The mining market balance

[EnergyVampire]

sub

Added to Watchlist: https://bitcointalk.org/index.php?topic=90136.0

[1713554425]

1713554425

[1713554425]

1713554425

[Dargo]

I don't see that it's as bad as you make out. We know that Bitcoins will be mined at the same rate over time regardless of the speed of the network. Setting ASICs and FPGAs aside for the moment, let's assume someone buying a $1300 GPU rig today could have it paid off in 12 months. How did they do that? By getting a small share of the coins mined over that time period, essentially determined by their share of the total hashing power of the network over that period. Let's say this miner goes for another 12 months after the rig is paid off, earning $700 over the cost of the rig. Net profit=$700 (I'm not including income from selling the rig since you can't sell ASIC rigs except to other miners). Now let's suppose ASICs come out 24 months from now and this same person decides to spend $1300 on an SC BFL single. Many others will be doing the same, so the difficulty will skyrocket. But this miner also has 40 Gh/s now. Why can't this miner pay off the new rig, and keep going to make a profit? Maybe it will take longer than 12 months, or maybe it will take less if this miner gets the ASIC rig early in game before very many others pile in. The difficulty will skyrocket, but this person's hashing power has also skyrocketed. I see no reason to think this person's fractional share of the network hashing power can't be enough to eventually pay off the rig and get into profitable territory. Obviously I have hypothetically supposed ASICs won't be out for 24 months, but this is a thought experiment, so that isn't relevant. If GPU mining can be profitable now, I don't see why ASIC mining can't be in the future. The difficulty will increase dramatically, but so will the hashing power/$ ratio. I'm not really committed to this line of reasoning, but I don't at the moment see where it goes wrong - would like to know why you think ASIC mining is a losing proposition. The analogy to the dollar auction was just too loose for me to figure out what you are thinking.      

Let's suppose buying a 1300$ ASIC rig could be paid off in 4 months. Me, the manufacturer, I know that you're going to pay it off in 4 months, because I just calculate it using the price/difficulty ratio. So I decide, because of the crazy demand, to up its price to 3900$. Now, your ROI is 12 months, and I make a lot more money. It's still a good ROI, but you, the miner, take all the risk. If the Bitcoin price crash after 4 months, I don't care, and your ROI become 14 or 15 month.

If you're angry at me for tripling the price like that, why should I give a fuck? You can't buy better anyway, because a 3900$ investment with a ROI of 12 months is better than a 400$ investment with a ROI of 80 months.

If, after a couple of months, the ROI on my products become 24 months, and they are selling less, I could just reduce the price. I decide to sell them at 1950$, the ROI become 12 months again, and I start selling again. You bought them at 3900$ 2 weeks ago? Tough luck buddy, deal with it. My marginal cost is 100$ anyway, so I can play with the price like that to maximize my profit.

Gotcha - I'm not assuming manufacturers are going to do this, at least not to the extreme of your example, but I suppose they could. Assuming the difficulty increases more gradually (as well it might if BFL is the only supplier), ASIC rigs purchased early in the game would be worth more since you can get more coin with them. If price of rigs drops gradually as difficulty increases, this may offset partially or completely any advantage of getting in early, but the time to ROI may not be affected much. Also, if BFL was really thinking this way already, I would think the projected prices of their rigs would be a bit higher than they are now. Anyway, I see what you are thinking now - it's possible, but not inevitable IMO.  

[Brunic]

Gotcha - I'm not assuming manufacturers are going to do this, at least not to the extreme of your example, but I suppose they could.

They could, that's the problem. They will be in a position to do it, without consequences. We could try to guess all day if they will do it or not, but that's not the point. The point is, because of that technological shift, an exclusive group can achieve complete dominance over another group. And that happens in the heart of the security system of Bitcoin.

Also, if BFL was really thinking this way already, I would think the projected prices of their rigs would be a bit higher than they are now.

Maybe, maybe not. I'm not in a crusade against BFL. I hope their project succeed, because they worked for it. I'm only looking to create a balance of power in the mining market.

