The University of Cambridge’s Judge Business School just released what might be the most statistically sound and feature-rich model on Bitcoin’s power consumption to date.
The Cambridge Bitcoin Electricity Consumption Index (CBECI) provides estimates for Bitcoin’s real-time and annual electrical appetite with a live data feed that updates every 30 seconds. These data points are divided into three categories: upper bound, estimated and lower bound consumption. Together, they give a liberal, average and conservative spread for Bitcoin’s actual power use.
CBECI’s team provide all three figures in order to weigh all possibilities, crunching a hodgepodge of various network and mining data. The estimated figure — currently at 7.5 GW for real-time and 53 TWh for annualized consumption — is the tool’s best guess for an accurate appraisal of Bitcoin mining’s electrical cost.
Cambridge partly structured its model after a comprehensive research study by Marc Bevand on the energy efficiency and market presence of ASICs.
Bevand pooled data from popular mining hardware manufacturers — some, like Genesis, were more generous with their products’ information — and, using the network’s hashrate to gauge how many miners could be running, used this ASIC data to derive an estimate of the blockchain’s total electricity consumption.
Bevand breaks his numbers down into an upper-bound category that assumes miners are using the least efficient hardware available, a lower-bound category which figures they use one of the top-three most efficient ASICs, and an estimated category that meets in a more realistic middle.
The Cambridge model also follows a similar logic in comparing hashrate, hardware efficiency and profitability, but it then factors in the efficiency of the data centers that house mining farms and the average cost of their electricity. Its lower-bound model, for instance, assumes that miners are utilizing the most efficient hardware possible and that their facilities operate with a 1.01 power usage efficiency (PUE). The upper bound assumes the opposite and a 1.2 PUE. For its best-guess estimate, Cambridge takes an average of the hardware efficiency of the other two models and applies a 1.1 PUE. Each model then assumes the global average price for electricity is $0.05 kWh (a value derived from “in-depth conversations with miners worldwide,” the post reads).
Cambridge concludes its report with the model’s limitations: Assuming an average global electricity cost doesn’t account for dynamic factors like region and seasonal circumstances, and the mining specifications manufacturers provide might not be wholly accurate (Cambridge may not have been privy to data from the most efficient hardware, either).
Visually, Bitcoin’s annualized electrical appetite looks like this:
On the tool’s website, you can adjust the average electricity costs to play with the model — if you raise it to the max at $0.20, for instance, the estimated power consumption drops to 32 TWh, while dropping it to $0.01 raises it to 62 TWh.
If you were worried about using 53 TWh per year to secure the internet’s endogenous monetary system, Cambridge also offers a comparison section to see how Bitcoin stacks up to some of the world’s other electrical needs.
On average, 25,082 TWh of energy are produced every year and 20,863 TWh of this is consumed. Bitcoin accounts for 0.21 percent and 0.24 percent of this total energy, respectively. The study also makes the pointed observation that idle devices left on in U.S. households each year could power the entire Bitcoin network four times over.
Oh, and Bitcoin could power all tea kettles in the U.K. for 11 years (or 1.5 in the EU and U.K.).
It also compares Bitcoin’s use alongside other countries’, and yes, Bitcoin uses about as much power as a small country (Switzerland or Nigeria, for instance).
But let’s put this into perspective with other industries: Gold mining, according to figures cited in a 2014 CoinDesk article, consumes 131 TWh each year, and this doesn’t include recycling and refining processes for jewelry. Between ATMs, branches, transportation and server uptime, banks and credit card companies burn roughly 100 TWh annually. We haven’t factored in the internet yet, either, which Standford Research Fellow Jon Koomey estimates might account for 10 percent of the world’s total electricity consumption.
That’s 50 times what Bitcoin puts out, and yet no one questions whether that’s worth it.