The end of the ASIC

[Rival]

Hardware sales and pre-orders have reached a fevered pitch. New IPO's are being sold left and right. Everybody who knows anyone in chip or board manufacturing is getting into the game. But it is almost game over and no one appears to see it coming.

Here is an example from BFL. If you ordered a unit in April, 2013 you should have been aware that there were hundreds and hundreds and hundreds of Th/s ordered before you. They will all come online before you get yours plugged in. There was a large multiple of the entire global hash in April 2013 already ordered. There was already never any chance of getting your unit hashing at a profit. You need to have ordered no later than October 2012 to break even no matter when you got it. Even if they had delivered in a month the delivery of all of the units in the queue before you would have driven the difficulty level past a break-even ROI.

This is the problem facing all orders across all manufacturers. Everyone focuses on the other companies and forgets the damage to the difficulty inflicted by their own orders. If you did not order early enough all you can do is prevent other people from reaching expected ROI.

In the next year we will probably find ourselves in an unlikely environment where it is no longer cost effective to manufacture, sell, or buy ASIC miners. The only miners left will be those that have already paid for themselves and cost nearly zero to run. The days of buying a box, plugging it in, and watching $20 bills come flying out of it are just about over. No manufacturer will survive the difficulty levels we are creating. What good is 50Th/s for $1.00 if it only returns 50 cents over 50 years?

The golden age of bitcoin mining is coming to a close... but sadly few can see this. It is being destroyed by it's own success, another "Tragedy of the Commons" writ large. There is so much hashing power currently ordered and paid for that the difficulty will wipe out any possible earnings. There is a practical limit to current technology when it comes to chips, and baring some magical breakthrough, the hash rate will eventually plateau as it becomes uneconomical to create more. Any new offering will only be a drop in the ocean, or cost far too much to develop and produce.

Where the bitcoin experiment may lead no one today can say. But the future of mining bitcoins is pretty much set in stone. I am glad to have been a part of this unique and fleeting moment in history.

[1714808025]

1714808025

[1714808025]

1714808025

[1714808025]

1714808025

[1714808025]

1714808025

[1714808025]

1714808025

[1714808025]

1714808025

[empoweoqwj]

Well if you will order from BFL ....

[wormbog]

The days of money-printing machines never last long. But that was never Satoshi's endgame for the mining side of bitcoin. ASIC developers will continue to make faster and more efficient machines. Miners will buy them hoping to turn a profit. But the larger goal is to keep the overall network hashrate so high that no organization, not even a coalition of governments, can overwhelm and break the system.

Even miners who are selfishly motivated to turn a profit for themselves are helping protect the entire bitcoin ecosystem as a side-effect of their profit chasing. It's a work of genius, and amazing to watch.

[DeathAndTaxes]

As an individual miner waiting on a delayed pre-order it can be very bad, from a security standpoint it is very good. 

The units once bought can only do one thing and their operating cost is very low.  Global hashing power would need to rise to something like 20PH/s before even the least efficient ASICs will no longer be cost effective (operating cost only).  The higher efficiency ones will require closer to 100 PH/s before they are operated at an loss (once again looking only at electrical operating cost).

So while we may see no new (or more likely less) units sold the network will be much much more secure and there is no financial incentive for a miner to take their already purchased rig offline.  Even if a miner sells their rig it almost certainly just miners another miner puts it right back online.

Sales will slow after that but the most efficient units will still be sold (at significantly reduced prices/margins) and slowly that will edge out the least efficient ASICs.  In time the network will be 200 PH/s or more.  If the price rises significantly we could even see the network hit an exahash.

[wtfvanity]

Hardware sales and pre-orders have reached a fevered pitch. New IPO's are being sold left and right. Everybody who knows anyone in chip or board manufacturing is getting into the game. But it is almost game over and no one appears to see it coming.

