You are too late to the ASIC game. There are at least three companies that have ASICs developed already, and you are a fool if you think those three companies aren't working on improved versions (well, maybe not BFL, lol).
That's A LOT of competition and developing an ASIC takes a lot of money and a long time.
These companies are selling their machines WAY over their cost because that is what the market will bear right now. ASICs in large quantities are cheap. Expect them to get cheap and competition to get tougher. And most importantly, difficulty will skyrocket.
In the not too distant future (within a year), we will be able to buy ASIC machines with about 100 GH/sec power that run at about 100 watts and cost a few hundred dollars. The difficulty will be such that the ROI for such a machine will be 1-2 years, so mining will be mostly for fun and small profits, not the cash cow it is now.
I think you're greatly underestimating the greed of the people building these devices.
The only way this turns into a "cheaper for faster" scenario is if someone steals an existing design - downsizes it to a much smaller process and straight up sets the price at something reasonable - then you might see a price war.
~
As far as coming late to the game - I agree... the only strategy that would provide for success is if the project is entirely privately funded (investors with very deep pockets) and as soon as they have a working design they start cranking out units and reduce the process.
What's needed is something in 45nm range (or less) to really make a killing as a hardware designer.
