Hacker News Reader

Comment by mrb | Hacker News Reader

66 GHash/sec. Or about the same as a cluster of 110 AMD Radeon HD 7970 cards!

Right now, 66 GHash/sec generates about $210/day (11 bitcoins/day).

If you have a machine that generates $210 a day, why would you sell it for $1,300? Wondering why they wouldn't plug them all in and not start selling any until they've dropped the value of bitcoin below their profit margin...?

Mining is extremely risky right now. If someone else turns on ASICs before you do, the resulting difficulty spike would reduce your profits significantly. It's even possible (although unlikely) that some of these ASICs will lose money. But selling ASICs locks in the profit immediately.

Very simple: they did NOT have the funds to develop these ASICs by themselves. They HAD to pre-sell them via pre-orders to bootstrap their business.

Avalon is literally the work of 2 chinese guys (ngzhang and yifu). They are not a big company with lots of money.

If they had enough to build one, they could bootstrap themselves up from that. (It takes about 6 days for an ASIC to generate enough money to make a new one.)

You can't build one ASIC; you have to order a batch of thousands and you have to pay over $100K for the masks weeks before you receive the chips.

$100K is nothing for the kinds of profit numbers I'm seeing here.

Firstly, their true NRE costs are closer to $200-300k IMHO.

Secondly, $200-300k is not "nothing". I bet they tried to raise investor money, but didn't find any. $210/day/unit is a lot, but Bitcoin still remains relatively a risky investment.

If these things can print money, then they're worth $200-300k.

If they can't, then they're not worth $1.k/ea.

The problem is that they can print money if they are released before their competition, but we don't know when the competition will ship. (In retrospect we can see that Avalon batch 1 shipped before BFL, but there was no way to know that ahead of time and batch 2 is still unknown.) So I can understand why outside investors would refuse to touch it and why the Avalon developers might have decided to reduce their risk.

Still, I don't understand why they don't delay/cancel the shipments? eEven if it's unethical.

No need to sell any of these machines before somebody else hits the market.

The major cost of generating bitcoin is not the hardware itself, but the power.

I know people who have unplugged their entire setup and flown it half way across the US to get power 10% cheaper.

That's no longer true for ASICs. This box consumes only $1-2/day of electricity, so it's likely that it will consume less than $1,300 of electricity over its entire lifetime.

If you sell machines, you get $1300 cash. If you mine, you'll get continually decreasing returns, and you end up with bitcoins instead of USD.

The blog referenced in another thread mentioned that you have to pay them in bitcoins and not in cash...

But they immediately cash out the Bitcoins anyway.

Smart guys.

The $210/day for $1,300 only holds for a single incremental device. Since the network produces a mostly fixed number of bitcoins per day regardless of the hashing power thrown at it, there are severe diminishing returns for running a lot of these devices. I think the current math is that if you ran 300 of these, each would only be producing $100/day. 600 and each would be producing $70/day, etc. (Pre)selling them scales much better and is far less risky since it insulates them from the fluctuating BTC/USD exchange rate and the possibility of competitors diminishing their share of the network down the line.

300 at 100 a day is $30k/day. If they're selling them for $1.3k, let's assume a 50% markup; it would only take about 6 days to make back their investment. 12 days if they're selling them at cost.

I don't really follow the maths here.

Presumably they want to sell much more than 300 devices (enough that running them themselves would be less efficient than selling them) and turning on these 300 for themselves would make customers much less willing to buy more from them in the future.

> Presumably they want to sell much more than 300 devices (enough that running them themselves would be less efficient than selling them)

That still doesn't make sense to me. If these machines work, they're money trees. You build them, run them, and use the ridiculous proceeds to build and run more of them, or to get out before the market becomes too crowded.

At scale you could fill racks with these things for a relatively small outlay.

> ... and turning on these 300 for themselves would make customers much less willing to buy more from them in the future.

The truth outs. In a gold rush, sell shovels.

When there's a gold rush, you want to be the guy selling the shovels.

Thank you!

Reply on news.ycombinator.com