Crypto arguments from the head, the heart and the hands
2022-06-07 (updated 2023-01-13)
My previous post on the climate-impact on cryptocurrencies garnered some attention on reddit and Hacker News, and I had originally planned a follow-up to address some of the criticisms I received in those comments. I never finalized the article1, but it has been sitting in my drafts long enough now that I decided to post the outline of it anyway2.
The title and structure of this post is a reference to the idea laid out in Buster Benson's book Why Are We Yelling, that most arguments can be categorized in three categories: those of the head, heart and hands. In order, those arguments boil down to disagreements about the truth, the meaning, and the usefulness of the argument.
Arguments of the Head (Truth)
The correct calculation yields ~2,000 kWh per block, not transaction
I addressed one of these comments directly in the HackerNews thread, but for completeness, here's how you can verify yourself that this argument is false.
We start by looking at the total energy consumption of the Bitcoin network as a whole. As of the time of writing, that figure is 200 TWh per year. 200 terawatt-hours is the same as 200,000,000,000 (200 billion) kWh. Next, we find the number of Bitcoin blocks mined each day. Since the Bitcoin network is designed to yield a block at a rate of about one block every 10 minutes, we expect this value to be around (1 block / 10 minutes) * (60 minutes / hour) * (24 hours / day) * (365 days / year) = 52,560 blocks / year. Empirically, we can also see that the number of blocks mined daily tends to track very closely to the theoretical average. Dividing our two figures, we find that (200,000,000,000 kWh / year) / (52,560 blocks / year) ≈ 3,800,000 kWh / block, which is about 1,900 times larger than the figure I quoted for the effective kWh per transaction.
To bring the calculation all the way home, we can calculate the kWh per transaction by dividing our kWh per block figure by the number of transactions per block. There is no theoretical average for the number of transactions in a block since a block is limited by memory (1 MB) rather than number of transactions, but we can see empirically that the number of transactions per block tends to hover around 1,900. (3,800,000 kWh / block) / (1,900 transactions / block) = 2,000 kWh / transaction, which is very close to the figure I used in the original post.
Bitcoin uses mostly renewable energy
According to Digiconomist, about 25.1% of the energy used to mine Bitcoin comes from sustainable sources. This is the lowest percentage for multiple years, partially due to many miners having to move their operation from China to other countries after China cracked down on cryptocurrencies in 2021.
25% may still seem like a large number, given that many sources for sources of global energy seem to indicate that the global average share of sustainable energy is much lower. However, most of those statistic include energy usage for applications that are not powered by an electric grid such as shipping and airlines. A more applicable comparison is to look at what percentage of energy consumed through the electrical grid comes from renewable sources. Globally, that percentage is 26.4%3.
Bitcoin uses less energy than ATMs
This one we can also address with some quick maths. An standard ATM draws between 60W and 145W. Let's go with the larger number to be conservative: that then gives us an energy usage of 145Wh per hour per ATM. Yearly, that is (145Wh / hour) * (24 hours / day) * (365 days / year) = 1,270,200 Wh / year, or 1,270 kWh / year4. There's about 3.2 million ATMs in the world, so assuming that those machines are running 24/7, that gives us a total energy usage of 1,270 kWh / (ATM * year) * 3.2m ATMs = 4,060,000,000 kWh / year, or about 4 TWh / year. So in our worst-case scenario, ATMs use about 50 times less energy than Bitcoin yearly.
Bitcoin uses less energy that Christmas lights
This would indeed be quite alarming if it was true. The comment I link to provided a source from oral testimony in front of the U.S. House of Representatives Financial Services committee, which may give the statistic the veneer of being authoritative. However, it's important to remember that a wide range of people get asked to give testimony in front of Congress for an equally wide variety of reasons. In this case, the speaker was Brian Brooks, CEO of the crypto company Bitfury, who claimed in his testimony that "bitcoin’s total global energy usage of about 188 terawatts is somewhat less than the total annual energy usage of Christmas lights (around 201 terawatts)". There's no source for his claim that Christmas lights use around 201 TWh (which is what I assume he means by terawatts) per year, but I was able to find an estimate of the energy usage of various forms of lighting in the United States from the Department of Energy. In this report, the DoE reports that as of 2007, around 6.63 TWh worth of energy is used to power "Decorative Holiday Lights" each year. They also helpfully note that at that time, around 5.2% of those lights were LEDs, which are much more energy efficient than traditional holiday lights.
