Sorry but you cannot use one variable to decide when these machines shut off. Difficulty is important, but so is price. At my electric rates, each TH/s of Avalons running are only using $218 of electricity per month. My decision is very different at $50 btc than at $200 btc.
That is a good point but difficulty follows price so if you are the marginal miner (product of cost per kWh and rig efficiency) you are going to get squeezed out regardless. If price quadruples so will difficulty eventually.
I estimated some break even difficulties (electrical cost = value of BTC mined) in this thread:
https://bitcointalksearch.org/topic/break-even-difficulty-by-hardware-efficiency-power-cost-value-of-btc-281279Device Process Break Even Diff (mil)
-----------------------------------------------
GPU various 60
FPGA various 400
Avalon 110nm 2,400
ASICMiner 130nm 2,700
BFL (SC) 65nm 4,200
KNC 28nm 18,250
Bitfury 55nm 23,300
Hashfast * 28nm 25,100
Cointerra * 28nm 25,100
BFL (Monarch) * 28nm 25,100
* Indicates devices has not shipped and efficiency based on projected specs
The break even difficulty is based on $100 exchange rate and $0.10 per kWh. To adjust to other prices simply multiply or divide (i.e. for $200 exchange rate the break even difficulty would be doubled, for $0.20 per kWh power rate the break even difficulty would be halved).
Assuming price remains ~$200 the first interesting point is ~10B difficulty. At 10B difficulty even with $200 exchange rate and $0.05 per kWh Avalon and ASICMiner gear is negative operating margin. While miners might cling on for a while as difficulty moves higher and higher month after month even those "slowest" miners will get the picture. Burning $100 in cash to produce $50 in value isn't particularly "fun".
At around 8B difficulty BFL rigs in high cost areas (>=$0.10 per kWh) will be unprofitable but we may see some secondary sales (high cost miners selling to low cost miners). However by 20B or so no BFL rig will make sense without "free" power.
.