Quantifying Bitcoin’s Carbon Emissions
Update 3/20/2021: @lmaonade80 on Twitter asked where the carbon emission values for renewable power came from, which highlighted I forgot that bit of data. Here’s a study from 2019 which refutes the USEIA’s claim that hydroelectric power is carbon neutral.
Like most of the things I write, this article contains uncomfortable truths based on publicly available facts and data.
It’s the research and calculations that nobody wants to perform because the results might break the market narrative — a narrative the vocal minority believes drives the market.
I’ve never been a fan of that narrative because it abuses the lack of knowledge in the general public and lack of deep concepts explained in regard to Bitcoin, cryptocurrencies, how they operate, how they are different, and what those operational differences mean. Blame my inner Foucaultian, but that sort of intellectual abuse for profit is a form of exploitation. You’re not going to make the world a better place exploiting people, even if it turns out in the end that you were correct.
Karma happens. I digress.
Denying or by any other means attempting to logically reduce the carbon emissions revealed by this method is the wrong way to go about disproving the calculations I present.
To prove this inaccurate you need either better data or find a flaw in the method.
There are no flaws in the method, except perhaps that I always tend to err on the low side of any assumption. There’s no need to be gratuitous when much of the calculations are derived from other calculations. It’s actually surprisingly simple to follow and calculate it for themself. What’s difficult is collecting enough data to generate ANY sort of meaningful result. That’s a challenge and I attempted to employ the best possible data from the most reliable sources for that data and never is it the same paper/organization/publication. Hopefully this will encourage industry participants to provide just an improved amount of visibility into their operations that better, more accurate calculations can be made. The type of information necessary should be evident to such parties in reading this piece. All data sources are included as links embedded appropriately in the text.
Starting Point: Network Difficulty
The method used to calculate Bitcoin’s carbon emissions for the year of 2020 begins with quantifying an estimate of consumption. Luckily we are not without a starting point as the Bitcoin network itself provides a clue: the mining difficulty. Mining difficulty is a parameter of network operation which increases (or decreases) based on the number of miners competing to solve blocks. This parameter is what keeps Bitcoin blocks being solved roughly every 10 minutes. We are able to use this parameter to find an assumed number of hashes required to solve a block at that difficulty. It’s important to note that direct measurement isn’t possible and any reporting by miners would have to be taken at face value. As each hash is calculated the same way, and each calculation has a quantifiable energy consumption based on the latest in industry ASIC manufacturing advancements, we are able to begin to thumbnail how much energy would be consumed at any given difficulty.
This process begins with the mining difficulty at any given moment (in these calculations, annualized) which reveals the average number of hashes required to find a suitable block solution.
We can fast forward through some complex mathematics and use the values already calculated and peer reviewed from Cambridge University, known as the Cambridge Bitcoin Electricity Consumption Index (CBECI). Not only does the CBECI do a lot of heavy lifting, but it annualizes consumption across adjustments to difficulty as miners enter and leave the network. There are other means by which to calculated electricity consumption, however the CBECI tends to trend a bit higher which we use to reflect non-mining electrical consumption of the network at large. While not completely precise, it certainly provides a more comprehensive picture of network electricity consumption, as it is a logical error to omit non-mining operations.
The annualized electricity consumption of the Bitcoin network as calculated by the CBECI for 2020 was 87.06 tWh (terawatt hours).
Step Two: What Kind of Electricity Production Is Used?
This data is absolutely controversial, but the best visibility one might have is provided by CoinShares 2019 Bitcoin Mining Network Report, which was the latest information provided at the time of this writing.
CoinShares provides data to support the assumption that 74.1% of electricity consumed was from various renewable resources, while 25.9% was from non-renewable resources which are predominantly coal. CoinShares did an excellent job in first cataloging the geographic locations of as many mining farms as is possible, but also to document the electricity production within those regions. It is this location-specific research which provides as accurate a view as possible as to renewable vs non-renewable consumption.
This reveals that 64.51146 tWh of electricity was produced by renewables, whereas 22.54854 tWh was produced by non-renewables.
Step Three: Emissions Per Generation Method
Each power generation method identified in the CoinShares report has a rate of carbon emission per kilowatt which one might obtain from the US Energy Information Administration.
The margin of error at this point is potentially so dramatic that we can cut a statistical corner by using an average emission rate per type: renewable or non-renewable. Using values from our previous calculations we arrive at the facts that renewable consumption emitted 3,412,500,000 pounds of carbon per million terahash per second and non-renewable sources emitted 250,000,000 pounds of carbon per terahash.
The Tale of the Tape
Finally, at long last, we can calculate the amount of annualized carbon emitted by Bitcoin in 2020.
10,483,112,250 renewable + 25,648,964,250 non-renewable = 36,132,076,500 total annualized pounds of carbon
Carbon emissions are expressed in metric tons, which is a simple conversion that results in a total carbon emission for the Bitcoin network in the year 2020 as 16,389,253.71 annualized metric tons.
I’ve been known to move a decimal around by accident every now and then, but these numbers seem to check out. The method is solid based on all available data and any independent calculations should arrive at a total within an acceptable margin of error.
One cannot solve a problem unless one can first quantify it. Now we have the numbers. It’s time to work on solutions.
Comments and feedback welcome.