Here is an example from BFL. If you ordered a unit in April, 2013 you should have been aware that there were hundreds and hundreds and hundreds of Th/s ordered before you. They will all come online before you get yours plugged in. There was a large multiple of the entire global hash in April 2013 already ordered. There was already never any chance of getting your unit hashing at a profit. You need to have ordered no later than October 2012 to break even no matter when you got it. Even if they had delivered in a month the delivery of all of the units in the queue before you would have driven the difficulty level past a break-even ROI.

This is the problem facing all orders across all manufacturers. Everyone focuses on the other companies and forgets the damage to the difficulty inflicted by their own orders. If you did not order early enough all you can do is prevent other people from reaching expected ROI.

In the next year we will probably find ourselves in an unlikely environment where it is no longer cost effective to manufacture, sell, or buy ASIC miners. The only miners left will be those that have already paid for themselves and cost nearly zero to run. The days of buying a box, plugging it in, and watching $20 bills come flying out of it are just about over. No manufacturer will survive the difficulty levels we are creating. What good is 50Th/s for $1.00 if it only returns 50 cents over 50 years?

The golden age of bitcoin mining is coming to a close... but sadly few can see this. It is being destroyed by it's own success, another "Tragedy of the Commons" writ large. There is so much hashing power currently ordered and paid for that the difficulty will wipe out any possible earnings. There is a practical limit to current technology when it comes to chips, and baring some magical breakthrough, the hash rate will eventually plateau as it becomes uneconomical to create more. Any new offering will only be a drop in the ocean, or cost far too much to develop and produce.

Where the bitcoin experiment may lead no one today can say. But the future of mining bitcoins is pretty much set in stone. I am glad to have been a part of this unique and fleeting moment in history.

Shouldn't this be under mining speculation and not economics?

You're saying bitcoin is going to be destroyed because of the collapse of this golden age of mining? Or you're just saying mining interest will wane from newbies because they can't make easy money?

[Rival]

While it could have gone under mining speculation, I felt it focused more on the economic impact as a whole. Much of the bitcoin economy currently is dedicated to hardware purchases and investments in mining securities. The effect of hardware sales and development cannot be understated in the bitcoin economy as it exists today.

I personally believe that should the hashrate plateau prior to widespread acceptance of bitcoin, the effect would be detrimental to the bitcoin economy. It is a bit of a race in my eyes. Although, from a security standpoint it would obviously be a good thing as stated by others.

[LainZ]

While it could have gone under mining speculation, I felt it focused more on the economic impact as a whole. Much of the bitcoin economy currently is dedicated to hardware purchases and investments in mining securities. The effect of hardware sales and development cannot be understated in the bitcoin economy as it exists today.

I personally believe that should the hashrate plateau prior to widespread acceptance of bitcoin, the effect would be detrimental to the bitcoin economy. It is a bit of a race in my eyes. Although, from a security standpoint it would obviously be a good thing as stated by others.

It seems to me that hashrate plateau is a necessary but insufficient condition to widespread acceptance.

When I am telling people about investing in bitcoins, I have noticed that one the things that bugs them the most is mining.

These kind of people do not want to know about money printing machines.

They want to invest, but at the same time, they know their neighbors could be producing them like crazy.

[damiano]

I think once we hit 500+mill difficulty the price of of bitcoin will rise violently

[Adrian-x]

It is evident to me mining difficulty has more impact than halving (or at least halving magnifies price in relation to difficulty)

More to the point how many GPU miners are still running?

For me I'd need to see a 50% drop in price today to make it unprofitable. I suspect at least 70% of GPU miners are still at it, and are waiting for the coming ASIC oversupply.

@ OP
In Q3 and Q4 of 2010 GPU mining had a bigger impact on difficulty than ASIC's today, (CPU mining was rendered useless) the result proponents of Bitcoin pushed the price up.  

While I agree with you. There are other variables like price that change the dynamic so no need to panic.

[tonto]

If short-term ROI is your goal, then you shouldn't be in the game.
 
If I mine one coin a day, or one coin a year, I'm still participating in the network, and I won't care.
 
