IEA (2019), Bitcoin energy use - mined the gap, IEA, Paris https://www.iea.org/commentaries/bitcoin-energy-use-mined-the-gap
Of all the potential implications of blockchain for the energy sector, the energy use of cryptocurrencies – and bitcoin in particular – has captured the most interest.
As the price of bitcoin skyrocketed in 2017, attention turned to the cryptocurrency’s energy and environmental footprint. High-profile news articles reported that electricity use of the bitcoin network had equalled that of medium-sized countries and was on track to consume as much electricity as the United States in 2019 and all of the world’s energy by 2020. A widely reported article in Nature Climate Change warned that Bitcoin emissions alone could push global warming above 2°C.
With bitcoin value tripling in recent months and Facebook announcing its new Libra coin, interest in the energy use of cryptocurrencies is again on the rise.
In this commentary, we explain why and how bitcoin uses energy; dig into published estimates of bitcoin energy use and provide our own analysis; and discuss how these trends might evolve in the coming years.
In order to understand why and how bitcoin uses energy, we first need to understand its underlying technology: blockchain. Blockchain offers a new way to conduct and record transactions, like sending money. In a traditional exchange, central authorities (e.g. banks) verify and log transactions. Blockchain removes the need for a central authority and ledger; instead, the ledger is held, shared, and validated across a distributed network of computers running a particular blockchain software.
The lack of a centralised, trusted authority means that blockchain needs a “consensus mechanism” to ensure trust across the network. In the case of bitcoin, consensus is achieved by a method called “Proof-of-Work” (PoW), where computers on the network – “miners” – compete with each other to solve a complex math puzzle. Each guess a miner makes at the solution is known as a “hash,” while the number of guesses taken by the miner each second is known as its “hashrate.” Once the puzzle is solved, the latest “block” of transactions is approved and added to the “chain” of transactions. The first miner to solve the puzzle is rewarded with new bitcoins and network transaction fees. The energy use of the bitcoin network is therefore both a security feature and a side effect of relying on the ever-increasing computing power of competing miners to validate transactions through PoW.
The energy use of the bitcoin network is a function of a few inter-related factors (some of which respond to the changing price of bitcoin):
- mining hardware specifications, notably power consumption and hashrate;
- network hashrate, the combined rate at which all miners on the network are simultaneously guessing solutions to the puzzle;
- “difficulty” of solving the puzzle, which is adjusted in response to the network hashrate to maintain the target block rate of one block every 10 minutes; and
- energy consumption by non-IT infrastructure, such as cooling and lighting.
The rising price of bitcoin, particularly as it rose to all-time highs in December 2017, drove huge increases in hashrate and difficulty, and the development and deployment of more powerful and energy efficient mining hardware.
The IT infrastructure for bitcoin and other cryptocurrencies has evolved rapidly over the past decade. In the early days of bitcoin (2009), hobbyists used standard central processing units (CPUs) to mine bitcoin. By October 2010, miners started to use more powerful graphics processing units (GPUs) as mining difficulty increased. By June 2011, miners – increasingly large and more industrial operations – used more powerful (but less energy-efficient) field-programmable gate array (FPGA) hardware, and a year later, moved to application-specific integrated circuits (ASICs).
ASICs are purpose-built chips, in this case, to mine bitcoin. The latest ASICs are both more powerful and more energy efficient – around 50 million times faster (H/s) and a million times more energy efficient (H/J) in mining bitcoin than the CPUs used in 2009.
Diverse methodologies, limited data availability, and highly variable conditions across the industry (e.g. mining hardware used; electricity costs; cooling needs) make estimating bitcoin energy use extremely challenging (Koomey, 2019). Therefore, all estimates must be interpreted with caution.
