Based on an electricity price of $125/mWh and power consumption of 650W/gigahash, Bitcoin costs $9m a day, or about $3.3b a year to run. The total value of Bitcoin is about $7.5b. Incidentally, the US mint and federal reserve combined 2013 budget was $8.1b for 2013 and the total value of US currency in circulation is about $800b.

This raises a few interesting questions: will the logistical cost of running Bitcoin rise linearly with its total value? If so, are Bitcoin's running costs too great to make it a viable global currency?

Sources: http://www.federalreserve.gov/publications/budget-review/fil...

http://www.treasury.gov/about/budget-performance/Documents/2...

http://bitcoincharts.com/charts/mtgoxUSD#rg60ztgSzm1g10zm2g2...

The fundamental problem with bit-coin is anyone willing to pay for twice the current mining costs can double spend and destroy the currency. As such there is no way even 100 billion in mining can protect 10 trillion in value. QED: the long term viability of bit-coin as a cheap world wide currency is impossible.

Anyway, the secret service and possibly the FBI one run anti counterfeiting operations so it's slightly higher than that. Beyond that, I think there is a fair amount of diplomatic operations that support the USD but it's still far far cheaper to run relative to value than bit-coin right now.

A 51% attack only gives you the ability to double-spend coins you already control. It doesn't give you the ability to crack the private keys of others. So there is no straightforward economic incentive for a 51% attack unless you own a huge amount of Bitcoin, and thus would be hurting yourself.

You can double spend with wallets you own so 1BTC > 2BTC > 4BTC ... any number you want fairly quickly.

I think this is false. Double spending doesn't create coins, it screws someone who accepted a coin. My understanding of the 51% attack is that you can basically reliably snip the end from the block chain. That doesn't let you put arbitrary things into it.

But it'd be difficult to keep such an attack secret, and the value of bitcoin would zero pretty much immediately.

So the only way to make money on this attack would be to short a very significant amount of bitcoin. I don't know it that would be illegal - does it count as insider trading if you cause the value change?

So the mining costs do not have to cover the entire value of all bitcoins. They only have to be large enough to discourage political attacks, and large enough that a bet to short twice the mining costs worth of bitcoins would be noticeable.

Still, it's an interesting point that I haven't seen raised before.

EDIT: woops, other siblings of this posts are right. The protocol validates all chains before accepting the longest one as true, so you can't print or steal bitcoins even with 51%

You seem to be assuming that while USD has human protections, the only thing protecting BTC is the chain. But if someone were to actually destroy 10 trillion USD, there would be social consequences as well; it's not like people would just accept it quietly.

More importantly, what gains would the attacker get to compensate an expenditure of such big amounts of money?

Given that ASIC mining runs more like 10 W / GigaHash [1] (not 650W as stated on blockchain.info/stats), we are actually looking at more like ~$150K per day, or $56M per year to run.

At the moment, the value of bitcoin is $7.2B [2], so we can assume that if it scales up to $2T (about the same as M1 of the dollar[3]), then if cost scales linearly, it will cost about $15B per year to run the bitcoin network. That is probably less than it costs to pay the mint, federal reserve, and secret service.

1) http://millybitcoin.com/does-bitcoin-mining-use-a-large-amou... 2) http://www.bitcoincharts.com/bitcoin/ 3) https://www.cia.gov/library/publications/the-world-factbook/...

Do we know how prevalent the ASIC rigs are? I'd be really interested to see a rigorous study of Bitcoin's energy costs. Even so, it seems like the estimates on Blockchain.info are off by an order of magnitude. I've amended my comment to include the power consumption on which my estimates are based.

Theoretically the total cost of the energy expended on mining will be equal to the value of the mining rewards and fees received by the miners, minus the cost of the hardware amortized over time and a small profit.

Additionally, mining operations will gravitate to areas of the world with excess electricity, since it will be cheaper to mine and thus more profitable.

