Estimate of ASIC pre-orders: 13 to 15 PH/s (diff 1.8B to 2.1B) by end of 2013

[DeathAndTaxes]

Predicting mining 6 months out is like trying to predict today what the weather will be in 2089. However here is my guess.  

First I have always been in the "price drives difficulty but difficulty doesn't drive price" school of thought.  So all of the following is basd on price being relatively stable (say $200 +/- 20% long term with no more than +/-35% in the short term).  If exchange rate moves up or down significantly and "sticks" (i.e. falls to $100 and stays there for weeks/months) then you can expect the following to be pushed higher or lower respectively.

For Q1-2014.
I think we will see pretty aggressive hashrate growth in Q1 of 2014.  If the estimate holds and we are roughly 12 PH/s on 01/01/2014 my best guess is that will rise to at least 30 PH/s over the next three months.  As an upper limit I am reluctant to put one but if I had to pick a number I am fairly confident I would say less than 60 PH/s.  This is simply because there is a lag in chip ordering and thus hashrate growth is constrained by wafer batch size.  While in theory an ASIC company could order hundreds or even thousands of wafers I doubt any ASIC company wants the risk of holding a small fortune in chips in the event something happens to Bitcoin and/or mining.   So no matter how delusional miners are I think there will be cap on how fast (not how high but how fast) hashrate will rise.  

Without significant exchange rate increase I don't see mining companies being able to sell the amount of rigs necessary to push hashrate up another >20PH/s (for visualization that is >20,000 KNC Jupiters).   If we are looking at 20 to 50 PH/s added in Q1 at $10 per GH/s thats $250M to $500M.  I don't think there are that many rich people who are bad at math.   Even $4 per GH/s is $80M to $200M which is likely too high given the annual mining revenue and amount of existing hardware so my guess is hardware prices are going down a lot in early 2014.  I expect every major vendor to be <$3 per GH/s with prices heading closer to sub $2 per GH/s as the quarter progresses.  The vendors with the most aggressive pricing may be pushing up against $1 per GH/s.   I would be surprised to see <$1 per GH/s for complete "ready to mine" systems although raw ASICS in bulk (chips on a reel) might get that low.

For the miners who thought power doesn't matter and bought rigs in countries with high power rates (>$0.10 per kWh) we probably will see a lot of used hardware up for sales.  This will be a migration of hardware from high cost miners to low cost miners albeit at a significant price reduction.  This will be more frequent with lower efficiency rigs (anything worse than 1 J/GH).  This resale of hardware to lower cost miners will increase its economic lifespan and ensure difficulty isn't going down (other than minor fluctuations).  It isn't until an Avalon rig is paying more than the value of a BTC in electicity in low cost areas that equipment will be retired.   At current exchange rate that isn't going to happen until north of 30 PH/s.

I also think more than one vendor will go bankrupt, be exposed as a fraud, or simply exit the space in the quarter.  There might be a new entrant or two but net-net the number of vendors will shrink.  Margins will be falling and competition increasing.  The weaker players probably won't make it.  It wouldn't surprise me to see an OEM type arrangement emerge (i.e. AMD doesn't make graphics cards, they make GPUs and sell them to a dozen or so OEMs who make graphics cards) with one or more vendors simply supplying SHA-2 processors not complete systems to a handful of OEMs.

For Q2-2014.
Your guess is as good as mine.  To be honest, even the Q1 outlook should just be taken as my opinion but I think it is at least grounded in probable events.   Going beyond a full quarter out is just throw a dart at the board territory.

[geofflosophy]

Predicting mining 6 months out is like trying to predict today what the weather will be in 2089. However here is my guess.  

First I have always been in the "price drives difficulty but difficulty doesn't drive price" school of thought.  So all of the following is basd on price being relatively stable (say $200 +/- 20% long term with no more than +/-35% in the short term).  If exchange rate moves up or down significantly and "sticks" (i.e. falls to $100 and stays there for weeks/months) then you can expect the following to be pushed higher or lower respectively.