[Dargo]

Gotcha - I'm not assuming manufacturers are going to do this, at least not to the extreme of your example, but I suppose they could.

They could, that's the problem. They will be in a position to do it, without consequences. We could try to guess all day if they will do it or not, but that's not the point. The point is, because of that technological shift, an exclusive group can achieve complete dominance over another group. And that happens in the heart of the security system of Bitcoin.

Also, if BFL was really thinking this way already, I would think the projected prices of their rigs would be a bit higher than they are now.

Maybe, maybe not. I'm not in a crusade against BFL. I hope their project succeed, because they worked for it. I'm only looking to create a balance of power in the mining market.

And you have my best wishes on that. I'm all for the balance of power. I do worry that it will be hard to catch-up, but i'm totally behind the intended result. Would I be willing to put some coin behind a project like this? Maybe, but I guess I'd want to see which showed the most promise of success first. Will be following. Also, I'm glad this isn't simply motivated by the reactionary "BFL is evil" attitude that seems so prevalent on this board.

[nedbert9]

I don't see that it's as bad as you make out. We know that Bitcoins will be mined at the same rate over time regardless of the speed of the network. Setting ASICs and FPGAs aside for the moment, let's assume someone buying a $1300 GPU rig today could have it paid off in 12 months. How did they do that? By getting a small share of the coins mined over that time period, essentially determined by their share of the total hashing power of the network over that period. Let's say this miner goes for another 12 months after the rig is paid off, earning $700 over the cost of the rig. Net profit=$700 (I'm not including income from selling the rig since you can't sell ASIC rigs except to other miners). Now let's suppose ASICs come out 24 months from now and this same person decides to spend $1300 on an SC BFL single. Many others will be doing the same, so the difficulty will skyrocket. But this miner also has 40 Gh/s now. Why can't this miner pay off the new rig, and keep going to make a profit? Maybe it will take longer than 12 months, or maybe it will take less if this miner gets the ASIC rig early in game before very many others pile in. The difficulty will skyrocket, but this person's hashing power has also skyrocketed. I see no reason to think this person's fractional share of the network hashing power can't be enough to eventually pay off the rig and get into profitable territory. Obviously I have hypothetically supposed ASICs won't be out for 24 months, but this is a thought experiment, so that isn't relevant. If GPU mining can be profitable now, I don't see why ASIC mining can't be in the future. The difficulty will increase dramatically, but so will the hashing power/$ ratio. I'm not really committed to this line of reasoning, but I don't at the moment see where it goes wrong - would like to know why you think ASIC mining is a losing proposition. The analogy to the dollar auction was just too loose for me to figure out what you are thinking.      

Let's suppose buying a 1300$ ASIC rig could be paid off in 4 months. Me, the manufacturer, I know that you're going to pay it off in 4 months, because I just calculate it using the price/difficulty ratio. So I decide, because of the crazy demand, to up its price to 3900$. Now, your ROI is 12 months, and I make a lot more money. It's still a good ROI, but you, the miner, take all the risk. If the Bitcoin price crash after 4 months, I don't care, and your ROI become 14 or 15 month.

If you're angry at me for tripling the price like that, why should I give a fuck? You can't buy better anyway, because a 3900$ investment with a ROI of 12 months is better than a 400$ investment with a ROI of 80 months.

If, after a couple of months, the ROI on my products become 24 months, and they are selling less, I could just reduce the price. I decide to sell them at 1950$, the ROI become 12 months again, and I start selling again. You bought them at 3900$ 2 weeks ago? Tough luck buddy, deal with it. My marginal cost is 100$ anyway, so I can play with the price like that to maximize my profit.

