If they're planning to have the capacity to pump out 900 units per day on a long-term basis, then everyone who buys one is going to be even more screwed than if it was a scam, though. Assuming their mixture of devices averages out to about 30 GH/s per unit, which seems like a reasonably conservative assumption, that means they have the capacity to turn out 27 TH/s of kit every day. That's nearly a petahash/s every month. It won't take long for the break-even point to reach years. Unfortunately most of the costs of ASIC design are non-recurring and BFL will want to make the most of their investment, so if they're rational they'll drop prices hugely once that happens - making the break-even point appear reasonable to new customers, but meaning older ones may never make their costs back at all.