Verify, don’t trust

It’s no secret that a good part of the Bitcoin community hates Digiconomist for contributing to exposing the digital currency’s extreme environmental impact.

Every now and then, this leads to an ambitious attempt to discredit the Bitcoin Energy Consumption Index and its associated conclusions, hoping to put the genie back in the bottle. An example of this is an article that was published the other day titled “The reports of bitcoin environmental damage are garbage”. Regardless of the quality of the content, these stories are eagerly hyped up by the Bitcoin community. Ironically, the best defense against such articles is Bitcoin’s own motto and raison d’etre: verify, don’t trust.

Update: at the time of publication a new article titled “Renewable Energy Will Not Solve Bitcoin’s Sustainability Problem” was undergoing peer-review via the energy journal Joule. The content could not be disclosed at the time, but addresses several of the remaining arguments that were initially left untouched in this post.

Verification approach

The numbers and conclusions presented by the Bitcoin Energy Consumption Index are easily verifiable with a simple back-of-the-envelope calculation, that has been used by both critics and proponents of the digital currency since 2014. This approach requires one to know the total Bitcoin network computational power/hashrate, and the computational power/hashrate of the machines available to miners along with their respective energy use. Both variables can easily be obtained from public sources. We can observe that the current Bitcoin network computational power/hashrate equals 42.6 EH/s per today (January 31, 2019). Likewise, we can find that the most popular (and most efficient) machine over the past few years (Bitmain’s Antminer S9) produces around 13 TH/s at an energy expense of 1300W. Miners suggest this advertised performance is slightly optimistic, and the average performance is closer to 1500W per machine. Either way, the network could contain 3,276,923 of these machines (42.6 EH/s divided by 13 TH/s). Multiplied with the energy use of 1500 watts per machine, we find that the network runs on 4.915 gigawatts. This translates to an annualized energy consumption of 43.06 TWh.

Flawed criticism

The former reveals the first mistakes in linked criticism. The author optimistically argues that “the Bitcoin network uses maybe between 25–35 TWh per year” without further substantiation (even though a lack of substantiation is a primary point of critique). In fact, the criticism completely ignores the former approach, despite quoting Jonathan Koomey (who has previously argued that this bottom-up approach is the only correct approach to determining Bitcoin’s energy consumption), and despite the fact that this approach is represented in the Bitcoin Energy Consumption Index (Minimum TWh per year line). This at least raises the appearance of cherry picking and the pot calling the kettle black.

Of course, the observant reader will notice that the minimum TWh per year in the Bitcoin Energy Consumption Index equals 47.01 TWh per year rather than 43.06 TWh. But as mentioned in the introduction of the page, this number includes both the Bitcoin and Bitcoin Cash network. The latter network employs the exact same mining algorithm (SHA256), and Antminer S9 machines can be used to mine Bitcoin Cash as well. The combined network hashrate is reflected in the key statistics table, and comes down to 46.5 EH/s. This is equal to 3,576,923 Antminer S9 machines with a total consumption of 47 TWh per year. One might prefer to consider only the Bitcoin Core network, but as shown the difference is minimal (and therefore wouldn’t change the ultimate conclusion).

Riddled with errors

Likewise, the observant reader would notice that the current economics-based estimated reflected in the Bitcoin Energy Consumption Index is currently exactly equal to the estimate produced by the previous approach. This is the natural economic outcome under adverse market conditions (the Bitcoin price has been declining throughout 2018) where only the most efficient mining machines can survive. Rather than noting this, the author of the criticism only notes “since the price of bitcoin has fallen significantly, based on the assumption on his site (which is a fixed cost to income ratio), the costs to miners must also have fallen equivalently”. This is where the article resorts to blatantly false statements, as the cost-to-income-ratio varies and is transparently disclosed in the key statistics table on a daily basis.

The author also asserts that the premise on which the Bitcoin Energy Consumption Index is built (miner income and costs being related) “is wrong”. This is rather surprising, as it shouldn’t take long to explain that a miner earning $10,000 per day can afford to spend more on electricity than a miner earning $1,000 per day. This is basic economics, and used in alternative papers as well like “A Cost of Production Model for Bitcoin” and “The Cost & Sustainability of Bitcoin”. Yet, the author seems to imply that these assumptions are solely used in the Bitcoin Energy Consumption Index. Although giving that much credit is appreciated, this would obviously be a bit too much. It’s also not as difficult as the author suggests to find independent articles using the machine-based approach to estimate Bitcoin’s electricity consumption, e.g. this piece by Tim Swanson, or this academic article by Max Krause and Thabet Tolaymat published in Nature (November 2018).  

The criticism continues to get worse, as it claims that the methodology in the Bitcoin Energy Consumption Index would somehow lead to the conclusion that “the Bitcoin network would consume more power than was produced by the sun”. But as mentioned in the “Forecasting” paragraph on the Index page: “Based on 100% of revenues already being used to cover electricity expenses, the Energy Consumption Index would thus predict little change in Bitcoin’s energy consumption”. This is where the mistakes are becoming rather painful, and could have easily been avoided with some peer review before publishing.

Carbon footprint

The author’s ultimate conclusion is that “crypto’s carbon footprint is extremely moderate for being the world’s largest computer network”. Herein the author ignores that even with his own optimistic 25 TWh per year energy consumption estimate, the Bitcoin network still has an average per transaction electricity footprint of 300+ KWh (processing 81 million transactions in the whole of 2018). Even with a bizarrely optimistic emission factor of 10g of CO2 per KWh (note that pure hydropower may have an emission factor of 4g/KWh), that’s still a carbon footprint of 3 kilograms of CO2 per transaction. By comparison, a VISA transaction has a carbon footprint of 0.4 grams (a factor 7,500 difference).

Conclusion

Beyond doubt, new criticisms that try to put lipstick on a pig will continue to appear. Bitcoin’s dirty secret has, however, been made public – and there is no way to make it go away. As shown in this article, no amount of optimism can make Bitcoin look like an environmental non-issue. Moreover, anyone can easily verify this conclusion with a rather simple back-of-the-envelope approach. It’s about time the Bitcoin community starts investing their time and effort in seriously addressing the problem instead. In the meanwhile: verify, don’t trust.