A few days back I posted a rather casual statement about the fact that I thought Bitcoin’s energy consumption to be harmful for the environment. This attracted quite a bit of criticism from crypto fans who accused me of spreading false information, because the energetic costs of cryptocurrencies allegedly compared quite favourably with the traditional banking system with all its baggage of servers, branch offices and the like.
My crypto-critics did not do a particularly good job at articulating their points credibly, but I don’t take such accusations lightly, so over the weekend I set out to do some research on my own.
It is important to note that I do not have an opinion on the evils of cryptocurrencies, and I have lots of personal reservations over the “good deeds” performed by banks.
Finding hard, reliable data on this topic has proven to be a challenge – however the comparison between “bitcoin” and “the banking system” from an energy consumption standpoint has been attempted several times, so I will summarize these attempts here.
Banking system energy consumption
The most common approach divides the banking system into three components, branch offices, servers and ATMs; estimating the consumption of each of these and totaling them yields about 100 TWh of energy (examples of such estimates here, here and here). I was surorised at how the language seems to be exactly the same in all cases, making me wonder if I am not simply seeing one same original estimate reported several times.
Of these, 26 TWh are attributed to servers, a number that does not conflict with another independent estimate which pegs the consumption of all data centers in the world at 205 TWh (Data Centers Knowledge, feb 2020). Banks have big data centers, but so do tech giants such as Apple, Google, Microsoft, Amazon, Facebook or Twitter just to name a few, so attributing 10% of the all data center consumption to banks seems reasonable.
The DOE moreover estimates that the datacenter energy consumption for the US alone to be 70 TWh in 2020 which again seems reasonable vs the global total of 205. (DOE – OSTI, feb 2016)
Cryptocurrencies energy consumption
This also has been the subject of many studies, coming to wildly different totals. The Cambridge Bitcoin Electricity Consumption Index estimates 138 TWh while the Bitcoin Energy Consumption Index estimates around 76 TWh, while other estimates go as low as 39 TWh; all in all, I decided to use the conservative figure of 43TWh of energy (equivalent to 5 GW of power) coming from this study which takes into account the fact that many crypto are moving from Proof-of-Work consensus to Proof-of-Stake consensus, a far less energy-hungry protocol.
One last number we need to estimate is the number of transactions; for crypto, it stands at roughly 320k per day, while it is estimated to be about 65 million per day for the traditional banking system.
Based on these data, cryptocurrencies at the moment are between 79 and 280 times less energy efficient than the traditional banking system.
Therefore, I definiteky stand by my “shooting-from-the-hip” comment.