Recent published estimates of bitcoin’s electricity consumption are wide-ranging, on the order of 20‑80 TWh annually, or about 0.1-0.3% of global electricity use (Bendiksen & Gibbons, 2018; Bendiksen & Gibbons, 2019; Bendiksen, Gibbons & Lim, 2018; Bevand, 2018; BNEF, 2018; De Vries, 2018; Digiconomist, 2019; Krause & Tolaymat, 2018; Morgan Stanley, 2018; Rauchs et al., 2018; Stoll et al., 2019; Vranken, 2017).
These figures can appear large when compared to countries like Ireland (26 TWh) or emerging technologies like electric vehicles (58 TWh in 2018), but small when compared to other end-uses like cooling (2 020 TWh in 2016). Nonetheless, bitcoin mining is a highly mobile industry, allowing it to migrate quickly to areas with cheap electricity. Localised hotspots and electricity supply issues can emerge quickly, generating strong backlash from regulators and the public.
Bitcoin has also been compared on a per-transaction basis to VISA payments, the broader banking system, and gold mining. However, comparisons on a per-transaction basis are not meaningful in the context of PoW blockchains, particularly because the energy required for the networks to function is independent of the number of processed transactions. A recent peer-reviewed article compared the energy intensity of mining bitcoin (17 MJ/USD) to the mining of other metals like aluminium (122 MJ/USD) and gold (5 MJ/USD).
By far, the most frequently cited estimate in news media is the Bitcoin Energy Consumption Index (BECI), which uses a top-down approach that assumes miners spend (on average) 60% of their revenues on electricity at a rate of 0.05 USD/kWh. These key assumptions have been criticised to overestimate electricity consumption; indeed, BECI estimates represent the high range of published estimates to date.
Bendiksen, Gibbons (2018; 2019) & Lim (2018) also use a top-down approach, but undertake significant data collection efforts on existing mining hardware and mining locations to inform their assumptions and analysis. They also conduct sensitivity analyses around key uncertainties, including electricity costs and capital depreciation schedules. Under their central assumptions, they estimate that the bitcoin network consumes between 35 TWh (May 2018) and 41 TWh (November 2018; June 2019) per year.
Other researchers have calculated lower-bound estimates using a bottom-up approach (e.g. Deetman, 2016; Morgan Stanley, 2018; Valfells & Egilsson, 2016). This approach assumes that all miners are using the most efficient mining hardware to achieve the network’s hashrates (TH/s). The Bitmain Antminer S9 series (0.1 J/GH), used by two-thirds of miners worldwide, is typically used as a benchmark.
Using this approach, we can estimate that thebitcoin network (excluding cooling) consumed 31 TWh in 2018. Based on data collected from mining facilities in China, cooling and other ancillary demands accounts for 30% of electricity use overall, thereby adding another 42% to the lower-bound estimate. Therefore, we estimate that bitcoin mining consumed around 45 TWh in 2018, which aligns well with the latest peer-reviewed estimate of 45.8 TWh as of November 2018 (Stoll et al., 2019).
With the recent run up in price and hashrate, energy consumption is expected to be much higher in 2019. Through the first six months of 2019, bitcoin mining has already consumed an estimated 29 TWh.
While these early estimates provide a rough indication of bitcoin energy use today, it is clear that researchers need more data, in particular from mining facilities, to develop more rigorous methodologies and accurate estimates.
Headlines concerning the environmental impacts of bitcoin re-emerged last October, when a commentary article from Mora et al. in Nature Climate Change concluded that “…projected Bitcoin usage, should it follow the rate of adoption of other broadly adopted technologies, could alone produce enough CO2 emissions to push warming above 2°C within less than three decades”.
A closer look reveals serious issues in the study’s methodology and assumptions, notably around bitcoin adoption rates, the efficiency of mining hardware, and the assumed electricity mix (Masanet et al., 2019, Nature Climate Change, In Press). Crucially, the use of country average (and in some cases, world average) emissions factors inflates the GHG estimates, since bitcoin mines are typically concentrated in renewables-rich states and provinces.