For Q1-2014.
I think we will see pretty aggressive hashrate growth in Q1 of 2014.  If the estimate holds and we are roughly 12 PH/s on 01/01/2014 my best guess is that will rise to at least 30 PH/s over the next three months.  As an upper limit I am reluctant to put one but if I had to pick a number I am fairly confident I would say less than 60 PH/s.  This is simply because there is a lag in chip ordering and thus hashrate growth is constrained by wafer batch size.  While in theory an ASIC company could order hundreds or even thousands of wafers I doubt any ASIC company wants the risk of holding a small fortune in chips in the event something happens to Bitcoin and/or mining.   So no matter how delusional miners are I think there will be cap on how fast (not how high but how fast) hashrate will rise.  

Without significant exchange rate increase I don't see mining companies being able to sell the amount of rigs necessary to push hashrate up another >20PH/s (for visualization that is >20,000 KNC Jupiters).   If we are looking at 20 to 50 PH/s added in Q1 at $10 per GH/s thats $250M to $500M.  I don't think there are that many rich people who are bad at math.   Even $4 per GH/s is $80M to $200M which is likely too high given the annual mining revenue and amount of existing hardware so my guess is hardware prices are going down a lot in early 2014.  I expect every major vendor to be <$3 per GH/s with prices heading closer to sub $2 per GH/s as the quarter progresses.  The vendors most aggressive with pricing may be pushing towards $1 per GH/s.   I would be surprised to see <$1 per GH/s for complete "ready to mine" systems although raw ASICS in bulk (chips on a reel) might get that low.

For the miners who thought power doesn't matter and bought rigs in countries with high power rates (>$0.10 per kWh) we probably will see a lot of used hardware sales as hardware moves from high cost miners to low cost miners albeit at a significant price reduction.  This will be more frequent with lower efficiency rigs (anything worse than 1J/GH).  The ability for used hardware to be sold to miners with low power costs means the difficulty isn't going down (other than minor fluctuations).  It isn't until an Avalon rig has higher electrical cost than BTC produced even in <$0.05 per kWh areas that we will see the first of the hardware heading to the landfill.   At $200 exchange rate that isn't happening until >30 PH/s.

I also think more than one vendor will go bankrupt, be exposed as a fraud, or simply exit the space in the quarter.  There might be a new entrant or two but I think on the number of vendors will be less by the end of Q1 then it is now.   It wouldn't surprise me to see an OEM type arrangement emerge (i.e. AMD doesn't make graphics cards, they make GPU and sell them to a dozen or so OEMs) with one or more vendors simply supplying SHA-2 processors not complete systems.

For Q2-2014.
Your guess is as good as mine.  To be honest, even the Q1 outlook should just be taken as my opinion but I think it is at least grounded in probable events.   Going beyond a full quarter out is just throw a dart at the board territory.

Very well put sir.

The one thing I'd like to add is that KnC has said that they will be putting out a second generation of mining hardware in Q2 of 2014 after halting shipping for December January and February; I think that we can expect this new miner to be the biggest and baddest mo-sucka on the market, and it's going to have to be. With the network hashrate pushing up against 20+ PH, we're looking at 1 TH of mining power being worth maybe 2 BTC long term if it first comes online in March 2014.

I see the price of bitcoin rising another 10x by the end of 2014, so the mining hardware game will keep on keeping on. Month over month growth is going to slow substantially, but general point is that the hardware that is turning profits toward the middle of next year has not even been sold yet.

[DeathAndTaxes]

IIRC KNC said they would be halting sales if every single other vendor did the same.  It was an empty promise.  No other vendor will be doing so and thus KNC will be shipping units every single month.

As for the size of future units well that gets interesting.  It is unlikely sub 28nm processors will hit the market before 2016 (maybe late 2016).  The hashpower of an individual unit is going to be capped by efficiency.  All current 28nm offerings are around 0.8 J/GH at the wall.  Given multiple smart teams ended up around the same place without undervolting/underclocking I don't think efficiency is going to get significantly better on 28nm platform.

