I'm going to move this note up to a top level thread so I can refer to it from various other posts. When thinking about mining, the key point to consider is the "value" of your mining asset. That value is set by mathematics. The only uncertainty in the equation is how rapidly the difficulty increases over the life of your investment in mining.
I have seen a number of online calculators trying to help you figure this out. All of them are wrong. The only way I have figured out how to get a good estimate of the future production of a mining asset is to model it in a spread sheet with the expected difficulty increase for each bump (increase in difficulty). I am happy to make this spreadsheet available to anyone who wants a good tool to calculate ROI. Just send me a PM with a real email address.
The key point here is for any difficulty increase projection, a miner has a well defined value in terms of BTC produced. One way the decrease in BTC value affects miners is each mining investment must be balanced against the BTC that can simply be purchased instead of the mining power. At $400 BTC price, a mining investment must be twice as efficient as a $800 BTC price. I see people saying things like the value of your mining rig will increase over time if BTC goes up. True for fiat value, but the BTC value of the rig is pretty much set the day you turn it on. If you can buy BTC for less than you can buy and operate the hashing power, then you should buy BTC instead of hashing power.
I load the spreadsheet with the current difficulty, expected difficulty increases and then it iterates to an answer with assumed difficulty increases. This result shows the payback in BTC of one terra-hash of mining power put into production on the date of the 9/25 bump. The spreadsheet assumes the specified difficulty increase occurs every 12 days. If you keep your miner (local or cloud) in production 100% of the time, you should be within a few one thousands of a BTC to these numbers. I put the end date when the miner is accumulating less than 1/1000 BTC every 12 days (bump period). Average difficulty increase has been about 13% for the last ten bumps.
What this table tells you is if you put a 1000 THS miner in production 9/25/2014 and difficulty increases are 13%, then it should produce roughly 1.510 BTC for you between now and 4/23/2016. Depending on your situation, you should certainly not pay more than 1.5 BTC for the miner (or in dollars, roughly $600 as of 9/25/2014). Instead, you can buy and hold the BTC and be at the same place today. Similarly, at 13%, a 480 GHS miner will produce 0.725 BTC in its productive life time.
If you think the difficulty increases will slow down to 10%, then your result for the 1 THS will be in between, but not more than 1.908 BTC/THS.
p.s. On thinking about this some more, there might be a way to formulate this as a set of equations in N unknowns. The knowns would be current difficulty, difficulty increase rate and hashing power. The unknown would be number of bumps until the mining power is "worthless" and the number of BTC produced. I think the number of BTC produced can be interpolated from the starting BTC produced per bump and "0" BTC produced per bump. This is because BTC produced has a linear dependence on difficulty. Maybe I'll come up with something. I think one of you young technical guys who took discrete math more recently than me should work it out.