Bitcoins Should Be Called BTUcoins, and That’s a Problem
Because electricity is not priced optimally, we all pay for excessive Bitcoin mining.
(This post is coauthored with Catherine Wolfram)
You may be reading this blog post in your dorm room, while your GPU is mining for Bitcoin using that zero marginal cost (nerd-speak for “free”) electricity coming out of the wall. If you did not understand that sentence, let’s back up. Bitcoin is one of a number of virtual currencies, which can be used to pay for lots of things you can buy on the internet. Bitcoin use cannot be traced back to you, which makes it super popular for shady transactions.
While neither of us is a monetary economist, we learned in graduate school that currencies are usually issued by governments. In some places, these currencies are backed up by valuable physical resources like gold. Traditionally every currency has a physical manifestation, which we refer to as money, but that actually makes up a tiny share of the money supply. The rest is electronic and tracked by commercial banks and the federal reserve. The key is that the small pieces of paper currency have value because the government prints a limited number of them, can validate that they are real, and promises to honor non-fake currency. This makes the whole system work.
Bitcoin’s scarcity, validation, and hence value, comes from solving complex math problems. And, complex math problems require lots of computing power. Turns out that computing power is exactly what’s found in that nice computer (= gaming rack) your parents bought for your term papers. On a larger scale, it’s in dedicated servers.
There have been a number of articles recently about just how much energy all these complex math problems use. This site claims that cryptocurrency mining (i.e., solving the complex math problems) today accounts for more annual energy than the country of Bulgaria. This site helps you figure out how much money you can make mining Bitcoins. The basic calculation seems to suggest that the less you pay for electricity, the more money you will make.
The fundamental problem is that worldwide electricity prices are not designed to sustain this onslaught of new Bitcoin miners.
- The most obvious problem is that prices do not (for the most part) reflect the local or global environmental damages from electricity generation. Specifically, those electrons powering your computer, and generating “coin”, more than likely come from a fossil-fuel power plant that’s contributing to global warming and local air pollution. This has been a long-running theme on this blog, but the explosive growth of Bitcoin mining promises to really make us cringe.
- Another problem is that, unlike most goods and services, the prices for electricity are set by regulators (read politicians) and not by markets. This point is made ultra-clear in the developing world, where power theft can be a problem. For example, this paper, shows that power theft in India increases around election time. Politicians are not very motivated to cut down on theft, even though it’s bankrupting the local utilities. This is problematic even if you have millions of rural households using free electricity to power a single lightbulb. It is a real problem if someone with a Bitcoin server farm moves in to access the free electricity.
Politicians usually have at least 23 goals they’re trying to achieve with electricity prices — including fairness and gradual changes, if there are any at all — all while guaranteeing that elderly customers can run life-support systems and that we can pay for pet programs.
The mish-mash of electricity prices that arises doesn’t lend itself well to something like Bitcoin mining.
- In addition to power theft, there are instances where the price of the electricity is zero for more benign reasons. For example, college students living in dorms often get electricity as part of their housing package, which means that students don’t pay more if they use more. On the margin, then, their electricity is free. The same for many apartment renters, as in many older, large buildings you don’t have individually metered apartments.
These problems are not unique to Bitcoin mining. If electricity prices do not reflect marginal costs for environmental, political or other reasons, it’s bad for the overall economy. But, most humungous electricity consumers – think aluminum smelters – are much more centralized and aren’t growing at the same astronomical pace as cryptocurrency mining, so it’s easier to track their use.
While it is impossible for us to track the physical locations where Bitcoin is generated, we hypothesize that it is most likely generated in American dorm rooms and regions with cheap electricity (e.g., India and China and Mexico) and less likely in places with expensive electricity like Denmark, Germany and Hawaii. If you can find prime numbers, you are smart enough to figure out that you should run computers in the places where electricity is cheapest. While China and India are working hard on improving environmental regulations (which might make up for some of Scott Pruitt’s war on the environment), there is a long way to go to get to electricity pricing that reflects its full cost to society, including pollution.
Put differently, the Winkelvoss twins, who seem to have redeemed their Facebook losses with big Bitcoin wins, are indirectly helping to drown polar bears, kill trees and guarantee that utilities in the developing world move towards bankruptcy.
But, you say, what is the alternative? We had a lively discussion at the EI lunch table about this. Max likes his currency traceable and issued by a government. This allows for things like taxation and easier monitoring of illegal transactions. Some graduate students called him an old grouch (which is correct). But, after all, we are a bunch of energy economists and eventually got to thinking about the energy implications. The question we could not answer on the spot was whether maybe the energy penalty for Bitcoin is not so much bigger than that for cash or credit cards.
