The thing is that even if they create a device that generates 100 MH/s for $600 but consumes lets say 25W, its going to outprice the GPU's anyway, because long time the investment will be worth it and the GPU wont be able to stay profitable at 100Mh/s at 25W. Thats a MH/J of 4, and the most efficient GPU's can only do slightly above 2. As people buy more and more of this ASIC's the difficulty go up and at some point the gpu's will generate bitcoins at a rate that wont pay for the electricity they consume.
It would not. The investments needed for the design and starting up fab process are enormous. The hardware unit cost may be $600, but that would not include the vast investments thrown in for R&D and setting up the production facility. It is a long term investment and I find it kinda weird how people decide to go after it as risks associated with bitcoin are rather high in long term. The bitcoin prices may collapse for a prolonged period rendering your investments useless and that would be a grand fuck up, millions of dollars thrown in the air. Unlike GPUs, noone is going to buy that hardware because it can do nothing but 2xsha256 calculations.