Would you care to share the maths behind that statement?
The back of my napkin tells me that @ 40MH/W FPGA's are good to ~200M difficulty, ASIC Miner chips @ ~133MH/W are good until 666M difficulty - all assuming a $0.22/kWh power cost.
IMHO the difficulty can't keep rising at the current rate because it is a hell of a lot easier and therefore cheaper to go from 30TH/s to 60TH/s than from 400TH/s to 800TH/s. The next doubling to 1.6PH/s will be even more costly, unless the people building the next generation of ASICs are basically giving them away, and as the process technology shrinks down to the bleeding edge 20nm processes, the cost per wafer will actually increase...
see
http://mining.thegenesisblock.com/ As you can see from the red line on the graph, the latest difficulty hike was much larger than predicted!
Bitcoin used to buy ASICS - RISKY (as most will never get ROI on their bitcoins)
USD to Asics - less risky (with btc price going up this can attain ROI with a higher chance -however, you could be able to buy more bitcoins with the cash you used to buy the asics)
risk risk risk
Many got ROI very quickly after receiving preordered unit. The key is to have the device soon, but this preorder game is like gambling
I was chatting to someone yesterday that just received his batch #3 Avalon, which cost about 100BTC all up, and was looking at 2-3mnths ROI, with today's diff hike, he will not achieve ROI unless BTC/USD price goes up. I have a few BFL devices on back order from April/May, and they will probably not cover their power costs by the time they arrive.