Quote from: markm on March 06, 2012, 10:27:39 PM

Many people doubted strongly that an ASIC would come along at all, due to supposedly HUGE startup costs.

If the claims about how prohibitive the initial research/development costs would be are valid, surely it would seem incongruous to have the first few units produced be dirt cheap. Quite the contrary, surely it is to be expected that the first production units would be sold to a select clientele at a premium price? If the price of these first few units were in the "reasonable" price ranges you folk would like to see them at doubtless that would have been taken as further proof it must be a scam, due to the evident failure to attempt to recoup some of the "prohibitive" development costs...

-MarkM-

Well no. If one has a fixed cost of $1M then to amortize that over 1000 chips will be +$1000 over build cost per chip. If you ammortize that over 20,000 chips it is $50. If you amortize it over 100,000 chips it will be it will be $10 per chip.

ztex (as an example) won't get much lower prices even if he builds 10,000 boards BUT an ASIC w/ huge upfront costs acheives significantly lower cost by scaling production.

Keeping prices high for a select few means less units moved and higher cost per unit. Both company and customer lose. There also is the risk with high prices of slow volume and someone comes along w/ cheaper/faster/cooler unit before you ammortize. Then you are left holding the proverbial bag.

An example:
Hypothetically (excluding profit margin, risk, volume pricing on production runs, etc)
$1M NRE cost.
$100 wafer cost per chip (1 GH/s)

Sell 1000 chips = $100 ea + $1000 NRE share = $1100 per chip ($22,000 per 20GH/s box)
Sell 5,000 chips = $100 ea + $200 NRE share = $300 per chip ($6,000 per 20 GH/s box)
Sell 25,000 chips = $100 ea + $40 NRE = $140 per chip ($2,800 per 20 GH/s box)
etc...

So someone selling a few units at very high cost seems counterintuitive to the economics of ASICS. With ASIC your non sunk cost will be very low. It is in your best interest to undercut everyone else, be the clear choice and line up tens of millions of dollars in orders.

If you want to sell a few units at high premium get early access to 28nm chips. No sense and pricing them low as your cost won't go down much when ramping up volume.