How much does it cost to produce one Bitcoin? (Market Value Comparison)

[odolvlobo]

I found the following figures:

"According to Blockchain, over the last 24 hours Bitcoin miners used an estimated 135,950.77 megawatt hours of electricity." That is round about $195.000 a day. Daily there are 5.000 new Bitcoins mined. So that would equal to $39 cost of electricity per Bitcoin. Far away from current market price. Sure, to "produce" a Bitcoin you need hardware.

How did you get the $195,000 number?  That's 0.143 cents per kilowatt hour, which is about 100 times cheaper than average price of electricity in the US.

I doubt blockchain.info's estimate is real. ASICs hardly use any electricity at all.

[fallinglantern]

don't trust the blockchain.info electricity estimates. They're still based on GPU mining AFAIK. It's reasonable to assume a current gen ASIC consume 1 watt per GH. From there, the rest of the math is fairly straightforward and you'll realize that the vast majority of cost for mining a bitcoin is capital, and not operation.

[An amorous cow-herder]

The actually sales volume of all the mining equipment would be a very interesting figure to know though, can you explain your steps to calculate it?

Its basicly a fairly simple guestimate. If mining equipment is expected to actually return more money than it costs than obviously the hardware market as a total has to be smaller than the value of bitcoins mined. If it is expected to return less than the money invested for mining equipment on average than the hardware market is actually bigger than the BTC earnings. Currently the chances to break even on mining are nearly none existant unless you get a batch 1 preorder that actually ships on time.
Another way to get to the same result is to check the total network hashing speed in GH/s and multiply that by the average cost per GH/s. This is imho a pretty dangerous situation, if too many miners have to sell BTC to recoup at least a part of their investment this could create a negative spiral which in turn cause the hardware manufacturers to go bankrupt (since lower prices will make mining hadware less atractive). Which could negativly effect the bitcoin economy as a whole (companies going bankrupt could cause other other unrelated companies, such as merchants, from even touching bitcoin with a 10ft stick).
Not saying it will happen. But its i risk i wouldnt underestimate.

[Impaler]

We have speculated that mining could be and likely is overshooting so any estimate based on assumptions that miners will remain profitable is highly suspect.  If that was the basis for your estimate then I'm declaring it a worthless estimate.

I would be much more inclined to believe estimates derived directly from Hash-rates, observing the raw hash rate increase and comparing that to the average cost of equipment over the period to get an idea how much money was spent of equipment over the period.  I'll do that right now in fact.

Over December we saw ~6 million GH/s added to the network, Rival provides a price estimate of $41 and $36 per GH/s in early a late December respectively, I'll go with a round $40 to be conservative.  This would thus total to $240 million dollars of Mining equipment purchase in the month of December. 

Butt, Rival's numbers are for immediate delivery of equipment from E-bay though and this is likely skewing the estimate to the high end because most equipment is purchased direct but with a delay, immediate delivery comes at a premium.  A better price estimate would be derived from a survey of manufacturers and that might lower the estimate by half.  Rival can you get a hold of such numbers?

Compared to the value of mined BTC, if we estimate 4k coins a day and ~$900 then you would be at roughly $100 million gross mining revenue per month, which looks to be at least in the ball park of whats spent on mining. 

This is perhaps unsurprising at first if we naively assume that the sale of coins are what fund mining activity, then naturally if X revenue is generated from coin sales by miners then they will reinvest X in mining equipment.  But I do not think that very many newly mined coins are actually sold, the markets churn over a small number of high-velocity coins while most new coins go immediately into the miners personal stash.  Mining expenditures are largely new investments made with outside money, not with revenue from coin sales.  This is why I think the situation will inevitably end in overshoot as miners lose profitability and begin to release their hoarded coins to pay ongoing operating costs.

[mestar]

I've been saying for a year that ASICs have set off a kind of mining bubble, in which miners just pocket the BTC they produce as their profits which dries up the supply on the exchanges and inflates the price.  Eventually the system will break down when the marginal (electric) costs of mining don't even equal the profits and hash rates actually drop as a response to the overshoot.  When miners are right at the edge of profitability again they will operate like real mines, they will sell everything daily into the exchanges and this will drive down values as happened after the 2011 crash.  The market now can not absorb 5k coins per day being liquidated, that would drain all the USD off the exchanges in a week flat.  The present market is maintained by the bottling up of the newly mined coin stream by ASIC hoarders.

Excellent post. 

This current mining profitability is maintaned by two more factors:

-It's not easy to get mining hardware, you have to wait months.  (In the GPU era bubble, you could start mining in hours)
-most famous exchange does not do USD withdrawals, yet all the other exchanges seem to follow their price.  This also relieves the selling pressure.

When those two things are resolved, things will get interesting. 

[komar]

[mestar]

Nice charts. 

Is the Avalon #2 miner the most efficient on the Hash/Watt basis?
What is your assumed electricity cost?
Why would blue line on the first picture double, with difficulty almost the same?

[komar]

Nice charts. 