Gotcha - I'm not assuming manufacturers are going to do this, at least not to the extreme of your example, but I suppose they could. Assuming the difficulty increases more gradually (as well it might if BFL is the only supplier), ASIC rigs purchased early in the game would be worth more since you can get more coin with them. If price of rigs drops gradually as difficulty increases, this may offset partially or completely any advantage of getting in early, but the time to ROI may not be affected much. Also, if BFL was really thinking this way already, I would think the projected prices of their rigs would be a bit higher than they are now. Anyway, I see what you are thinking now - it's possible, but not inevitable IMO.  

Be very clear that it is entirely probable that current network capacity will double within 3 months of BFL shipping ASIC based products.

BFL's intent with their pricing will not be known to us unless they reveal it.  It is easily discerned that BFL's pricing is sufficient, based on their estimations, to recover NRE costs and to profit.  Once NRE costs are recovered the profit margin increases.  Going forward once a miner, or potential miner, cannot justify purchasing a BFL product based on quoted pricing due to much higher difficulty - long or infinite ROI - BFL would be compelled to lower unit pricing to maintain sales volume.  The point is that once NRE cost has been recovered BFL has significant latitude to reduce unit pricing thereby promoting a more rapid network capacity growth.  Both of these events would equate to negative ROI impact for early adopters of ASIC products.

To put this in to perspective.  Estimations are that ASIC chips can be manufactured for $.10 - 1.00 per GH.  Once NRE is no longer a concern, even including all other costs of doing business, you can see for yourself the latitude BFL has in pricing ASICs.

[Dargo]

Im sure some people will call for a fork, or even go for it, but I cant see it gathering any real traction. Whatever other algorithm you decide upon, it can be accelerated by another ASIC, so its going to happen again at some point. Will you keep forking and changing the protocol? It makes no sense, its gonna happen, so it may as well be now. I think most people will understand that and not follow a fork which only purpose is cutting off the legs of the first vendor(s) who invested heavily in to bitcoin.

The idea is already germinating:

"So, in this arena - price wise what would be a very reasonable estimate of how much it would cost BFL?
(so I know if the major effort to very slightly possibly invalidate sha256 would be worth the while to kill their company)"
-kano (posted in "Lancelot" official discussion thread)

Much as I like and respect kano, this is the attitude that has me slightly concerned.

[dave3]

You could make an argument that it would be in the best interest of miners if BFL has a monopoly on the bitcoin mining hardware business.  Ideally the price of the mining hardware would remain as stable as possible.  Yes, it may drop over time, but hopefully not too quickly.  If there's an ASIC competitor, then it could end up being a race to the bottom.

So if you think BFL will have no competition, ASICs might be a good investment.  If you think they will have competition, maybe not such a good investment.

[rjk]

Im sure some people will call for a fork, or even go for it, but I cant see it gathering any real traction. Whatever other algorithm you decide upon, it can be accelerated by another ASIC, so its going to happen again at some point. Will you keep forking and changing the protocol? It makes no sense, its gonna happen, so it may as well be now. I think most people will understand that and not follow a fork which only purpose is cutting off the legs of the first vendor(s) who invested heavily in to bitcoin.

The idea is already germinating:

"So, in this arena - price wise what would be a very reasonable estimate of how much it would cost BFL?
(so I know if the major effort to very slightly possibly invalidate sha256 would be worth the while to kill their company)"
-kano (posted in "Lancelot" official discussion thread)

Much as I like and respect kano, this is the attitude that has me slightly concerned.

+1

[bitlane]

Much as I like and respect kano, this is the attitude that has me slightly concerned.

Don't bother respecting Kano. He's a crybaby at best. He called for my BAN from the forum for posting a fake screenshot of my CGMiner performance and talking about a Malware app that was circulating around the bitcoin world (THAT I NEVER EVEN HAD, OR POSTED A LINK TO).
I can think of far better reasons to BAN me, than THAT....LOL

If he's not crying about, or fighting, in regards to the development of CGMiner/BFGMiner....then he's up on 'his soapbox' about something else.

He's done great work for CGMiner and it's backend, but at the end of the day, he's just a little girl.

[Dalkore]

Yes, this is true.  My question is, why should miners drive off a cliff if they know one is approaching?