Besides, my coin today (or in a year) will be worth a 1000 times what it is now in a few short years.  I'll be ahead
 
I'm the guy who used to participate in distributed.net, SETI@home, folding@home, etc. (for free)  I'm doing it for the value of the service, not necessarily the value of the currency.

[DeathAndTaxes]

If I mine one coin a day, or one coin a year, I'm still participating in the network, and I won't care.

That is a legitimate view.  Securing the network in a decentralized fashion even if it means low or no long term profit.  Nothing wrong with that.

Besides, my coin today (or in a year) will be worth a 1000 times what it is now in a few short years.  I'll be ahead

This is a flawed way of looking at it.  Say you can buy a miner X for 1 BTC today.  If that miner doesn't produce >1 net BTC over its lifetime you are always behind.  The exchange rate is irrelevant.  Why?  You always have the option to just buy 1 BTC today and thus have 1 BTC in the future.

However note the prior statement as a non-economic reason to still mine.  Just don't confuse non-economic good with "profit" which is actually a loss.  Personally I see nothing wrong with it.  If 10,000 people buy relatively small rigs (say 10GH/s avg) thats 100 TH/s helping to keep the network decentralized.  Maybe someday someone will convince netgear or the like to add low powered SHA ASIC into their mainstream router and you could have tens of millions of people mining in a decentralized fashion.

[Nagle]

If short-term ROI is your goal, then you shouldn't be in the game.

Mining is all about short-term ROI. The return on the equipment drops rapidly.

Besides, my coin today (or in a year) will be worth a 1000 times what it is now in a few short years.  I'll be ahead.

Mining and speculation are separate businesses.  If you want to speculate, you can do that without mining.

[Impaler]

The problem is that most miners ARE speculating, and see mining as a cheaper means to obtain coins then purchasing them on an exchange.  Most miners sell only enough coins to cover their operating costs, this is what is keeping the price afloat now after the bubble, very few coins actually reaching the market (especially with Gox being such a poor way to sell coins now).

As difficulty peaks and the glut of mining equipment really hits home miners will have no profit left after paying electric bills and will be forced to sell most of their coins or run their machines at a loss.  As more coins come onto the market prices will drop further reinforcing the trend.  As we saw in 2011 the peak of hashing comes after the peak exchange rates, and then the exchange rate bottoms out after the hash peak.  This time around I expect the same pattern with 3-6 months between the hash peak and the market bottom.

[crumbs]

@DaeathAndTaxes:
I'm not sure ASICs are a boon to network security.  Higher hashrate does not translate into more security if *per-hash cost drops by the same amount.*  
In other words, if the total hashrate of the network increases tenfold, while hashes/sec costs drop tenfold, the network's security remains the same.

If we assume that security is improved through distributing the hashrate, we start to see that ASICs are concentrating hashpower (the "why sell what you make when you can mine it" argument, and empirical data).  Not sure how that could be seen as securing the network.

Finally, the highly specialized nature of ASICs, what you see as a good thing that guarantees that these miners will remain online, rather than turned into gamer boxen, i see as another centralization danger.  These miners, as they grow toward obsolescence & their profitability plummets, will get sold at that point to professional miners (who have lowest energy & location costs).

And ASICs have finally wiped out the last traces of hobbyshopping from bitcoin mining.  Yeah, there will be novelty USB miners, but...

[DeathAndTaxes]

@crumbs.  ASICs have always been possible, and they always will be possible.  Ignoring that doesn't make the network more secure.

The "good miners" making some kind of agreement to use inferior technology (i.e. 1/10th the hashrate @ 10x the cost) doesn't make the network stronger.  The attacker will simply bypass that and use the more efficient technology.  

Say at one time the network consisted of $20M in GPUs.  Was the network secure?  No not against any real threat.  An attacker with $20M in funding wouldn't build $20M worth of GPU just because that is what the idiot defenders are using.  I mean this isn't some kind of ritualized duel at sunrise.  No instead the attacker would just spend $1M on some low end (130nm or older ASICs) and have enough hashing power to 99% attack the network at 5% of the cost.