So, we did some blogmath. The calculations can be found here. Turns out that a banknote uses about 0.8 watt-hours per transaction. To put this in perspective, a 60 watt lightbulb (which is your average incandescent – now illegal in many places) uses, well, 60 watt-hours if left on for an hour. A credit card transaction uses about 7 watt hours per transaction. So, about eight times more energy than the cash transaction. Now, get ready for it. A Bitcoin transaction uses 100,000 watt-hours per transaction. That is 100kWh, so about the equivalent of 115,000 cash transaction. This is why the annual electricity consumption is surpassing Bulgaria, and by some estimates, on track to exceed the US by mid 2019. If we use a number of external costs from electricity consumption of 6 cents per kWh which assumes the US fuel mix, which is much cleaner than e.g. China’s, we are talking about almost $2.5 billion worth of free riding. This is a conservative number, with the true number potentially much higher.
So, what on earth are we doing? We are destroying the environment in order to subsidize an untraceable currency which enables and encourages tax evasion and other illicit behaviors. We thought that refrigerated pet food was going to take the cake for dumbest and most frivolous new use of electricity. Move aside kitty, here comes the itty-bitty coin to take the prize!
Suggested citation: Auffhammer, Maximilian & Wolfram, Catherine. “Bitcoins Should Be Called BTUcoins, and That’s a Problem.” Energy Institute Blog, UC Berkeley, January 8, 2018, https://energyathaas.wordpress.com/2018/01/08/bitcoins-should-be-called-btucoins-and-thats-a-problem/
Maximilian Auffhammer View All
Maximilian Auffhammer is the George Pardee Professor of International Sustainable Development at the University of California Berkeley. His fields of expertise are environmental and energy economics, with a specific focus on the impacts and regulation of climate change and air pollution.
Wind-farms and solar panels are expected to cleanly power cities and nations. Why can’t they be expected to cleanly power bitcoin miners?
So I wonder, did defenders of refrigerated pet food offer you similar criticism? Regardless, the problem is inefficient energy pricing, not the use to which we put energy. I do not believe it is my place to judge the appropriateness of how we use energy more than it is my place to judge the appropriateness of how we use any other resource. Price the resources efficiently and we economists should be happy. So, instead of judging bitcoin mining or refrigerated pet food, I judge the politicians, the regulators and the stakeholders who not just allow inefficient pricing, but openly advocate for inefficient pricing. Instead of some readers calling for a movement away from electronic transactions and asking our economy to rely on paper transactions, I prefer that we recognize the incredible ways that energy has improved our lives and push to price it appropriately.
I wish my friends and colleagues would do the same. It seems that the world of policy dialogue is filled with those who would judge how we use energy. To say that this misses the point should be obvious. Sadly, it is not. Certainly putting bitcoin in the headline attracted readers who likely do not regularly read this blog. In amusement, I noted that on my browser it also attracted a small advertisement for “cloud based bitcoin mining”. (I am struck by the coincidence that the one person I know who is in the coin mining business is also in the internet advertising business.) So if a topical, and dare I say sexy, subject such as bitcoin mining sheds more light on bad pricing practices, then all is good. However if, as I suspect, this dialogue provides regulators with flawed policy reasons to aim at how we use energy, rather than efficiently pricing energy, then a disservice has been done.
I’d also love to learn whether Max has observed any aluminum smelting in dorm rooms, given the zero marginal cost energy.
You should come visit the sweet miniature smelter I have set up in my office here!
You should use it to power lawyers’ computers. They get a higher yield per kwh.
A few corrections:
(1) Energy consumption is a mutable design decision (made differently by different currencies): The computational work that bitcoin miners do is computing hashes (unique numerical signatures of far larger volumes of underlying data) on bundles of transactions with arbitrary properties (i.e. a certain number of trailing zeros in the hash) that unlock coins for the miners. This is the mechanism that pays people to verify transactions (a problem that must be solved in a distributed system of transactions). It is not related to finding prime numbers (a casual implication in your post) and it is not technically necessary for the system of transactions to work. The issue is how to control the rate of release (discovery?) of bitcoins, which are regulated by the underlying software to be issued at an ever decreasing rate, asymptotically approaching a static final number. Currently, there is a “hardness” factor that scales the difficulty of the hashing requirements as more miners enter the system to a level that roughly produces the expected number of new coins. This method of throttling coin production is, but does not NEED to be computationally or energy intensive. It is an arbitrary design decision and one that could be changed. There are already the less computationally intensive solutions employed by other currencies. Even a fair/verifiable lottery would work. So, my opinion, it is a mistake to extrapolate under the assumption that the characteristics of the system will not change and it is dead wrong to assume that they can’t.
