hardly anyone would want to mine on GPU's anymore, it would cost way too much! Diff , therefore network hashrate, would drop significantly to make up for the reward halving.
Well, those who mine on GPUs and pay for electricity average rates or higher would mostly be forced to drop out.
When they drop out, the difficulty would drop.
That lower difficulty will make it so others who pay a lower rate (e.g., up to $0.07 per kWh) for electricity can continue mining at a profit after the halving event.
When conditions like that exist, BTC becomes more vulnerable to a 51% attack.
The reason we are currently at 24 Thash/s is not because that is the minimal level necessary to protect the Bitcoin network from a 51% attack, it is because with bitcoins issued at the rate of nearly $3 million USD worth each month, 25 Thash/s was the level of hardware acquisition that occurred from miners competing.
Is the bitcoin network just as protected from a 51% attack with a total of 12 Thash/s? Probably.
Since ASICs are 100x more power effeciency, this leaves room for a substantial increase in the amount of miners being able to run this type of hardware(50x considering reward halving) without hitting a power consumption ceiling .. meaning still being quite profitable.
Yes, the power consumption limits that were the upper ceiling for "hobbyist" and relatively "small-time" miners is gone with ASICs.
(hopefully that wont lead to a 51% attack, in the wrong hands it could =( ).
Economically it doesn't make sense for someone to invest the millions of dollars in ASIC hardware to harm bitcoin or to double spend (which will harm bitcoin). Not all actions are for economic reasons. Having many ASICs distributed among many miners is the countering force. The hash rate isn't busting out anywhere and at least two manufacturers now are claiming "real soon".
We'll know how this all pans out "real soon" I guess.