So the "cost" of running Bitcoin will rise with price, but fall as the block rewards continue to halve every 4 years (unless transaction fees make up the difference). The users essentially pay for it through inflation and transaction fees. It's really quite elegant.

How is that even a fair comparison? Mining isn't just "creating" new bitcoins, it's verifying the transactions of the whole network. To make a somewhat fair comparison we'd have to include the electrical costs of all the physical banks and bank networking systems in the US, which is much, much more expensive. Even that's not a fair comparison, either, since bitcoin is a global currency, but let's not go there.

It's true that it's not a fair comparison. The corresponding fair comparison is interesting, though. The thing I worry about is long term. As technology gets better, running banks should become more energy efficient. Because bitcoin mining is inherently competitive, it can't become more efficient in aggregate - if it's cheaper per hash, people will generate more hashes (and/or the difficulty will rise).

The US Mint and the Federal Reserve aren't the only institutions -- there are many other similar institutions duplicated across every sovereign power. Also factor in all the national/international banks, too.

>so why not go for a nice central-custodian system like that run by VISA, Mastercard, or your friendly neighbourhood bank which is protected by the state, rather than for one that relies on the fact that protecting it wastes so much money that every attack will be very costly (if this was not clear, I have explained this in detail here)

This article brings up a good point about consumption, but this introductory question is dumb. The short answer is "because that's the entire point".

That's the entire point for (most? all?) the people involved in creating bitcoin. It's the central point for many of the people involved in bitcoin today. But there's a large number of people who just want a cheap and easy means of electronic payment.

They have that, it's called VISA, Mastercard, etc.

Bitcoin promises to be cheaper, and to not require blessing of payer or payee.

Yes, and one of the reasons BTC is cheaper is because it's a market, and not "a nice central-custodian system like that run by VISA, Mastercard" or any other monopolist.

Yes, but from the point of view of the people I am describing that is an implementation detail, not "the entire point."

So the biggest problem with this post is that blockchain.info specifically says that they assume 650W per GigaHash. This is just out of date.

Modern ASIC miners (which are doing the majority of the mining) are doing more like 10W per GigaHash.[1]

So we only need 1/65th of what is being estimated.

1) http://millybitcoin.com/does-bitcoin-mining-use-a-large-amou...

10W/GH is a pretty old number. A lot of miners are now around 6W/GH on average from butterfly labs & asicminer. Newer miners that have been released in last few months are 1.5W/GH from KNC & BitFury. In a couple of months it's going to be 0.7W/GH from Cointerra and Hashfast.

The amount of hash power in reaction to these cheaper and more efficient miners is going to go up although, so the total power consumed is going to be similar in the end.

I assume you mean 10 joules per GigaHash (with ASICs), a joule being a watt-second.

Put another way, such an ASIC doing a GigaHash per second would consume 10W.

It's bad enough that science journalists routinely confuse energy and power units -- let's not do it here on HN!

What’s the source of that - that the majority of Bitcoin miners are using ASICs? From everything I’ve read they’re new, in scarce supply with lead times of up to a year, and expensive.

"The majority of Bitcoin miners are using ASICs" and "the majority of the mining is being done by ASIC miners" are two distinct claims.

Interesting numbers. Still, it's pretty benign compared to the costs and environmental horrors of gold mining.

https://en.wikipedia.org/wiki/Gold_cyanidation

What about the costs of market distortions and mis-allocation of resources when a fiat currency can be inflated with almost no cost?

How much energy (and other resources) were expended in the dot-com boom, and the housing boom? Think of all the houses that were built to replace houses that were perfectly usable, and are now empty? And all the IT equipment built for dot-coms that died after 6 or 12 months?

Probably need to include at least some of the costs of wars and military forces, as those are necessary to maintain the "full faith and credit of the XYZ government" that is the fiat currency.

