You don't take in account things like:
- you will be able to install more kH/s per same power line
- it will be less noisy, possible no noise at all if you don't overclock
- it will take less room. or, the other way, you will able to pack more in the same space
- it will generate less heat
I don't have any hard stats but I'd bet that if you took away all the large miners that the average miner probably puts down less than 4,000 khash (about 8 x 7950) which isn't even close to maxing out their power line. I think the average person's wallet is much more of a limiting factor (so again, price trumps power use).
Noise is really only a problem for people mining from their desktops. Anyone with a dedicated mining operation probably isn't so concerned about the noise that they'd pay 2x the cost for the same hash power just to make things a little quieter.
Again, the average miner probably has two or fewer dedicated mining rigs - saving a tiny amount of space again probably not worth paying a significant premium for your hashing power.
Less heat is probably the most important of those points. However this has varying impact and is obviously more important to people in hot climates whereas people in colder climates can just pipe in external air for free cooling.
I'm not saying people shouldn't develop FPGAs here - just that the price needs to come down to be much closer to the GPU price per hash since I doubt most people would be willing to pay a large premium to save some power noise and heat. Due to the ASIC arms race in the SHA-256 world it seems like people over here on the SCRYPT side are itching for some "miracle cure" but so far all efforts seem to have confirmed that it is much harder/costlier to build an FPGA/ASIC for SCRYPT mining.