Ok. I asked this in another thread already, but HERE, I particularly do not understand where a lot of you are coming from.
First off, no matter what you WILL achieve ROI, so I have to assume you mean POSITIVE ROI. But even given that niggle, there is virtually no way these machines will not reach break even in a fairly short time, after which your only ongoing expense is electricity, which is not that great of an expense on these devices.
So, those of you crying "no ROI" over and over again, WHAT FREAKIN" TIME FRAME ARE YOU TRYING TO BEAT?

Any frickin' timeframe. Please see
here.
I mean really. An ivestment that is likely to break even in less than a year??? That's insanely positive by most any business metric you care to apply! ...
If you don't break even in a year, you never will. Never. Unless conventional difficulty predictions are just a bunch of fud, spread by evol gobment infiltrators.
Damn am I tired of seeing estimates from the TGB calculator. If your link to TGB is going to be accurate, we will need to see 1.6 Ph added to the network every week between December and May. In total, that amounts to just over 5 times all pre-orders for all
companies who have currently offered pre-orders ..then consider how people around here are already reluctant to spend money on miners when we are not even 1% to 1 billion difficulty. Now, unless Intel decides they'd like to produce a miner, 28nm efficiency is going to become the de facto. That puts an equilibrium of slight profitability with a $100 BTC and electrical costs for the average person when difficulty is at about 15 billion. If manufacturers want to continue selling mining rigs, they'll have to be very near $1-$1.5 per Gh when difficulty is at only 3-5 billion. Then again, maybe TGB will be correct with the jumps in difficulty of 4, 10, or eventually 40 billion in a single month.

Anyway, this thread has gone to shit. Is it even 1 in 10 posts is on topic anymore?