Well, if you're willing to ignore the existing base of bitcoin already mined (and even in some cases the sunk cost of buying hardware that can be used to mine them) go ahead and do your calculation based solely on the price per unit energy as a starting point.
If you use this calculator[1] you can see that if your 50GH/s array spends 300w which you had to pay for, you are burning about $0.48/day at the current price, assuming all of that from before. This is not top-of-the-line, but it is some relatively modern ASIC equipment from BFL. Previous generation miners, 5 and 10GH/s "Jalapenos" from the 65nm fab. They are supposedly rolling out 28nm "Monarch" now which are at least 10x as powerful and certainly more than just marginally better at power efficiency.
If you look at this chart[2] you can see an estimate of the next difficulty and adjust your calculations for next week, it becomes pretty clear that more people are turning on their Gigahashes than turning them off, even if that was only marginally the case two weeks ago. You have to imagine those people are mostly those with newer equipment.
I don't have a graph of price overlaid with difficulty, but it can't be too hard to find one. A few months ago it was possible to scale them and demonstrate a relatively good fit where price and difficulty were roughly correlated, but since difficulty shows no signs of falling yet, I suspect that fit is not so good anymore.
[1]: http://www.bitcoinx.com/profit/
[2]: https://bitcoinwisdom.com/bitcoin/difficulty