No, wait, I think I get it:
1. Lots of demand. Price spikes.
2. Supply increases,
3. Mining difficulty ratchets up.
Because mining difficulty is a function of mined bitcoins (and not a function of time) this price estimate isn't a refutation of Moore's law. It basically means that the demand for bitcoin, quantified I-don't-know-how, is going to double quite a bit faster than dollar-normalised processing power.
(Can anyone tell me whether increases in mining difficulty happen smoothly or sharply? I wonder whether the current market is smart enough to price a cliff in the rate.)