it really depends when you'd get it and where the difficulty is by then. if difficulty goes up by a factor of 10x, then that Singe SC at 40gh will be making the same money, roughly, as 4gh today. which isn't bad at all, and definitely profitable. now, if difficulty goes up 100x....
and that's the rub right now. we can guess and estimate for weeks on end (as we have done in various threads) what the difficulty will be in 6 months or more, but in the end it's still just a guess.
I'm currently estimating a medium-term increase in difficulty by a factor of 30. Target is around 70-75 million difficulty.
My comparison is currently that a BFL Single SC is hoped to deliver about 30 times as many hashes/sec per dollar as a typical GPU gives today. Some of the other ASIC ventures are hoping for roughly that same performance level, so it seems plausible.