay for that electricity versus just buying bitcoins, but it would make solving the problem quicker.
I already included all the computers existing today, see:
...
Simple crude calculation:
2
160 bitcoin addresses
2 billion computers worldwide
13 million addresses checked per day on a computer
22
160 / ( 2e9 * 3e6 * 365 ) = 667e27 years
...
2 billion that is.
Of course it is arguable on how many computers to add each year, and about how fast they are going to get.
But even if you let is grow in a few orders of size, the point of my explanation stays the same. It is just too hard to do.