After three years, the University of Cambridge has implemented a major update to its Bitcoin Electricity Consumption Index (CBECI) to more accurately assess the global energy footprint of Bitcoin miners.
The conclusion?
The previous power estimates were greatly overestimated.
“The first and most noticeable discrepancy appears in 2021, where our previous CBECI model estimated an electricity consumption of 104 terawatt-hours (TWh), 15.0 TWh higher than the revised model estimate (89.0 TWh),” the report read.
A terawatt-hour (TWh) is a unit of energy equivalent to outputting one trillion watts for one hour. For context, if used for one hour per day, an average incandescent lightbulb consumes 21,900 watthours in a year.
The university’s 2022 power estimate was also adjusted down by 9.8 TWh, from 105.3 TWh to 95.5 TWh, putting Bitcoin’s electricity consumption that year in roughly the same league as U.S. tumble dryers (108TWh).
The need for revision arose from the university’s former methodology that every “profitable” hardware model released within the past five years “equally fueled the network hashrate.”
Though effective for most of Bitcoin’s lifespan, the methodology started exhibiting shortcomings in 2021 after China’s mining ban.
“This led to overestimating the number of older hardware and underestimating the proportion of newer hardware,” the report’s author Alexander Neumueller told Decrypt.
ASIC hardware devices have become “considerably more efficient” and powerful over time. ASICs are “Application Specific Integrated Circuits," machines specially designed to mine Bitcoin as efficiently as possible.
Given that the ban created a shortage in data center capacity, Cambridge said it is “reasonable to infer that mining operators would have already replaced all old machines with newer models.”
Cambridge model good, but not perfect
Shortcomings also emerged during “exceptionally profitable mining periods,” when the old hardware distribution estimates presented a “disproportionally large number of older devices.”
With its new methodology, Cambridge incorporated recent hardware mining deliveries, though many “assumptions and simplifications” still applied.
Nevertheless, the report’s findings reinforce those from a Coin Metrics study in June, which used the blockchain-based fingerprints left behind by mining machines to determine which hardware was dominating the network.
Though the report’s findings were not incorporated into Cambridge’s new methodology, Neumueller said that he holds the Coin Metrics authors’ work in “high regard.”
Karim Helmy—lead author of the Coin Metrics report—told Decrypt that he was glad to see Cambridge’s updated figures taking public company hardware data into account.
However, he believes its figures are still inaccurate.
“The new methodology continues to overestimate energy consumption in bull markets,” he said.
For example, Cambridge estimated a “considerable increase in energy consumption per terahash” between 2020 and 2021 which “is unlikely to have actually occurred in practice.”
Cambridge’s findings were also supported by CH4 Capital founder Daniel Batten, whose fund invests in companies using Bitcoin mining to clean up the environment.
His model estimates Bitcoin’s current power demand to be 13.095 GW, versus Cambridge’s 12.89 GW estimate. This figure presents the power it takes for all active mining devices to function.
In 2023, Cambridge estimates that the Bitcoin network has consumed 70.4 TWh of energy–so far.