So at 0.8 J/GH at the wall and standard US household outlet being limited to 1440W continual load that limits capacity to 1440/0.8 = 1.8 TH/s.   If efficiency improves to say 0.7 J/GH it is 2.0 TH/s and 2.4 TH/s at 0.6 J/GH.

Units will get cheaper but I don't expect individual units to get much bigger than Cointerra rig.  That is pretty much pushing what a non-dedicated branch circuit can handle.

[geofflosophy]

IIRC KNC said they would be halting sales if every single other vendor did the same.  It was an empty promise.  No other vendor will be doing so and thus KNC will be shipping units every single month.

As for the size of future units well that gets interesting.  It is unlikely sub 28nm processors will hit the market before 2016 (maybe late 2016).  The hashpower of an individual unit is going to be capped by efficiency.  All current 28nm offerings are around 0.8 J/GH at the wall.  Given multiple smart teams ended up around the same place without undervolting/underclocking I don't think efficiency is going to get significantly better on 28nm platform.

So at 0.8 J/GH at the wall and standard US household outlet being limited to 1440W continual load that limits capacity to 1440/0.8 = 1.8 TH/s.   If efficiency improves to say 0.7 J/GH it is 2.0 TH/s and 2.4 TH/s at 0.6 J/GH.

Units will get cheaper but I don't expect individual units to get much bigger than Cointerra rig.  That is pretty much pushing what a non-dedicated branch circuit can handle.

There's a big difference in efficiency between Cointerra and KnC, so I suspect that KnC will make up this difference in their second batch. They rushed those 28nm chips out and were successful, now they can spend some time tweaking them to get them better while their competitors play catchup. There are a lot of hardware errors on those chips as well, so I expect that they will be able to make significant gains from Gen 1 to Gen 2 of their chips.

Moral of the story is that the hashing power will keep going up. A story hit the wire the other day about BitPay processing a $1m order to ButterFly Labs, this was no doubt for Monarchs, so there's that hashing power coming online once it is released... 2014 will continue to be a bloodbath, but at least the network will be stable and secure!

[masterOfDisaster]

The analysis is for sure describing a feasible scenario. And I share your concerns regading validity of such a forecast. One thing that came to my mind which can mess up such scenarios completely (and I'm sure there are tons of other imponderabilities that could be considered...) is the possibility to redistribute the hashing power to other crypto currencies once it is no longer profitable to let the ASICs mine Bitcoins.
Some miners might continue mining Bitcoins just because they believe the price will rise. But in that case it would be better to simply buy the Bitcoins at an exchange and mine a more profitable coin
There are some other crypto coins using PoW and SHA256. One candidate which lately had a bit of an increased difficulty e.g. is Peercoin (which has not only PoW, but PoS as well; albeit the PoW is "Bitcoin compatible") and there are several others which might (or might not) be financially rewarding

[World]

Phoenix Technologies new update
http://www.sbwire.com/press-releases/phoenix-technologies-increase-batch-size-of-bitcoin-mining-hardware-373334.htm

[Puppet]

Phoenix Technologies new update
http://www.sbwire.com/press-releases/phoenix-technologies-increase-batch-size-of-bitcoin-mining-hardware-373334.htm

So even the scammers are upping their (non existent) volume lol.

[rampalija]

nice explanation and forecast of difficulty, thx man. I can see that no miner is vorth of buying anymore

[CMMPro]

nice explanation and forecast of difficulty, thx man. I can see that no miner is vorth of buying anymore

And this ^^^ is why the hashrate increase we are talking about in this thread won't occur for quite a while.
No one is interested in buying anything new right now...there is a glut of supply and no demand until things rebalance themselves.  

There is going to be little market left for new mining equipment in the near future.

I would guesstimate that almost all of the current undelivered customers are smartly considering refunds at this point.

Miners these days have zero tolerance for mining at a loss.