Ive been looking for a good analogy to help explain this, like something in game theory, but I havent found anything. Its really a quite interesting situation by itself, and I cant believe its unique, but the closest analogy one Ive found, is the well known dollar auction; its quite different in many aspects, but the parallel is that in both cases, there is a large potential profit to be made, yet anyone participating in the game, even rational actors, are forced to behave in way that is irrational and leads to individual losses for everyone, except the seller who makes a windfall profit.

Now if everyone truly understood this, BFL would hardly sell anything. But the reality is that a lot, if not most potential customers dont fully understand this, and price/difficulty will be determined by those that understand this least. And heck, even if everyone understood it, there would be some willing to bet they were one of very few taking the risk and therefore think they will still come out on top. Seeing how many gamblers we have, thats probably a LOT of people .

Wonderful example.   There is another zero-sum game that comes to mind.  Its in a book in my library, let me see if I can find it online, if not, I will just type it up.

[pusle]

It is possible to make permutating algorithms for FPGAs rendering ASICs pointless.

Like this thread stated earlier, FPGAs and GPUs are not driven by the bitcoin related market.
This means that for each new process node (28nm, 14nm etc) the generic hardware is available to all and prices are not influenced by bitcoin mining.

For the ASIC market to function properly there would have to be a lot of competition to drive down margins.
They must also make enough money to follow the advancements of the silicon industry at all times.
Currently bitcoin is nowhere near the size needed to support this.
This fact makes it much easier/cheaper for irrational players to attack the network compared with a system where ASICs didn't have any advantage.

With the current algo bitcoin really needs ASICs to protect itself.
If there is a 51% attack,  I think algorithm changes are inevitable.
 

[Meni Rosenfeld]

I think the stated problem is exaggerated. There is an equilibrium where the rig price is so close to the marginal cost that it's not profitable to reduce it further. This can work with 1 manufacturer but the more manufacturers, the more stable the equilibrium. Without fixing prices, no manufacturer can arbitrarily increase prices too much. The equilibrium also shouldn't be too sensitive to the BTC rate - doubling the rate won't cause the rig prices to double, as the difficulty will also increase so they won't generate twice as much revenue.

The healthiest way to converge to this equilibrium when there's just 1 manufacturer which is leaps and bounds ahead of everyone else, is to start by pricing it similar to the competition, and gradually decreasing the price. Based on their announcement BFL don't plan to do it responsibly in this way, which is a cause for some concern.

That said, I will support an initiative to develop an ASIC mining solution which is not by a company who tries to hide the fact that their target market is Bitcoin miners on one hand, or that plans to centrally do all the mining itself on the other.

Yes, this is true.  My question is, why should miners drive off a cliff if they know one is approaching?

Ive been looking for a good analogy to help explain this, like something in game theory, but I havent found anything. Its really a quite interesting situation by itself, and I cant believe its unique, but the closest analogy one Ive found, is the well known dollar auction; its quite different in many aspects, but the parallel is that in both cases, there is a large potential profit to be made, yet anyone participating in the game, even rational actors, are forced to behave in way that is irrational and leads to individual losses for everyone, except the seller who makes a windfall profit.

Now if everyone truly understood this, BFL would hardly sell anything. But the reality is that a lot, if not most potential customers dont fully understand this, and price/difficulty will be determined by those that understand this least. And heck, even if everyone understood it, there would be some willing to bet they were one of very few taking the risk and therefore think they will still come out on top. Seeing how many gamblers we have, thats probably a LOT of people .

There's a very simple game-theoretic model for this, and the key factor is time. Someone who buys a rig knows his profits will decline as more rigs are bought, but he hopes to make up for it in the time until this happens. This leads to a natural and steady decline in prices and increase in difficulty.