And definitely need to factor in all of the energy and materials put into huge volumes of cheaply imported products (including oil), that is only possible because fiat currencies enable huge trade imbalances. And all of the energy and materials put into re-creating the manufacturing industry in other countries, as those products become "cheaper", enabled by fiat currencies. And all of the energy used by those products, as their "low" prices make them more pervasive than they would normally be.

You don't have to continue this train of thought too far to realize that the costs of maintaining Bitcoin are minuscule when compared to the true costs of using and maintaining a fiat currency.

Do we really need to dig massive open pits, extract 3 ppm gold molecules, dump the rest, cast the gold into blocks and store them in locked vaults to have gold reserves?

The Blockchain.info number he bases it on comes from:

* Electricity consumption is estimated based on power consumption of 650 Watts per gigahash and electricity price of 15 cent per kilowatt hour. In reality some miners will be more or less efficient.

ASICs (which I believe are the majority of the hashrate nowadays) are more on the order of 10 W/GH (Block Erupter claims 7.5 W/GH/s, the Butterfly Labs 5 GH/s ASIC looks like it's around 6.5 W/GH/s).

First of all, thank you all for the insightful and civilised discussion of my recent post, and than you for pointing out that the number on blockchain.info is off by almost two orders of magnitude. I have updated my post to account for this.

My main argument was actually slightly tangential to this, in that I am arguing that Bitcoin still needs the protection of the state to survive. I have expanded on this topic in my post today http://www.oditorium.com/ou/2013/11/what-does-it-take-to-att...

and back in April I have written a more general post on the various threats to Bitcoin http://www.oditorium.com/ou/2013/04/fifty-ways-to-kill-bitco...

So my central point is: those 50MW (or whatever they will be when the price/hashrate goes up) are simply the costs of maintaining the ledger. As I have noted in the comments to my post, I agree that this is lower than the energy expended handling cash. However, and this is crucial, it is probably about 50MW higher than the cost of maintaining the ledger in a trust-based system that relies on the courts to maintain the integrity of the ledger.

I am all for running an electronic currency, but IMO the hash-rate protection feature of bitcoin is redundant and wasteful.

PS. If anyone reading this know the guys at blockchain.info can you ask them to update their site? It is not helpful if the information on what is arguably the #1 source for bitcoin related data is off by a factor of 50x for the better part of the year now.

I don't get the argument. The power requirement of running the entire world's monetary infrastructure is probably much bigger, no?

Yep that is true. The author discusses this here: http://www.oditorium.com/ou/2013/04/krugman-bitcoin-resource...

That's kind of a weird post - it conflates very different things. Mining burns power to protect against a specific class of attack, not any and all attacks. Bitcoin isn't magically immune to heists, and physical protection may be slightly easier (in that you don't need physical access as frequently) but no less necessary for large wallets.

The financial industry uses up 20+% of US population (albeit this is more than just the currency), but it gives you an idea of how many resources need to be diverted to the process of allocating capital. Many communism nations failed because (along with many many other reasons) there weren't a lot of people figuring out the best allocation of resources.

The current cost of fraud from chargebacks and identity theft from traditional credit cards seems like it would dwarf any eventual power consumption of Bitcoin miners.

You realize that Bitcoin isn't immune to fraud, it just puts the fraud burden on the consumer rather than the merchant, right?

There have already been plenty of Bitcoin frauds; pyramid schemes, merchants disappearing, computers hacked and Bitcoins stolen, entire exchanges disappearing, etc.

Sure; there's a reason I specified chargebacks and identity theft.

These are two areas that are largely solved with a push payment system rather than a pull payment system.

I'm looking forward to seeing how other types of fraud are addressed in the future, but it's fairly predictable we'll see further scams and fraud along the way.

Why do you think that Bitcoin fraud would be handled differently that fiat currency fraud; that is, by offering credit, doing chargebacks, and so on, just like people do with fiat currency?

I mean, it's not like we don't have options like wire transfers for transmitting currency without the possibility of chargebacks; but as you notice, most people choose to use credit cards, which offer the buyer protection, rather than using wire transfers for purchasing goods online.