[TooDumbForBitcoin]

So at 0.8 J/GH at the wall and standard US household outlet being limited to 1440W continual load that limits capacity to 1440/0.8 = 1.8 TH/s.   If efficiency improves to say 0.7 J/GH it is 2.0 TH/s and 2.4 TH/s at 0.6 J/GH.

Units will get cheaper but I don't expect individual units to get much bigger than Cointerra rig.  That is pretty much pushing what a non-dedicated branch circuit can handle.

Will the 2014 buyer be a guy mining at home?

Somebody looking to own 0.01% of the 20P network hashrate (what a Monarch buyer is trying to get in today's 4P network - but sadly will end up with 10th of that) will be interested in a 2T miner, but won't the bigger players be willing to buy as big as can be made?  They won't be limited by electrical infrastructure.

[DeathAndTaxes]

So at 0.8 J/GH at the wall and standard US household outlet being limited to 1440W continual load that limits capacity to 1440/0.8 = 1.8 TH/s.   If efficiency improves to say 0.7 J/GH it is 2.0 TH/s and 2.4 TH/s at 0.6 J/GH.

Units will get cheaper but I don't expect individual units to get much bigger than Cointerra rig.  That is pretty much pushing what a non-dedicated branch circuit can handle.

Will the 2014 buyer be a guy mining at home?

Somebody looking to own 0.01% of the 20P network hashrate (what a Monarch buyer is trying to get in today's 4P network - but sadly will end up with 10th of that) will be interested in a 2T miner, but won't the bigger players be willing to buy as big as can be made?  They won't be limited by electrical infrastructure.

The point is that mining is perfectly parallel.  It makes little sense selling a 6 TH/s unit for dedicated power connections and a 2 TH/s unit for "the home guy".  Just make a ton of 2 TH/s units and the small miners buys 1 or 2 and the big miners buys 100.  Higher economies of scale, less supply chain complexity, and higher margins.

The only reason for multiple units is when a customer can't afford the unit price but as the price per GH/s falls that becomes less of an issue.

[DeathAndTaxes]

I would guesstimate that almost all of the current undelivered customers are smartly considering refunds at this point.

I would guess most customers can't get a refund or at a minimum vendors make if very difficult.  Those ordering early even with credit cards are likely beyond the chargeback window.  Even KNC which had the most liberal refund policy went no refunds because orders were "in production" the last of the oct orders didn't ship until 3 weeks after they went "in production".

Still even if every miner got a refund the chips have in most cases already been purchased by the foundry.  It would be bad news for vendors because they likely would need to lower prices to sell the same chips again but either they would or they would mine themselves.

If the numbers are right and >10 PH/s of chips have already been ordered from foundries then it is only a matter of time before those chips are hashing (by someone, somewhere).

[Puppet]

Monarch is delayed (shocking!, lol), but this may raise some eyebrows:
"we should be able to product more than 1000 Monarchs per day. "
https://forums.butterflylabs.com/announcements/4414-monarch-information.html#post63779

Thats 12PH worth of 600GH monarchs (the 300GH model was only added later) per month.
Of course this says nothing about how much they actually sold, theoretically its possible they only sold 2 weeks worth of production and may not produce and sell more than that ever, but I kind of doubt that.

[btc_uzr]

Monarch is delayed (shocking!, lol), but this may raise some eyebrows:
"we should be able to product more than 1000 Monarchs per day. "
https://forums.butterflylabs.com/announcements/4414-monarch-information.html#post63779

Thats 12PH worth of 600GH monarchs (the 300GH model was only added later) per month.
Of course this says nothing about how much they actually sold, theoretically its possible they only sold 2 weeks worth of production and may not produce and sell more than that ever, but I kind of doubt that.

sure, BFL and 1.000 devices a day...AND they would ship them instead of mining with them oh their own as happened before...dunno why I can't believe it. 