Now in the future yes an attacker can use ASICs (but they always could) however when the hashrate rises to say 20 PH/s it is going to cost a lot more than $1M to 51% that.  ASICs alone don't secure the network but they do prevent an attacker from "cheating" and doing a 51% on the cheap.

These miners, as they grow toward obsolescence & their profitability plummets, will get sold at that point to professional miners (who have lowest energy & location costs).

Unsupported and I really doubt that. I don't think that any "pro-miner" wants to manage 50,000 USB Block Eruptors, connected to 6000 USB hubs, and 200 host PCs. No most likely they will be heavily discounted and sold to other hobbyist.  The fact that you have a variety of sized hardware from a variety of manufacturers is a good thing.  Those looking for max hashpower with minimal overhead/complexity are going to use a manageable number of identical units.  Managing 10 TH/s using 20 identical BFL minirigs is a lot easier than doing it with 208 Jalpenos, 3,187 Block Eruptors,  a handful of various KNC offerings, etc.  

[crumbs]

@crumbs.  ASICs have always been possible, and they always will be possible.  Ignoring that doesn't make the network more secure.

The "good miners" making some kind of agreement to use inferior technology (i.e. 1/10th the hashrate @ 10x the cost) doesn't make the network stronger.  The attacker will simply bypass that and use the more efficient technology.

I realize ASICs were always possible, but an attack on the network by an ASIC designed *exclusively* for such an attack is a bit irrational, don't you think?  
Why not design & build it in the open, fund your efforts with pre-order money, and *then* play with the network?  Why Spy vs. Spy when miners will *pay* you to build the silly thing, and fight each other for the right to send you money?  

Say at one time the network consisted of $20M in GPUs.  Was the network secure?  No not against any real threat.  An attacker with $20M in funding wouldn't build $20M worth of GPU just because that is what the idiot defenders are using.  I mean this isn't some kind of ritualized duel at sunrise.  No instead the attacker would just spend $1M on some low end (130nm or older ASICs) and have enough hashing power to 99% attack the network at 5% of the cost.

What "older ASICs"?  The attacker would design & manufacture enough '00 tech ASICs just to troll the network with a 51% attack?  I'm not even sure if the 1 mil figure makes any sense -- i suspect the investment needed is much higher.  What would be the finacial gain?  Bitcoin price would *zero out.*  Where's the profit?  That's one expensive prank.

Now in the future yes an attacker can use ASICs (but they always could) however when the hashrate rises to say 20 PH/s it is going to cost a lot more than $1M to 51% that.  ASICs alone don't secure the network but they do prevent an attacker from "cheating" and doing a 51% on the cheap.

You're forgetting that these ASICs are manufactured by ... a manufacturer.  Who may mine with them (this is one of sound business models), and, if successful, can easily have an instant 51% attack in *just his warehouse stock.*  Sure, there were the GPU manufacturers before, but they had bigger fish to fry. BTW, there is almost no difference in cost between making 10 or 10,000 ASICs -- so, why not?

These miners, as they grow toward obsolescence & their profitability plummets, will get sold at that point to professional miners (who have lowest energy & location costs).

Unsupported and I really doubt that. I don't think that any "pro-miner" wants to manage 50,000 USB Block Eruptors, connected to 6000 USB hubs, and 200 host PCs. No most likely they will be heavily discounted and sold to other hobbyist.  The fact that you have a variety of sized hardware from a variety of manufacturers is a good thing.  Those looking for max hashpower with minimal overhead/complexity are going to use a manageable number of identical units.  Managing 10 TH/s using 20 identical BFL minirigs is a lot easier than doing it with 208 Jalpenos, 3,187 Block Eruptors,  a handful of various KNC offerings, etc.

Just how much hobby fun can you have with a Block Erupter?  Why would anyone buy, oh, 10 tinker-proof, mine-at-a-loss identical things instead of *one*?  Mind you, other than adding a *truly insignificant amount of hashrate to the network,* there's nothing to be done with them.  (remember, we're talking about the future, the things are really obsolete)
I, of course, am talking about modular gear, along the lines of Avalon & BFL big guns, stuff that feels at home in racks or could be simply reconfigured to be happy in data centers.

