Wouldnt be so sure. Have you tried selling $1M worth of bitcoins lately?
Mt. Gox is exchanging around $2M per month in bitcoin versus only $1M in newly mined bitcoin. So there is apparent depth. The actual depth would depend on what percentage of miners are hoarding bitcoin versus spending. A change in this ratio at a rate greater than new participants in the Bitcoin economy would swing rates. Given the vast profit margins of internally mining you could afford to swing the market down quite a bit. I would set a threshold of around $4 / bitcoin and move to the selling hardware to miners model if selling 100% of bitcoin output was moving the market.
Of course, if the market can't absorb 100% of miner output then buying single purpose hardware for mining is a very bad investment. Paying in advance to fund the development costs of a company that doesn't have confidence in Bitcoin is an even worse decision.
I'm not quite sure why BFL would take on MORE RISK after funding the development of an ASIC chip.
Good business managers know that risk mitigation is way more important than quick profits.