They do more pr than engineering and they are dying. They currently try everything to survive, like send themselves a million through bitpay for some attention of news, claim ridiculous numbers and such for implying *there are soooo many BFL-customers out there, become one today ourself*  in order to suck new guys in joining the game lately.

[Puppet]

Monarch is delayed (shocking!, lol), but this may raise some eyebrows:
"we should be able to product more than 1000 Monarchs per day. "
https://forums.butterflylabs.com/announcements/4414-monarch-information.html#post63779

Thats 12PH worth of 600GH monarchs (the 300GH model was only added later) per month.
Of course this says nothing about how much they actually sold, theoretically its possible they only sold 2 weeks worth of production and may not produce and sell more than that ever, but I kind of doubt that.

sure, BFL and 1.000 devices a day...AND they would ship them instead of mining with them oh their own as happened before...dunno why I can't believe it. 

They do more pr than engineering and they are dying. They currently try everything to survive, like send themselves a million through bitpay for some attention of news, claim ridiculous numbers and such for implying *there are soooo many BFL-customers out there, become one today ourself*  in order to suck new guys in joining the game lately.

I dont understand why people draw such conclusions. The reason it takes 6+ months for them to ship your order is not because they have no orders, nor because they would self mine in any meaningful way. Simple evidence for that is the network speed, and the fact it doesnt make a lot of sense in the first place. Self mining is no where near as profitable as selling these devices.

The ridiculous long delays is just a combination of incompetence, unwillingness (slow shipping=low network difficulty=high perceived profitability=more preorders) and the sheer amount of orders they did get. Any other company would have stopped taking orders ages ago if they could not ship them in a reasonable time, but BFL keeps advertising and raking in orders they know they will never have to ship, since next year the import tax and electricity alone would make it silly for most customers to accept their 65nm product, and most of them will be switched to "mining by the GH".

BFL are devious, deceitful/dishonest, but they are not stupid, not bankrupt, not selfmining and not short on orders.

[RoadStress]

Monarch is delayed (shocking!, lol), but this may raise some eyebrows:
"we should be able to product more than 1000 Monarchs per day. "
https://forums.butterflylabs.com/announcements/4414-monarch-information.html#post63779

Thats 12PH worth of 600GH monarchs (the 300GH model was only added later) per month.
Of course this says nothing about how much they actually sold, theoretically its possible they only sold 2 weeks worth of production and may not produce and sell more than that ever, but I kind of doubt that.

sure, BFL and 1.000 devices a day...AND they would ship them instead of mining with them oh their own as happened before...dunno why I can't believe it. 

They do more pr than engineering and they are dying. They currently try everything to survive, like send themselves a million through bitpay for some attention of news, claim ridiculous numbers and such for implying *there are soooo many BFL-customers out there, become one today ourself*  in order to suck new guys in joining the game lately.

I dont understand why people draw such conclusions. The reason it takes 6+ months for them to ship your order is not because they have no orders, nor because they would self mine in any meaningful way. Simple evidence for that is the network speed, and the fact it doesnt make a lot of sense in the first place. Self mining is no where near as profitable as selling these devices.

The ridiculous long delays is just a combination of incompetence, unwillingness (slow shipping=low network difficulty=high perceived profitability=more preorders) and the sheer amount of orders they did get. Any other company would have stopped taking orders ages ago if they could not ship them in a reasonable time, but BFL keeps advertising and raking in orders they know they will never have to ship, since next year the import tax and electricity alone would make it silly for most customers to accept their 65nm product, and most of them will be switched to "mining by the GH".

BFL are devious, deceitful/dishonest, but they are not stupid, not bankrupt, not selfmining and not short on orders.

Why shouldn't they selfmine since only Paypal customers can get refunds? Considering their way of doing business it actually makes sense that they would mine for a period of time before sending the units out. Customers can't do anything about it and.