Indeed, the selection of mining locations depend on a balance of several key factors, including access to low-cost electricity, fast internet connections, cool climates, and favourable regulatory environments. For these reasons, China, Iceland, Sweden, Norway, Georgia, the Pacific North West (Washington State, British Columbia, Oregon), Quebec, and upstate New York are key bitcoin mining centres.
Around 60% to 70% of bitcoin is currently mined in China, where more than two-thirds of electricity generation comes from coal. But bitcoin mining facilities are concentrated in remote areas of China with rich hydro or wind resources (cheap electricity), with about 80% of Chinese bitcoin mining occurring in hydro-rich Sichuan province. These mining facilities may be absorbing overcapacity in some of these regions, using renewable energy that would otherwise be unused, given difficulties in matching these rich wind and hydro resources with demand centres on the coast.
Electricity generation in other key bitcoin mining centres are also dominated by renewables, including Iceland (100%), Quebec (99.8%), British Columbia (98.4%), Norway (98%), and Georgia (81%). Globally, one analysis estimates that the bitcoin is powered by at least 74% renewable electricity as of June 2019. Another analysis of data from 93 mining facilities (representing 1.7 GW, or about a third of global mining capacity) estimates that 76% of the identified energy mix includes renewables.
Based on these analyses and data from IPO filings of hardware manufacturers and insights on mining facility operations and pool compositions, bitcoin mining is likely responsible for 10‑20 Mt CO2 per year, or 0.03-0.06% of global energy-related CO2 emissions.
Apocalyptic headlines that bitcoin would consume all of the world’s energy by 2020 echo back to warnings from the late 1990s about the internet and its growing appetite for energy, including one Forbes article in 1999 that predicted that “[…]half of the electric grid will be powering the digital-Internet economy within the next decade”.
Since then, researchers have collected real-world data and developed and refined methodologies to establish rigorous estimates of the energy use of data centres and the global ICT sector, including by the IEA. The dire predictions about the energy use of the internet failed to materialise despite exponential growth in internet services, largely because of rapid improvements in the energy efficiency of computing and data transmission networks.
The outlook for bitcoin energy use is highly uncertain, hinging on efficiency improvements in hardware, bitcoin price trends, and regulatory restrictions on bitcoin mining or use in key markets. Bitcoin prices in particular are extremely volatile: between December 2017 and 2018, its value fell by 80%, but has nearly tripled since.
It is important to recognise that bitcoin is just one cryptocurrency, which is one application of blockchain, which is itself one example of distributed ledger technology (DLT). Ethereum (ETH), the second largest cryptocurrency by market value, processes more than twice as many transactions as the bitcoin network while using only about one-third of the electricity consumed by bitcoin. ETH also operates on a Proof-of-Work (PoW) consensus mechanism, but its founder has announced plans to move to Proof-of-Stake (PoS) in an effort to reduce its energy intensity. PoS and Proof‑of‑Authority (PoA) could help reduce energy use while also addressing scalability and latency issues. Other DLTs like Tangle and Hashgraph similarly offer the promise of lower energy use, scalability, faster transactions, and no transaction fees compared to blockchain.
Over the coming years, other applications of blockchain – including those within the energy sector – are likely to garner more attention. As the scope and scale of blockchain applications increases, these trends combined are likely to materially reduce the future energy footprint of its technology.
Sensational predictions about bitcoin consuming the entire world’s electricity – and, by itself, leading our world to beyond 2°C – would appear just that…sensational. That said, this is a very dynamic area that certainly requires careful monitoring and rigorous analysis – particularly, a careful monitoring of local hotspots.
The energy use of bitcoin and blockchain is just part of the blockchain and energy story. In our next commentary, we’ll look at how blockchain is already impacting the energy sector, dive into some of the most promising applications, and explore the technological, regulatory or market design challenges that await.
Bitcoin energy use - mined the gap
George Kamiya, Energy Policy Analyst Commentary — 05 July 2019