Bitcoin isn't a particularly good payment system, and I can't imagine that most people will wind up sending Bitcoins directly as payment for the vast majority of purchases. It's more of a replacement of bank transfers or wire transfers, which are used in particular specialized circumstances but not by the average consumer.

Do we really need a power station just for [more popular and frivolous technology than Bitcoin]?

high frequency trading (e.g.) Why not just throttle the markets so that trades are resolved discretely, at a human interval, say at 5 second increments?

How can I be the first one to notice how fucked this math is?

> Blockchain.info estimates the energy usage of Bitcoin miners to be 74,204.76 megawatt hours for a period of 24 hours. Now 74.2 / 24 = 3.1, so this corresponds to a power requirement of 3.1 GW.

I'm sure people have thought about this before, but is there no way for bitcoin mining to be performing useful computational work? Some algorithm that is CPU-intensive and verifiable, but also beneficial to humanity?

Indeed they have https://en.bitcoin.it/wiki/FAQ#Why_don.27t_we_use_calculatio...

You don't need to run a power station if you don't want to. Like people are not forced to make lolcat videos for everyone's amusement. They do so voluntarily. What's the problem?

Has anybody mentioned the power costs of all the different payment methods? I mean, datacenters, helpdesk-offices, sales and marketing. I think Bitcoin does a way better job.

How is that different for bitcoin? Accepting bitcoin payments for purchases still requires datacenters as well as support, sales, and marketing staff given that the products/services being sold don't automagically market, sell, and support themselves just because you accept bitcoin. It might even be worse as a transaction needs to be verified by the rest of the network, not just by a single authority.

Bitcoin would work perfectly well without datacenters. The client is all you need, you can run it on your computer.

I'm talking about the people from VISA that have to store and maintain accountholders data. None of this exists for Bitcoin.

Each bitcoin transaction needs to be verified by a majority of the network in order to mitigate the risk of a fraudulent transaction. Does the energy required to do this exceed Visa's per-transaction energy requirements?

It does seem a bit wasteful, but companies producing ASICs are fighting for lower power consumption. I would expect future power consumption per GH/s to go down.

Watts/(GH/s) will go down because of ASICs, but that will make mining more lucrative and they will have to up the difficulty which raises Watts/(GH/s).

While making GFlops/watt cheaper for all number crunching applications, no?

Not if it's just because we're building bitcoin-specific ASICs. It might help push technology, but it'll also pull brain-power that would otherwise have been focused elsewhere. It's hard to know what the counterfactual would actually look like.

> future power consumption per GH/s to go down.

But you'll need more hashes, so it'll go back up.

The current price of a bitcoin is almost exactly equal to its energy cost, and that's not going to change because if the price was higher people would spend more energy, if lower people will cut back.

Ripple contains a Bitcoin-like cryptocurrency called XRP that only takes a few seconds to confirm transactions and does not require mining:

https://ripple.com/

https://ripple.com/wiki/Introduction_to_Ripple_for_Bitcoiner...

It also allows accounting in any units.

Ripple moves IOUs based on trust, not hard gold-like asset.

XRP is trustless, like Bitcoin.

How much effort is spent on the equivalent administrative tasks for other currencies such as the USD?

"So currently mining Bitcoins requires power levels somewhere between the one produced by the Hoover Dam, and the world’s larges coal fired power station."

Emphasis on "currently"

Disruptive technologies are almost always costly in some way or another until more widely adopted.

Show your work. Why do you expect the power consumption of the Bitcoin network to go down?

When specific costs of a technology decrease it's for reasons specific to that technology. Yes, cell phones have gotten smaller & less expensive (as one example), but that doesn't tell us anything about the power consumption of the Bitcoin network.

what's the conversion rate on unicorns to leprechauns? i'm gonna need a perpetual motion machine to figure this one out...