The Carbon Footprint of Bitcoin | Christian Stoll

so um i would present in the next uh 21 minutes was my limit a paper that i joined he wrote with lena clarsen and willy gallasterfer so lena is a trained economist and uh has trained computer scientists and my background is in environmental policy climate modeling energy system modeling um so the paper i'm presenting goes back to an idea so i saw a presentation about uh the energy consumption of bitcoin back in 2015 at mit and i was wondering what actually the climate implications are because from a climate perspective electricity consumption per se doesn't matter but uh carbon emissions do and so uh the idea was born to actually calculate or estimate the carbon footprint of bitcoin the paper that came out of this research was then uh published last year in uh june in a journal called tool and to be honest we were quite surprised by the media attention that we received for this paper so i think it was a monday when we published the paper and the sunday night before i received an email from the guardian asking for for an interview and i was quite surprised and then on the monday when the paper came out university called me and told me that a german tv station wanted an interview and the entire thing escalated throughout the week so at the end of the week i received an interview request from cnn and the statement on cnn in the end received more than 1.2 billion views and went through media globally in total we or not we but the press office of human munich counted more than 3 billion visits of the results of this paper which is quite surprising that people actually care so much about 0.2 of global emissions but things but let's jump into the paper and have a look at what we actually did so the plan for today would be to first answer the question why is u2 actually matters as i mentioned i'm a trained climate economist and my research focuses on um on climate rather than on bitcoin um second we wanna or i would talk about how bitcoin actually causes co2 um then present the results of our paper so what the current footprint of bitcoin is and if we have some minutes left also look into how reliable such estimates are so why does co2 matter on this chart you see the carbon dioxide concentration in the atmosphere and you can see on this chart that the concentration in the atmosphere has never been that high why does this matter because greenhouse gases cause global warming so you have a greenhouse gas effect the more um greenhouse gases you have in the atmosphere and this warming is also not um equally distributed globally so on this chart number five you see the temperature anomalies over land and oversee and you see that over land uh heats up quicker than the sea and at uh at current levels we are already close to uh 1.5 degrees warming compared to green pre-industrial levels furthermore uh over land there's also a distribution so if you look at and the world map you see that especially around the poles there's even a higher temperature animally which kind of scares me considering how much methane and other stuff is tied up in permafrost so what is the challenge now some probably most of you are aware of the two degree target which was published or communicated after the paris um cop 21 in uh 2015 so the goal that global nations had to limit global warming um below 1.5 or 2 degree warming and on this chart number 7 you see how the emissions would actually have to develop to achieve this goal so on the left side of the chart you see the historic emissions so every year we emit more co2 into the atmosphere and on the right part and in this band you see how the emissions would have to develop if we want to achieve this 1.5 degree warming target and you see that basically we should or would have to reduce emissions immediately and arrive at net zero around mid centuries around 2050. that is quite a challenge um and if you look at what is currently happening so during the corona pandemic emissions have dropped since the largest drops in second world war but still we are still on a level comparable to i think 2012-ish so um that kind of shows how big the challenge is and we won't solve it by just behavior change but technology will play a major role in in this challenge so this is kind of the background um why co2 actually matters now you might be wondering what has bitcoin to do with all of that um easy answer first of all bitcoin how it's made and probably most of you know how it works in detail so the validation algorithm that validates transactions and ownership requires a computational intensive process and for this computational intensive process you need a specific hardware so typically asic mining devices are used recently and those consume electricity the more you have of those so that's a picture from a mining farm in iceland the more you have obviously the more electricity you need to run those and electricity depending on the source uh can translate into emissions so and the next step question is then what is actually the emissions caused by the electricity consumption of these mining activities and our research approach uh was the following so first of all we wanted to calculate the power consumption of bitcoin and then multiply it with the carbon intensity of the electricity used to derive the calm footprint so it looks quite simple but the devil is in the detail of collecting all the data just needed to derive this result so for instance key inputs to derive the power consumption or the hardware in use and the manufacturers of mining devices are typically quite sensitive with that data and you also need to know how efficient these operations are so depending on the size of mining if you have a small mining farm compared to a large one you have auxiliary losses or you have less auxiliary losses and on the carbon intensity side that's the tricky part you need to understand which carbon emission