If someone was offering 22nm SASIC it might be a bit more plausible.
Even if someone did have a 22nm SASIC you don't cut your price just to cut your price.
If it honestly had 50MH/W and a density of 1 GH/s per card (which looks like up to about 60GH per server that is insanely valuable. FPGA are >$1 per MH and "only" 15MH/W or so.
This imaginary product wins on raw hashing power (more powerful than a 6990), wins on efficiency (3x efficiency of best FPGA designs) and wins on density.
So imagine you have a product that is vastly superior to any other product on the market by any metric. Where do you price it?50% of the price of nearest inferior competitor? Really? How about just low enough to maximize profit while stealing 100% marketshare (up to your manufacturing capacity).
If this was real it would be an incredible steal at $1000+. If bitcoin was performing better I might buy one at $1200 assumming it had a warranty just to simplify my rigs. Replace a garage full of hot, noisy, bulky rigs (10GH/s) with a single cool and nearly silent workstation that I can keep in an airconditioned office. Even better put it in a datacenter (preferably one which accepts bitcoins and just collect a check each month). SIGN ME UP.
Anyone smart enough to build something like this is smart enough to know its value.
Someone releasing a product 3x better than anything else seen is suspicious but someone releasing a product 3x better at half the cost well that is a slam dunk scam. Free markets simply don't work that way.