[Adrian-x]

The problem is that most miners ARE speculating, and see mining as a cheaper means to obtain coins then purchasing them on an exchange.  Most miners sell only enough coins to cover their operating costs, this is what is keeping the price afloat now after the bubble, very few coins actually reaching the market (especially with Gox being such a poor way to sell coins now).

As difficulty peaks and the glut of mining equipment really hits home miners will have no profit left after paying electric bills and will be forced to sell most of their coins or run their machines at a loss.  As more coins come onto the market prices will drop further reinforcing the trend.  As we saw in 2011 the peak of hashing comes after the peak exchange rates, and then the exchange rate bottoms out after the hash peak.  This time around I expect the same pattern with 3-6 months between the hash peak and the market bottom.

I don't think most miners are selling to cover their overhead. I suspect they are saving.

I don't think you can look at 2011 in isolation nor 2012 - looking at what put Bitcoin on the map in 2010 is an exponential increase in difficulty not yet seen as an overall increase when looking at difficulty in logarithmic scale. As we saw in 2010 the peak in exchange rates comes after the quantum shift in peak of hashing. 





my thoughts referenced here

[Kaiji]

During the gold rush the people selling the mining equipment got rich instead of the actual miners.

[Impaler]

I don't think most miners are selling to cover their overhead. I suspect they are saving.

My point exactly, they save as many coins as possible and sell only what they absolutely must, If some miners are not even covering their electricity costs and are effectively subsidizing their mining then my case is only stronger.

I don't think you can look at 2011 in isolation nor 2012 - looking at what put Bitcoin on the map in 2010 is an exponential increase in difficulty not yet seen as an overall increase when looking at difficulty in logarithmic scale. As we saw in 2010 the peak in exchange rates comes after the quantum shift in peak of hashing.  





my thoughts referenced here

The graph clearly shows hash rates continues to climb for some time after the bubble bursts but then peaks and declines.  We are currently in the space between the two peaks, in 2011 that space was only about 2-3 months but it is clearly taking longer this time, also the hash rate growth is a considerably lower slope.  I suspect this is all to do with the fact that ASICs are not a commodity device like GPUs which anyone could buy off the shelf, thus the 2011 hash rate growth was limited only by willingness of people to invest in equipment, now their is a huge waiting list.

Also were likely to see a longer time period until hash peak because ASICs have a high capitol cost but a low electric cost, ware a GPU running constantly could consume electricity equal to its capitol cost in a year or two it will take an ASIC (at current prices) decades to do so.  This means ASIC mining is much more a race to recoup initial capitol in the face of rising difficulty, rather then a balancing act of ongoing costs vs ongoing coin generation.  The race is so tight that we may already be past the point were an ASIC ordered today would arrive in time to ever turn a profit, just one month delay in starting the machine up could wipe out it's whole profit margin.  But even then the machines marginal costs would be so low that it would not actually drop to unprofitable to run until the end of the year or roughly when difficulty passes 1 billion.  

But we might also speculate that hash rates growth has to level out at some point as the manufacturers hit their capacity, but at the same time they are all plowing forward full steam with expansion of their capacity and new players are entering the market with huge additional production.  Perhaps customers will rapidly dry up as people realize that the profitability window is or has already closed, if that's the case then I expect manufacturers to drop prices rapidly in the next few months, I wouldn't be surprised to see a monthly price decline in ASIC's that parallels the monthly rise in difficulty as only then can customers enjoy the kinds of returns that make people invest in mining equipment.

[Nagle]

During the gold rush the people selling the mining equipment got rich instead of the actual miners.

Yes, the Big Four:  Leland Stanford (groceries), Collis P. Huntington (hardware), Mark Hopkins (groceries) and Charles Crocker (hardware). As in Stanford University, Huntington, West Virginia, Mark Hopkins Hotel, and Crocker Bank.

Butterfly Labs is not going to be up there with those men.