My statement wasn't a value judgement on the choice to pay more per Ghps than a device is likely to earn, just that this seemed to be happening - and not just with Blade customers either. I'm certainly not saying that the decision to buy was stupid, just that the choice was apparently not motivated by profit.
Either that, or they have different assumptions about the likely future income of such a device than you or I or others do.
Although I agree with other sentiments -
if a device is deemed to be worth more to asicminer as a miner rather than as a sale unit, then we should mine with it. However that assumes we have the ability to deploy all our units - and given that that's not the case, as I understand it, selling them seems wise from a shareholder's perspective.
If ASICMiner only adds 100Ghps to the network over say the next 100 days and the network hashrate remains the same forever after, a Blade would take ~550 days to pay for itself, assuming constant uptime. So the Blade is worth far more as a sale unit than a mining unit, if people are willing to spend more than the device is likely to earn over a lifetime.
Not sure how you got that. Using those numbers I get that a Blade pays for itself in 148 days. Obviously unrealistic since everyone knows that more TH is going to come online everywhere.
That was total back of the envelope stuff. Using a larger envelope and a better pencil, I got a slightly better result:
The estimate is for a linearly increasing hashrate from 75 to 175 Thps over 100 days, and then 175Thps from then on starting at 75Thps on day 1.. That means about 8.5btc in the first 100 days, and then 0.2 btc per day thereafter. 76btc in 440 days. What did you get for each section?
The problem also comes down to scale. If AM could sell all of its chips at 71 BTC/blade, every single blade, that would be pretty cool. But you're not going to be able to find a buyer that will drop 450,000 BTC to buy AM hardware. You would, however, be able to mine with as much as you wanted.
So you sell the Blades until the price approaches what you think you could mine with it over the device lifetime. Then you mine.
If you sell for more than the device is likely to ever earn, then of course it's better than mining.