Is the Avalon #2 miner the most efficient on the Hash/Watt basis?

Nope. It has typical efficient for this moment.
According to this page, the next generation of ASICs will be 5 times more effective than avalons.

What is your assumed electricity cost?

$0.12 per KW/h. See computation parameters

Why would blue line on the first picture double, with difficulty almost the same?

It's a pink line actually. It's "amortized cost", means "what price should I set to cover my videocard's cost in 1 year".

[johnyj]

That's the question always lingering in many people's minds, is Mining Bitcoin really sustainable without constant major price rises ? for example if the price lingers around $1000-$1200 how long before mining collapses ?

IMO, the high cost of mining created a barrier for entry, so that any new comers will have to buy directly from the exchange, thus support the exchange rate

The large miners see the incoming huge demand for bitcoins, and they know it has a limited supply, they only need to hold all the mined coins to make sure the exchange rate kept stable. The capital and operating cost can be financed through loan, since the return will be huge after a couple of years. If they cash out the coins immediately, it defeats the very purpose of mining

[mestar]

It's a pink line actually. It's "amortized cost", means "what price should I set to cover my videocard's cost in 1 year".

Look at the blue line (or pink) in the first picture, in November 2012.  It doubles, and I don't understand why.

[komar]

Look at the blue line (or pink) in the first picture, in November 2012.  It doubles, and I don't understand why.

A reward changed from 50 coins per block to 25.

[mestar]

A reward changed from 50 coins per block to 25.

D'oh!

[Rival]

Butt, Rival's numbers are for immediate delivery of equipment from E-bay though and this is likely skewing the estimate to the high end because most equipment is purchased direct but with a delay, immediate delivery comes at a premium.  A better price estimate would be derived from a survey of manufacturers and that might lower the estimate by half.  Rival can you get a hold of such numbers?

I found it almost impossible to assign any realistic value to units which were pre-ordered. If you don't know when you will get your gear, you have no idea how many coins you might derive from it. A few weeks either way can have a huge effect. Given the fact that units are rarely available for immediate delivery from manufacturers, I was forced to look at the secondary market so as to at least glean some data which could be useful. It would be nice to try and derive a coefficient that could be applied to the numbers to determine cost/hash of deliveries from manufacturers. I just have no idea how it could be done. Consider BFL. The first to get units made a killing, the last ones got killed. Not sure how to average that, or if the numbers would tell us anything we don't already know intuitively.

edit: After additional consideration, I would suggest that the secondary market may be a roughly accurate average of the gh/s cost of delivered pre-order units, and not 50% as Impaler posits. The fact that they exist at all and are not pre-orders pretty much temporally locks them.

[Abdussamad]

Where did the op get the 5000 figure from? The idea goes that you have a block every 10 minutes and 25 new bitcoins are created as a result. So 25*6*24 = 3600 new coins a day.

[Maddin]

I just made a calculation:

Assumptions:

- BTC's a day: 4582    (average time for the past few months to solve 2016 blocks = 11 days.  2016*25 / 11 = 4582 )

- Gh/s acutally used: 15.000.000

-power cost: 0,12$ kw

- Questionable assumption: 1w per GH (is this realistic, what do you think? Knc Jupiter with 3.0 Th/s has 0.7w per GH and there is probably a lot of old hardware in use, so 1w per GH is maybe far too low)

15.000.000 GH * 1W/GH= 15.000.000 w/h * 24 = 360.000.000 w/day =  360.000 kw/day

360.000 kw/day *0,12$ per kw = 43.200$ for power per day

43.200$ / 4582 = 9.43$ per BTC (only power costs)

I don't know.. what do you think about this calculation guys ? Even with 5w per Gh it's only 47,14 $ per BTC

I'll do some more calculations and try to calculate the propotional hardware costs per BTC (with an uderlying assumption I don't know yet, any ideas?) 

[An amorous cow-herder]

- BTC's a day: 4582    (average time for the past few months to solve 2016 blocks = 11 days.  2016*25 / 11 = 4582 )

Thats a really good point. With the increasing hashrates the actual production is about 25% above target.

I don't know.. what do you think about this calculation guys ? Even with 5w per Gh it's only 47,14 $ per BTC

Sounds pretty plausible to me. But the power cost is a fairly insignificant part of the cost nowadays.

I'll do some more calculations and try to calculate the propotional hardware costs per BTC (with an uderlying assumption I don't know yet, any ideas?) 

Most basic assumption is that the hardware industry will naturally expand to the point were producing a BTC costs nearly as much as it is worth. And while that is (still) true for scrypt mining were hardware is readily available its more complicated since bitcoin hardware is a pre-order market nowadays.
Anyway, to get an estimate about hardware shipments check this thread.