It is very likely that the proportion of LED Christmas lights is much higher now than it was in 2007 given that LED usage as a whole has gone from virtually 0% to around 50% of the market. This means that the total energy use from Christmas lights has likely gone down in the past 15 years, but for a worst-case analysis let's assume that the energy use from Christmas lights remains the same in the U.S. Let's also assume that the U.S. is a representative country for estimating the amount of energy used per capita on Christmas lights (another assumption that will most likely overestimate the consumption as many other countries use little to no Christmas lights). The U.S. has around 334m people, and the world population is around 7,952m, so in our worst-case analysis, we arrive at the figure of 6.63 * 7,952 / 334 = 158 TWh / year. A more conservative estimate where the U.S. uses twice as many decorative lights per capita as the average country and where market penetration for LEDs in Christmas lights has reached 50% would yield an estimate of around 40 TWh, around 1/5 of Mr. Brooks' estimate. SEB CHECK THIS CALCULATION
Bitcoin use less energy than data centers
This one is pretty close. The best source I could find indicates that data centers consume about 190 TWh / year, with a clear trend in decreasing energy consumption from traditional data centers, and increasing consumption from cloud centers. Interestingly, the overall electricity usage from data centers is now decreasing year-by-year even as cloud computing is clearly on the rise, mainly due to the move to hyperscale server centers, which tend to be more efficient at the workload level. Bitcoin electricity consumption is already past 190 TWh / year and has consistently increased, so though data centers may have used more electricity in the past, that is no longer true. It's also worth noting that some of the electricity consumption from data centers may also be supporting cryptocurrency mining, so the gap between the two may be even larger.
There's also a usefulness argument here, as I would wager that the commenters who wanted to see this comparison would also concede that there is a lot of usefulness to the types of workloads that are being run in data centers. My guess is that the commenters expected data center electricity consumption to far exceed that of cryptocurrencies and so they meant to point out that there is a lot of waste in general computation as well.
General confusion about energy calculation
Though the comment I link to in this section header does not add
Arguments of the Hands (Usefulness)
Bitcoin uses less energy than X
I addressed some specific examples of this type of question in Bitcoin uses less power than data centers, Bitcoin uses less power than ATMs and Bitcoin uses less power than Christmas lights. In those two cases, we found that the statistics quoted in the comments were false, and we can consider them arguments of the head. However, given that aggregate energy usage for cryptocurrencies is still less than a percent there are of course a lot of things in the world that use more energy than Bitcoin.
Calculating energy usage per transaction is wrong as transaction rate for Bitcoin is constant
[The article does not account for transactions on the Bitcoin lightning network]
The title talks about cryptocurrencies, but the article only addresses Bitcoin
Mostly true, though I did try to give a brief explanation in footnote 1. Here's the slightly longer argument: Let's pretend that humanity collectively decided to wind down the Bitcoin experiment and every single Bitcoin holder sold their Bitcoins to reinvest in other cryptocurrencies. If these investors reallocated their money in such a way to not affect the relative weights of market-capitalization for the other cryptocurrencies, they would then buy into the cryptocurrencies based on the ex-Bitcoin asset-weighted mix of all other cryptocurrencies. For example, if Ethereum currently accounts for about 10% of the global cryptocurrency market-cap and Bitcoin accounts for around 80%5, we would now expect Ethereum to account for 10% / (100% - 80%) = 50% of the total cryptocurrency market capitalization. If each cryptocurrency also kept its same power efficiency, we'd expect the electricity usage for each currency to scale up by an equal amount. The full calculation using the numbers from the original article finds that energy usage would increase by 31.61% / 20.31% - 1 = 55.6%.
Even though there are cryptocurrencies that are more and less efficient than Bitcoin from an energy perspective, this calculation shows that the asset-weighted mix of all cryptocurrencies sans Bitcoin has worse energy efficiency than Bitcoin itself, and so my arguments from the previous article also hold for the cryptocurrency ecosystem as a whole.
Only proof-of-work cryptocurrencies are bad for the environment
Cryptocurrencies use very little energy on an aggregate scale compared to other industries
It's more important to consider energy production than consumption
The global banking system is worse for the climate than Bitcoin
Arguments of the Heart (Meaning)
Corporations are worse polluters than individuals
 Though I would like to credit my non-posting as a sign of personal growth and a proof that I now know not to pick fights on the internet, in truth it was more of a victim of shiny-object syndrome. ↩︎
 In good news, the energy-usage of most crypto-currencies have plummeted since I originally drafted the article above. Not necessarily due to technological improvements though - just an artifact of the cratering crypto-prices. Hooray! ↩︎
 Though the Our World in Data link I used reports the non-fossil fuel percentage energy mix as 36.7%, that mix includes nuclear power. I'm all for nuclear power, but since the Digiconomist definition excluded nuclear from their sustainable sources in arriving at their figure, I do the same here to arrive at the 26.4% figure. ↩︎
 Note that right off the bat, we see that one Bitcoin transaction consumes about as much energy as 2 years worth worst-case usage of an ATM, so I'd say this argument is not off to a great start. ↩︎
 These weightings are based on the study linked in the original article, which is re-linked here for easier reference. ↩︎
 Funnily enough, it doesn't appear like Brian Brooks actually read the section touching on Christmas lights in the hearing for which he submitted his testimony. In the video recording from the hearing, he ends his opening statement in the paragraph before he talks about the sustainability issues of Bitcoin, at the 30:30 minute mark. There's nothing abnormal about this: witnesses in congressional hearings often have limited time to provide their opening statements and can submit written statements that go into more detail than what their oral testimony would allow, but I still find it kind of funny that this very easily-disprovable argument seems to have garnered attention even though the main source of it didn't even address it in his opening arguments.
For complete transparency, I did not watch the full 4.5 hours of the Congressional hearing, so my apologies to Mr. Brooks if he did in fact voice this part of his testimony at a later stage in the hearing, though I still stand by the fact that he grossly overestimated his electricity consumption figure for Christmas lights. ↩︎