Controversy over a new academic paper spotlights how difficult it is to measure bitcoin's carbon footprint

Quick Take

  • A recent academic journal article concluded that carbon dioxide emissions from bitcoin mining could seriously undermine China’s climate change targets, sparking criticism and questions about the paper’s sources.
  • The report is the latest reminder that accurately measuring bitcoin’s carbon footprint is extremely challenging if not impossible.

Earlier this month, a peer-reviewed academic journal article drew the conclusion that without a policy intervention, carbon dioxide emissions from bitcoin mining could seriously undermine China’s climate change targets.

The report, co-authored by researchers from various Chinese universities and published in Nature Communications, instantly inspired similar-sounding headlines on a wide range of mainstream media including the BBC, CNN, The Guardian, Economist, CNBC and others.

The authors wrote that the annual energy consumption of bitcoin mining in China is expected to peak in 2024 at 296.59 terawatt-hours (tWh) and generate 130.50 million metric tons of carbon dioxide emissions. That would exceed the total annual greenhouse gas emission output of the Czech Republic and Qatar, and would pose a threat to China’s long-term ambition to become carbon neutral.

Crucially, however, the researchers left two fundamental questions unanswered, according to the paper's critics: where exactly are the mining machines located, and what is the energy mix there?

“I expected most of the paper to be about province-level data covering the energy mix of Chinese miners,” Nic Carter, partner at Castle Island Ventures and co-founder of Coin Metrics, wrote on Twitter. “But that's missing. Instead, they claim to have taken this into account... but don't show their work (!) They just assert they've quantified this.”

In the article, the researchers said they accounted for the carbon dioxide emissions of both hydropower and fossil fuel-based bitcoin mining in China. “As suggested by the actual regional statistics of Bitcoin miners, we assume 40% of miners are located in the coal-based area,” they wrote. But they didn’t elaborate on the origin of those "regional statistics."

After The Block reached out to the co-authors for clarification on this question, Shouyang Wang, the Chair Professor from the Academy of Mathematics and Systems Science of the Chinese Academy of Sciences, responded:

“We obtained the statistics on the broadcast location of each mining pool from BTC.com. Based on the location of each mining pool and the associated region, we are able to make the assumption that approximately 40% of the miners are located in the coal-energy area.”

In a follow-up message, Wang elaborated on what the authors meant by “broadcast location of each mining pool”:

“The pool regional statistics of BTC.com suggests an approximately 60% to 40% split between hydro-rich and coal-heavy regions in China. The ratio represents the computing power reported from Shenzhen (server location closer to hydro-rich regions) versus Beijing (server location closer to coal-heavy regions).”

From Wang’s statement, it appears that the researchers based this assumption on another: that the location of mining pools corresponds directly to the location of the individual miners. But that would be a misunderstanding of how bitcoin miners and pools function work in practice.

Bitcoin mining pools aggregate hashing power from any individual miners who want to connect to their service to mine blocks collectively. Although F2Pool is based in Beijing, that doesn’t mean at all the miners connected to the pool are located there, too. In fact, they can come from anywhere in China or anywhere in the world.

Wang also confirmed that they did not obtain internal statistics from any major mining pool about the exact geolocation of each of their miner customers. Instead, they used the data from Cambridge University, “which showed that 40% of the hash rate are contributed by coal-heavy regions such as Xinjiang and Inner Mongolia as of April 2020,” according to Wang.

But Cambridge University’s data is outdated by a year, and the makeup of the mining network is constantly changing. For instance, China’s share of the Bitcoin network's total mining capacity has declined significantly over the past 12 months.

Impossible to measure?

THE SCOOP

Keep up with the latest news, trends, charts and views on crypto and DeFi with a new biweekly newsletter from The Block's Frank Chaparro

By signing-up you agree to our Terms of Service and Privacy Policy
By signing-up you agree to our Terms of Service and Privacy Policy

Assumptions aside, the debate over the report's methodology does raise a question that is getting difficult to ignore as bitcoin and other cryptocurrencies make their way into the mainstream: how can the energy mix of the bitcoin network be accurately measured?

The reality is that the decentralized nature of bitcoin and bitcoin mining makes it incredibly difficult to quantify how much of the energy the network uses is sourced from renewables or not — let alone estimate the associated carbon dioxide emissions. 

The Cambridge Center for Alternative Finance made one of the most recent attempts to answer the question with its Cambridge Bitcoin Electricity Consumption Index (CBECI).

The CBECI estimates bitcoin’s total energy consumption with a theoretical lower and upper bound between 35 and 391 terawatt-hours annually. The lower bound assumes that all miners are using the most energy-efficient hardware available all the time, and the upper bound assumes that all miners are using the least efficient hardware all the time. 

The researchers behind the CBECI then give a “best guess estimate” — based on the assumption that miners use “a basket of profitable hardware rather than a single model” — that the network consumes 113.88 terawatt-hours annually. That’s around the same amount that The Netherlands consumes

The CBECI also provides a geographic breakdown of bitcoin’s hash rate based on data supplied by three bitcoin mining pools: BTC.com, Viabtc and Poolin. The report’s mining map as of April 2020 showed that miners in China’s Xinjiang and Inner Mongolia — two regions that are fossil fuel-based — accounted for roughly 40% of the global hash rate. Meanwhile, hydropower provinces like Sichuan and Yunnan accounted for roughly 25% as of a year ago, according to CBECI. 

But these numbers haven't been updated since April 2020, and the data came from contributed by only three mining pools that combined to account for only 35% of bitcoin’s total hash rate. Further, the CBECI’s mining map only captured the hash rate’s geographic breakdown from September 2019 to April 2020, a period that’s known as the dry season in China. 

As the CBECI notes in a disclaimer: “In some countries, and China in particular, mining operations tend to move between locations according to seasonal variance in renewables production. These migration patterns can only be observed when selecting a longer timeframe for the analysis.”

During the dry season, a significant number of miners in China’s southwestern provinces of Sichuan and Yunnan migrate north to Xinjiang or Inner Mongolia, where the energy mix tends to be coal-based. Upon the return of the rainy season, which spans from May to September, some may return to the south, where there is a lot more hydropower in the mix.

More accurate accounting would require year-long cooperation from enough major bitcoin mining pools to comprise a large majority of the hash rate. And it would need not only the IP geolocation of each mining machine — whether it’s in Sichuan, Yunnan, Xinjiang or Inner Mongolia — but the exact model of each mining machine. 

That’s because bitcoin mining machines have evolved significantly over the past years, and several old models are still in use. For instance, to compute the same amount of hash rate, Bitmain’s five-year-old AntMiner S9 consumes four times the amount of electricity used by the most efficient model on the market today, the AntMiner S19 Pro.

It’s next to impossible to know the exact composition of the machines in use, particularly during a bull market, when a large number of old models can still be used to turn a profit.

Another theoretical approach — albeit an even longer shot — would be to convince energy authorities in different countries to identify how many bitcoin mining farms there are locally and how much energy they consume.


© 2023 The Block. All Rights Reserved. This article is provided for informational purposes only. It is not offered or intended to be used as legal, tax, investment, financial, or other advice.

About Author

Wolfie joined The Block’s news team in 2020 and switched to the research side in 2021 to focus on crypto mining analysis. Prior to The Block, he had been a journalist at CoinDesk for three years. Wolfie has a background in financial journalism.