Let's assume that the current network hashrate is 12 TH/s and BFL starts churning out 1 TH/s rigs at a rate of 1 per day. After X days, there will be a total of X+12 TH/s, so let's assume the profit per day for a rig is 7200*0.999^X/(X+12) (the exponential factor is to account for reward halving etc.) Then a potential buyer at day N expects to receive $\sum_{X=N}^{\infty}7200*0.999^X/(X+12)$. That's a nonelementary expression but the first 10 values are 28410.8, 27810.8, 27257.5, 26744.2, 26265.7, 25817.5, 25396.1, 24998.4, 24622.1, 24265. So a buyer on day 0 can expect to get 28410.8 BTC out of the rig, so that's what he should pay for it, and that's more or less what BFL should charge for it; on the next day it should charge 27810.8 BTC, and so on.

Of course the offered model is extremely simplified but it demonstrates the point. Additional factors such as differing skills and opportunities among potential buyers, uncertainty regarding delivery times, and a bit of irrationality for good measure just enhance the possibility of a stable equilibrium.

[P4man]

Im not sure $\sum_{X=N}^{\infty}7200*0.999^X/(X+12)$ is an analogy that helps me explain it to someone who doesnt get it. I was rather hoping something involving nickels,  dice or playing cards

That said, you are assuming you know BFL production and sales rate and you assume its constant; that is the key factor. In reality we have no clue. We can only find out roughly, after the fact, once those machines start hashing. Moreover BFL can vary supply as they see fit. Start with a trickle and gradually increase it, constantly fooling its customers who are trying to guesstimate the speed at which hashrate will increase.

[Meni Rosenfeld]

Im not sure $\sum_{X=N}^{\infty}7200*0.999^X/(X+12)$ is an analogy that helps me explain it to someone who doesnt get it. I was rather hoping something involving nickels,  dice or playing cards

That said, you are assuming you know BFL production and sales rate and you assume its constant; that is the key factor. In reality we have no clue. We can only find out roughly, after the fact, once those machines start hashing. Moreover BFL can vary supply as they see fit. Start with a trickle and gradually increase it, constantly fooling its customers who are trying to guesstimate the speed at which hashrate will increase.

Right, it's not easy to decide how much to pay, but there is an equilibrium and whoever makes the best job estimating the future wins.

I don't think BFL has much reason to artificially limit supply, demand should be high enough as it is. If they wanted to do that they'd be better off investing in something with a smaller NRE in the first place.

As far as verbal analogies go, a dollar auction has some resemblance but the quantitative difference means not much of the intuition is transferred. A classic example of time-sensitivity is negotiation between two people dividing a pie, where the pie slowly rots/eaten away by someone else while they're arguing. They'd prefer to make some fair division early on rather than constantly making counter-proposals which will end up with each of them getting less.

[Transisto]

What about a limited lifetime price protection ?

Say a manufacturer drop the price on ASICs from 300$ to 280$ at a later date they would have to offer a 20$ in refund or credits for future purchases.

A more complex formula could be devised to account for production cost, USD devaluation and miners profits.

Such a scheme would give much more confidence to investors who wouldn't fear the lowering prices and reduced profits.

[P4man]

What about a limited lifetime price protection ?

Say a manufacturer drop the price on ASICs from 300$ to 280$ at a later date they would have to offer a 20$ in refund or credits for future purchases.

A more complex formula could be devised to account for production cost, USD devaluation and miners profits.

Such a scheme would give much more confidence to investors who wouldn't fear the lowering prices and reduced profits.

Put yourself in BFLs place and think about it.  As difficulty goes up because of these asics, its inevitable that future prices per GH will come down, pretty much proportionally. What you are suggesting boils down to BFL getting cash up front,  but really selling at the lowest future price from day one, and therefore assuming all the miner risks (while selling the opportunity). If they want to assume these risks, they would be better off mining themselves, then at least they get to keep the opportunity.

Your suggestion would be dangerous and likely very expensive if they maintain a monopoly, it would be financial suicide if a competitor (or large asic miner) would appear.