[sickpig]

Phoenix Technologies new update
http://www.sbwire.com/press-releases/phoenix-technologies-increase-batch-size-of-bitcoin-mining-hardware-373334.htm

So even the scammers are upping their (non existent) volume lol.

never heard about Phoenix Tech... their X3 model should mine at 1 TH/S with a Power Consumption of 650 W, this would mean 0.65W/GHs, to good to be true considering what D&T said a few posts above, and I have no reason to doubt his estimate on this matter

[Puppet]

Why shouldn't they selfmine since only Paypal customers can get refunds? Considering their way of doing business it actually makes sense that they would mine for a period of time before sending the units out. Customers can't do anything about it and.

Until recently BFL were the main provider of mining hardware. Arguably even today they are still the biggest provider. Had they mined with their sold but unshipped supposedly delayed hardware, two things would have happened:
- Difficulty would have shot up before they began shipping. We didnt see anything of the kind
- They would have shot themselves in the foot. One of the main reasons BFL did so well is that they started accepting preorders in august 2012 and didnt ship (or deploy) a single asic until, what was it, may 2013? Thats almost 10 months during which difficulty was low and mostly flat and prospective customers who used a mining calculator were willing to almost pay the weight of these machines in gold. Self mining would have dramatically reduced the perceived value of these machines. Even if they had the hardware, it would have been smarter to turn it off.

Now today you might argue it "makes sense" because BFL's impact on difficulty is dwindling, but then so is the profit potential of 65nm hardware. They claim to be shipping 1000 units per day now, even if its only half that, just how many days worth of production do you think fits in their datacenter? Its just not worth the bother if you can make fortunes selling vaporware and $250 worth of hardware for $2500. There is a reason Avalon, Bitfury, asicminer are all mainly selling gear, its simply more profitable than mining (and their margins arent even close to BFL's margins).

[ralree]

I dont understand why people draw such conclusions. The reason it takes 6+ months for them to ship your order is not because they have no orders, nor because they would self mine in any meaningful way. Simple evidence for that is the network speed, and the fact it doesnt make a lot of sense in the first place. Self mining is no where near as profitable as selling these devices.

The ridiculous long delays is just a combination of incompetence, unwillingness (slow shipping=low network difficulty=high perceived profitability=more preorders) and the sheer amount of orders they did get. Any other company would have stopped taking orders ages ago if they could not ship them in a reasonable time, but BFL keeps advertising and raking in orders they know they will never have to ship, since next year the import tax and electricity alone would make it silly for most customers to accept their 65nm product, and most of them will be switched to "mining by the GH".

BFL are devious, deceitful/dishonest, but they are not stupid, not bankrupt, not selfmining and not short on orders.

This +1.  AFAIK they tested / burned-in off main net, and were just super-slow at shipping.  I've seen no actual evidence of self-mining, even though people seem to like to think it happened (BFL also eats babies, BTW).  If you have some evidence of this, please post it!

[TooDumbForBitcoin]

6/1/13, network hash rate 156T.  10/1/13, network hashrate 1450T.

3% to 5% fluctuations in network hashrate on a daily basis.  Someone "hiding" 2% of the network hashrate would not be discovered.

So say a vendor builds 1T daily for its customers, but delays shipping by 3 days.  It can now hash with 3T paid for by customers.

Add 200G daily to the self-mine and you add 1T weekly to your self-farm, all paid for by customers.  ("we're shipping 800G daily, we're great".  Yeah, right.)

At the end of four months the vendor has 20T on the network, paid for by customers. ("we shipped 100T, we're great".  Yeah, right.)   A blip on the difficulty at 1.5%, but good for $10,000 daily in coinbase, $3.5 million annually, paid for by your customers, thanks.

Proof? none.  Do I believe it could be true? absolutely.  Are there 20T of customer's miners unshipped? plenty of threads say so.  Destroying the customer faith and future sales? not judging by 28nm "product" pre-orders. Unethical? yes ... that's the fun part.

EDIT:  They could even say, "we reserve 20% of our mfg capability for our own equipment, nothing wrong with that".  I guess, although 20T of customers waiting for miners paid for 8 months ago might disagree.