factor actually to apply for the electricity so for the electricity consumption i won't go through that in detail but this was our basic approach to calculate an upper and a lower bound so we assumed for a upper bound of electricity we assumed that the miners use all their revenues to buy electricity and for a lower bound we assumed all miners to use the most efficient hardware and we also calculated a best guess which included further assumptions and which we consider as our our best guest scenario so this was the approach it's a three-fold approach to estimate electricity consumption and um there were a lot of information which we needed to actually do so and we were quite lucky because during the time we conducted the study three major producers of a6 announced their ipos so their initial public offerings which required them to publish data which they hadn't done previously and so based on their ipo filings we could estimate which hardware they actually had sold which gave us information on the efficiency of the hardware used in the network which allowed us to derive the electricity consumption we had a few further inputs to do so so one was um the pool size and so from pool shares we derived the size of the single miners and we also conducted a bunch of interviews to understand how mining operations actually work so that's that's the screenshots from a data scraping we did uh on slushpool which probably some of you know that is the mining pool and we recorded the hash power contributed by single users to classify miners into three size buckets so small miners medium miners and large miners and depending on the size of mining operations we included pues a power utility effectiveness so meaning auxiliary losses so simple example if you mine at home you can open up your window you don't need additional cooling if you have a large mining farm you need cooling and if you have a very large mining farm your cooling will be more efficient than in a medium-sized mining farm so this was the basic idea behind that one so this chart shows uh the distribution of the pools in the bitcoin network at the time end of 2018 and um we classified these pools to say for instance private pools were then classified as large-scale mining because they were typically privately owned mining pools and that was how we derived electricity consumption which you see on that chart so these are the the lower bounds so the technological lower bound the economic upper limit and our best guess estimates which was uh 45.8 terawatt hours annually at that point in time end of 2018. so now the second part of the calculation interesting one in order to translate electricity consumption into carbon emissions you need to know where your electricity is consumed and we applied we calculate three scenarios to actually derive the geographic footprint of bitcoin mining so our first approach was based on a server a piece the pool server ip so we recorded and monitored the data that was provided by mining pools to derive a distribution between asia europe and america second we used a device ips so we used a iot search engine called showdown io to actually localize mining ips and so ip addresses of mining devices with a certain configuration and our third approach which we uh didn't use in the paper in the end was to set up our own node in the network and record um the blocks that were relayed so based on the scenario that we derived we then wanted to calculate the carbon emissions and this one is actually trickier so it looks quite simple so to calculate the carbon emissions from electricity consumption you need to know how carbon intensive your electricity is and here the challenge begins depending on which lens you take so you can say for instance i mine next to a renewable power source so i'm mining next to a wind farm i am renewable or you could say from a system perspective i am consuming the average carbon intensity of the electricity in the respective grid or you could say i'm causing with my mining additional load and the marginal emissions that i cause are the coal fire power plants the last one and the merit order that is actually added to fulfill my load and that is quite tricky and that's also where the discussion starts and there are a lot of guesses out there which deviate quite a lot and from so we took the system perspective and used an average emission factor in the end to derive our results which was then 22 megaton of co2 annually which is comparable to a major city so i think with kansas city as an example that emits a similar amount of carbon so how reliable are such estimates as i already mentioned the challenge is to get the emission factor right because if you assume mining next to a renewable energy source is uh carbon-free then your carbon footprint is much lower if you take the uh perspective of mining is adding loads to the to the grid then you end up with a much higher carbon intensity so that's quite quite a challenging one and you could also ask so what now you calculated the carbon emissions so what um there are much larger sources of carbon emissions definitely um and this research was also the starting point to look into other data centers that um require more electricity etc so it's i think it's a nice example to see that technology technological innovation uh also requires looking at externalities and um the conclusion here is not that we should ban bitcoin or regulate mining or something like that but it shows the general problem so from an economic perspective the most cost effective way would be to to use a carbon price to actually internalize the externalities that reside from these carbon emissions and the problem is not bitcoin percy but the problem as i mentioned at the beginning of my talk is the general greenhouse gas emission level having said that i think i'm quite close to the 21 minutes and i would stop my presentation at that point thanks