Much simpler would be for BFL to make public their production planning and simply state how many TH they will ship over what time period and contractually bind themselves to that. If they do that, they could auction off their hardware to the highest bidder to maximize their revenue, while buyers would still be able to make an informed purchase decision.  The problem when a potential competitor emerges remains though.

[nedbert9]

Im not sure $\sum_{X=N}^{\infty}7200*0.999^X/(X+12)$ is an analogy that helps me explain it to someone who doesnt get it. I was rather hoping something involving nickels,  dice or playing cards

That said, you are assuming you know BFL production and sales rate and you assume its constant; that is the key factor. In reality we have no clue. We can only find out roughly, after the fact, once those machines start hashing. Moreover BFL can vary supply as they see fit. Start with a trickle and gradually increase it, constantly fooling its customers who are trying to guesstimate the speed at which hashrate will increase.

Right, it's not easy to decide how much to pay, but there is an equilibrium and whoever makes the best job estimating the future wins.

I don't think BFL has much reason to artificially limit supply, demand should be high enough as it is. If they wanted to do that they'd be better off investing in something with a smaller NRE in the first place.

As far as verbal analogies go, a dollar auction has some resemblance but the quantitative difference means not much of the intuition is transferred. A classic example of time-sensitivity is negotiation between two people dividing a pie, where the pie slowly rots/eaten away by someone else while they're arguing. They'd prefer to make some fair division early on rather than constantly making counter-proposals which will end up with each of them getting less.

Using a very plain interpretation my opinion is that the price to pay is one that anticipates conversion of ~80% of the mining base to SC, and all the re-investment by miners that entails, and allowing a modest profit.

The question is the $38 dollar per GH question.  My estimation is at that price BFL stands to have minimum revenue of $5 million.

[P4man]

The question is the $38 dollar per GH question.  My estimation is at that price BFL stands to have minimum revenue of $5 million.

Even if we assume BFL ship after the reward halving, their hardware should be able to mine almost $8M per year at todays price. If they manage to only get $5M revenue from selling, they are doing it  wrong.

[Meni Rosenfeld]

The question is the $38 dollar per GH question.  My estimation is at that price BFL stands to have minimum revenue of $5 million.

Even if we assume BFL ship after the reward halving, their hardware should be able to mine almost $8M per year at todays price. If they manage to only get $5M revenue from selling, they are doing it  wrong.

This got me thinking about a different model: The glove game. BFL has the right glove and each of its potential customers has a left glove (BFL doesn't have a left glove because it doesn't want to speculate on Bitcoin). With n customers, the Shapley value for each customer is 1/(n(n+1)) of the total reward, and for BFL it is 1-1/(n+1). Meaning that with 100 customers, BFL should be getting about 99% of the reward.

Again this is highly simplified, but it does indicate that BFL may have more leverage than we'd like.

[Dargo]

Much simpler would be for BFL to make public their production planning and simply state how many TH they will ship over what time period and contractually bind themselves to that. If they do that, they could auction off their hardware to the highest bidder to maximize their revenue, while buyers would still be able to make an informed purchase decision.  The problem when a potential competitor emerges remains though.

What motivation would BFL have to do that? They probably can't estimate their production very precisely, it partly depends on the manufacturers they rely on, and there are bound to be unforseen snags. And given how they are treated on this board, the slightest deviation from their stated plans would raise a chorus of bitching, accusations of foul play, conspiracy theories, and calls for lawsuits (which might turn into viable lawsuits since they have bound themselves contractually). It would not be smart for them to bind themselves contractually to this, especially with a community that has treated them like crap from day one. But even if they did, we would be foolish to take the projected plans seriously. We already know they are not good at estimating the specs of their equipment, and even worse at estimating ship times. It would be nice, though, if they could give us more information that might allow a very rough calculation of their production rate. But it would probably be unwise for them to do even this, since if the information turned out to be off by a factor of even 2, the inevitable bitching would fire up. So I seriously doubt we are going to be able to make informed purchasing decisions - early adopters of ASIC mining equipment will be taking a big risk, but the big risk does carry the possibility of bigger reward IMO.