I've done the math repeatedly... It's not about up front costs, it's about price per Gh/S and how profitable said Gh/S are.
Mobile got it right with the idea of renting Gh/S or trading for them over at cex.io.
Even with an investment of .35 BTC, you can make 0.01 BTC/day. That was my first "trial" there and it was worth it. I'd invite you to check it out as well, because while I am "plugging it"... It's free to sign up and absolutely worth having Gh/S in your portfolio (they pay you!)
https://cex.io/r/0/neuros/0/At time of writing, CEX.io (the service mentioned above) is 0.1045 BTC/Gh/S. This equates to about ~20 USD at current market rates. This option would not break even in under 30 days (I like tight schedules!), except you have the option to trade Gh/S back for BTC, so your initial investment is "refundable," depending on Gh/S to BTC market rates. You can also continually buy in to further Gh/S with your earnings, compounding your income.
That said, when we're looking at bigger hardware, anything that costs less than ~$6.50/$7.00 USD will break even in 30 days or less at current BTC/USD market rates. However, looking at being "paid back" in USD or other fiat shouldn't be important - the BTC should be.
I myself have ordered a large machine with a friend (less than 1Th/S but not small) and estimates show it should "pay itself back" within 30 days. The machine is $6.50 USD per Gh/S.
Even if BTC falls to $100 and we assume a ridiculously high profitability decline of 0.1 or even 0.04 (unrealistic by the worst math), the machine pays itself back and turns a profit in a 90-day period. However, we are not looking to recoup USD, as we are looking for a set amount of BTC to break "even." Much of this BTC will go into EMA trading and Gh/S rental for continuing compounded income, hashing speed, and holdings.
Oh, I didn't even say hello, how rude.... Hello BitcoinTalk