[johnyj]

http://bitcoinwisdom.com/bitcoin/difficulty

The next difficulty jump is projected to be at least 30%. If this rate holds, a mining rig can only mine 3.33x coins of its first 10 day period income during its lifetime

Currently a BFL 230GH miner cost $4255 for immediate delivery (they are changing their pricing continuously), it will generate 0.0646 bitcoin per day, and 0.646 bitcoin during its first 10 day period. So it will only generate 3.33x0.646=2.15 coins during its lifetime, and that put the cost of each coin at $1979 without even the electricity cost

On CEX.IO, 230GH cost $8919 at today's bitcoin exchange rate of $843, it is more expensive than BFL but without the maintenance overhead

So the mining cost suggest a coin through mining will cost at least $1979 if the difficulty jumps 30% each time, and $1300 if the difficulty jumps 20% each time


One year ago I have projected that eventually the mining cost will drop close to the electricity cost after the market is saturated by ASIC miners, just like GPU era. But that is under an assumption that most of the miners will eventually ROI and use their rig as an electricity to bitcoin converter

The current situation is totally different, it seems we are still in the middle of this technology transition and at the same time the public awareness of bitcoin has greatly increased

The deployment of ASIC miners are 10 times faster then I projected, and this caused the exchange rate to lag behind. If the exchange rate is the market's valuation about bitcoin, and the mining cost is the bitcoin enthusiasts' valuation about bitcoin, now it is like a period when some miners were mining at a loss (The end of 2011)

[aminorex]

difficulty cannot be exponential because once every fab in the world is fully occupied making asics the increase will be linear until fujian can be paved over to build new fabs.

the marginal cost of mining is less than the average cost.  in the near term, while admittedly they have not shipped, cointerra for january delivery sold for 6k USD, 2TH/s, 1200w.   
$3/d for power, and 23 coins in 180 days at 25% difficulty increment (which is pessimistic, as per above), so $283/coin all-in on 6 month amortization.

In the long run, these chips will tend to the $10 range, and hardware will be inconsequential, but by then, say 9 months from now, the yield on 1200w will be about 0.01 btc, so costs are pretty constant until there is a big tech change, say 14nm process with merged litecoin a la gridchip.

[johnyj]

the marginal cost of mining is less than the average cost.  in the near term, while admittedly they have not shipped, cointerra for january delivery sold for 6k USD, 2TH/s, 1200w.  

In the long run, these chips will tend to the $10 range, and hardware will be inconsequential, but by then, say 9 months from now, the yield on 1200w will be about 0.01 btc, so costs are pretty constant until there is a big tech change, say 14nm process with merged litecoin a la gridchip.

The cointerra machines are for May delivery. Those pre-orders might get some profit,  but there are also pre-orders never get anything delivered. So for majority of people who want to get coin now, their choices are limited, they can do a pre-order and hope for the best or buy BFL's rigs directly

No one has ever expected the difficulty would rise so fast during 2013, and the 2014 might be even more surprising. Now the NRE cost for many chips have been paid, those companies can ramp up the production capacity at least 100 times more if they see a hash power war

If you compare bitfury and knc, you will see that the best move for ASIC manufacturer is to produce as many rigs as possible and squeeze many small miners out of the market so that they have to buy more rigs to keep their income from dropping

[Asrael999]

Let us  consider a miner looking today to buy new mining equipment. Our miner decides that their best mining option is a pre-order mining rig from KNC Miner, a Neptune (for our purposes our miner lives in the EU)
2nd batch Neptune costs $9995+VAT, -> $11,994 (VAT at 20%)
Our miner phones KNC speaks to them and hesitates a guess that the miner (a late 2nd batch order) will be delivered sometime in mid May (optimistic)

Our miner then looks at the growth of difficulty and estimates that mining power will continue to grow at an average of 1% a day between now and mid May (much lower than the current rate - optimistic ), meaning by 14th May difficulty will have reached  6,655,250,955, and in the first return period the Neptune will earn 3.17 BTC, the miner has free power.
The miner then assumes that difficulty will grow at a similar rate thereafter.
Therefore over its life the Neptune will earn 3.17/.1 -> 31.7BTC
The production cost per bitcoin is therefore $378.35 , the miner would not sell Bitcoin above this price

If the growth rate is 1.5% per day on a continuous basis the miner estimates the return will be approx 11.2 BTC over the life of the Neptune, the production cost per bitcoin (again free power) at his assumption is therefore 11,994/11.2, -> $1070.89

If the growth rate continues ad infinitum at the current rate of growth approx 1.8% per day, the return would be approx 6.2 BTC
-> $1,934.52 per BTC min sale price.

Two things -
1) this would only determines the price at which newly mined coins would be released to the market by new miners who take delivery of equipment in May - existing miners who have already hit ROI may sell at any price, but their share of produced coin will be declining.
2) on the 1.5% growth assumption we are adding 25,000 Thash every 11 days to the network by mid June, is that really likely?


All this really tells us is that there is a floor price below which newly mined coins will not be added to the market - this floor price is determined by expectation of Return/ghash in BTC and the cost/ghash in $ of the most efficient mining hardware.
This is not a floor on the price of bitcoin as supply from existing holders may still exceed demand.
As mining supply per week is a continuously shrinking % of the outstanding bitcoin stock shouldn't mining cost have less and less impact on the price of bitcoin?