(2) People are working on it: One major issue is the tight coupling between verifying transaction and mining coins. The proposed Lighting Network would decouple transactions from the work of adding blocks to the chain and would go a long way to reducing your energy per transaction metric: https://lightning.network/.
In the analogy between crypto-currency technologies today and the internet in the late 1990’s, Bitcoin is something like AOL. There is a bubble and there will be turnover and collapse, but there is also real value and we can look forward to innovation and improvements. As scholars of energy, you should also consider the parallels between bitcoin energy consumption forecasts and the wildly incorrect forecasts for computational energy consumption driven by the growth of the early internet. An example from our backyard, many buildings on UC Berkeley’s campus were built with jaw dropping HVAC system capacities to prepare for consumption (and waste heat) from computing loads that never materialized, leaving a legacy of oversized and inefficient cooling systems that violate UC Berkely’s policy of avoiding mechanical cooling whenever possible and (in some cases) require the addition of steam heat to ensure that they are operating above the point where the motors stall. Credulously accepting crazy forecasts can also lead to perverse environmental outcomes.
(3) If it can go up, it can go down: Due to the incentives for mining only when the payments exceed the marginal costs, you should also rest assured that the energy consumption of bitcoin will drop with prices just as fast as it increased. A forecast of ever increasing energy consumption is very similar to a forecast of ever increasing prices. If you think the prices will some day go down, then the price correction alone will significantly reduce power demands.
(4) Bitcoin is traceable and the majority of its users are not and have never been criminals: Do you really believe that a system with every transaction available for public verification and scrutiny can effectively mask user identities? If you do, then you must explain why the NSA spent so much money on its bulk surveillance system that “only” collected phone call metadata. In case there is any doubt, the issue has been covered in Science: http://www.sciencemag.org/news/2016/03/why-criminals-cant-hide-behind-bitcoin, and empirically, criminals have by and large abandoned it for currencies that get less scrutiny and work harder to mask identities https://www.bloomberg.com/news/articles/2018-01-02/criminal-underworld-is-dropping-bitcoin-for-another-currency.
(5) You’ve buried the lede: I would think that energy economists would be at least as interested in the possibilities and pitfalls of a public ledger of distributed grid-edge energy transactions powered by blockchain technologies that could enable market-driven procurement of grid services, the demand-side assistance to the integration of renewables, major shifts in the roles of utilities and regulators, etc. If every there was a bunch of hype calling out to be addressed by energy economists capable of recognizing wheat and chaff, that is it!
Who knows, maybe PV-proportional crypto-currency mining will some day play a vital role in soaking up all that mid-day PV we’ve been wringing our hands over, lest it break our power markets.
The energy cost of both a credit card transaction and a bit coin transaction are perfect reasons for not going to a cashless society. Bit coin, like cash, is difficult to trace, but its energy cost is far greater than cash. One would guess that a far greater number of transactions in bit coin could be illegal. I understand the arguments for a cashless society, but having every transaction traceable by anyone who wants to trace it a privacy issue. It appears that it is also and energy issue as well. I would assume that a debit card transaction is like a credit card transaction.
Fortunately such a profound lack of insight is not common in this blog. The authors clearly did not read the article posted at the link https://news.bitcoin.com/analysts-debate-power-consumed-per-bitcoin-transaction/ or anything else in some information that could be pulled out of context to support their conjecture. I actually have real work to do so I won’t go any further and just say I’m profoundly disappointed in this post
I am sorry our lack of insight is distracting. I suggest unsubscribing.
The article you reference still shows an estimated usage about 45% of the estimate cited in this blog, which is still a big number–equivalent to 50,000 cash transactions instead.
In fact, we did read the article you cite. You’ll see that the spreadsheet that we link to includes it as a reference. It suggests lower usage, but by a factor of about 3, not by an order of magnitude.
It seems like bitcoins are most used as a store of value rather than for transactions. If that’s the case, then dividing the energy used to create a bitcoin by the number of transactions is misleading, especially when comparing the result to the energy per transaction of instruments that are primarily used for transactions. It might be more appropriate to compare the energy used to create a bitcoin to the energy used to mine and mint a gold coin.
I’d give up on the illegal activity as a reason to ban BTC etc. (we should ban $100 bills and London if we’re worried about money laundering).
You’re right to worry about energy consumption, and there are technical ways to do this without undermining the value of bitcoin. (The most common example is a move from proof of work to proof of stake, but that switch would change power dynamics radically; a more likely solution would switch hashing from something GPU intensive to something, say, bandwidth intensive.) These ways will be pursued when it’s in the interest of 51% of the miners (game theory).
Going forward, I suggest you think of crypto markets in terms of 100% (unregulated) capitalism. Yes, it’s “wasting resources” but so is open sea fishing and fossil fuel consumption. Have we solved those issues yet?