Crypto Energy Use: Proof of Work vs Proof of Stake

Why Some Networks Use So Much Electricity

Cryptocurrencies need an agreed record of who owns what, and networks reach that agreement in different ways. In proof of work, used by Bitcoin, specialised computers race to find a valid answer to a puzzle, and the winner adds the next block and collects a reward. Racing costs electricity, and Bitcoin's 2008 design paper said so directly, comparing miners to gold miners spending resources: 'it is CPU time and electricity that is expended'. In proof of stake, used by Ethereum since 2022, participants lock up coins as a deposit instead. The ethereum.org energy page puts the difference as using ETH instead of energy to secure the network.

How Big Is the Bitcoin Figure?

Nobody can meter a decentralised network, so every figure is an estimate. The Cambridge Bitcoin Electricity Consumption Index, run by the Cambridge Centre for Alternative Finance, publishes a best guess between a floor and a ceiling. Read in October 2026, at an assumed electricity price of 5 US cents per kilowatt-hour, its best guess was about 176 terawatt-hours a year, a draw of roughly 20 gigawatts, within a range of about 90 to 338 terawatt-hours. A terawatt-hour is one billion kilowatt-hours, and the index notes that its annual figure assumes constant power use over a year.

What Changed for Ethereum

Ethereum switched from proof of work to proof of stake in the 2022 Merge. The Crypto Carbon Ratings Institute (CCRI) estimated in September 2022 that the old network used about 22.9 terawatt-hours a year as of August 2022 and the new one about 2,601 megawatt-hours, a cut of more than 99.988 percent. As of July 2026, ethereum.org still publishes the figure of roughly 0.0026 terawatt-hours a year. Energy is not the same as emissions, though: the page itself says consumption does not map precisely to environmental footprint, because projects use different energy sources.

1. Which date does the figure describe? Mining estimates move with hash rate, hardware and prices.
2. Which method produced it: a top-down estimate from miner revenue, or a bottom-up one from hardware and hash rate?
3. Is it electricity, which is energy, or emissions, which depend on the power mix behind the electricity?

How the Estimates Are Built

Cambridge describes a hybrid top-down approach: it builds a basket of real-world mining hardware and assumes miners are rational, so they only run machines that are profitable at the assumed electricity price. Three scenarios follow. The floor assumes every miner always uses the most efficient hardware in a data centre with almost no overhead, a power usage effectiveness (PUE) of 1.01. The ceiling assumes the least efficient profitable hardware at a PUE of 1.20. The best guess, in version 1.4.0 of the model read in October 2026, weights hardware by release date, adds a two-month deployment lag and assumes a PUE of 1.10. Efficiency is measured in joules per terahash; in the worked example on the methodology page read in October 2026, a device drawing 1,000 watts and producing 10 terahashes a second scores 100.

Why Sources Disagree

Two method families give different answers. A top-down estimate starts from miner revenue, assumes a share is spent on electricity at an assumed price, and divides. A bottom-up estimate starts from hash rate, works out which hardware could profitably produce it, and sums the power. Labels vary: Cambridge calls its hardware-based method hybrid top-down, whereas CCRI would class that kind of method as bottom-up. CCRI's September 2022 report put pre-Merge Ethereum at about 22.9 terawatt-hours a year as of August 2022, bottom-up, and cited a top-down estimate of about 78 from Alex de Vries as of early September 2022, which assumed 0.1 US dollars per kilowatt-hour and a revenue share close to 100 percent. That is a gap of more than three times for one network in one month.

Bitcoin ranges also move with date, model version and price assumption. The US Energy Information Administration (EIA), using Cambridge data, gave 67 to 240 terawatt-hours for 2023 with a point estimate of 120, and cited an end-of-January 2024 reading of 80, 170 and 390. Read in October 2026, Cambridge gave about 90, 176 and 338 at 5 US cents per kilowatt-hour; at 3 cents the ceiling was about 424, and at 10 cents the best guess was about 137.

The Grid and the United States

EIA's February 2024 article put US Bitcoin mining at 0.6 to 2.3 percent of 2023 electricity use, or 25 to 91 terawatt-hours, assuming about 38 percent of global mining was American, equal to the annual demand of more than three million to more than six million homes. The page carries a note, read in October 2026, that EIA discontinued its emergency data collection. It also describes miners taking part in demand-response programmes: in Texas, as of February 2024, the Large Flexible Load programme had enlisted up to 1,530 megawatts of large users to curtail at peak times, with miners among the major participants. Some sites sit beside underused plants, hydro dams or wells that would otherwise flare waste methane.

Emissions Accounting

Emissions add a layer: where the electricity is generated. Cambridge's emissions methodology multiplies each country's emission intensity by that country's share of Bitcoin mining electricity, using a mining map for the period from 1 September 2019, and assumes mining followed each country's share of world electricity production before then. Its floor and ceiling scenarios assume all-hydropower and all-coal supply respectively. The September 2022 White House OSTP report estimated 140 plus or minus 30 million tonnes of carbon dioxide a year for the largest crypto-assets, about 0.3 percent of global emissions, and listed noise, water use and electronic waste among local impacts.

Policy Responses

EIA's February 2024 article says mining relocated to the United States after China cracked down in 2021, and that members of Congress wrote in November 2022 and February 2023 asking for a 'mandatory disclosure regime' on miners' energy and emissions. In the European Union, Regulation 2023/1114 (MiCA) requires a crypto-asset white paper to include information on the principal adverse climate impacts of the consensus mechanism (Article 6(1), point (j)), requires service providers to publish the same on their websites (Article 66(5)), and tells ESMA to develop standards on sustainability indicators that weigh different consensus mechanisms and their energy use (Article 66(6)).

The Security Argument, and Proof of Stake Measured

The 2008 design paper ties energy to security: the majority decision is the chain with the greatest proof-of-work effort invested in it, and to modify a past block an attacker would have to redo the work of that block and all later ones. The expended resource is what an attacker would have to match; whether a given level of electricity use is a fair price for that is a value judgement these sources do not settle. Proof of stake is not zero-energy: CCRI's September 2022 estimate applied a modelled average of about 547 kilowatt-hours per node per year (bounds about 175 and 1,315) to 4,755 beacon nodes counted on a public dashboard. CCRI adds that many nodes are very likely cloud-hosted, so its self-hosted assumption may overestimate. This page does not quantify hardware manufacture or cloud concentration.

Numbers Age

A July 2023 comparison table on ethereum.org listed Bitcoin at 149 terawatt-hours a year and proof-of-work Ethereum at 21. The OSTP report put the whole crypto sector at 120 to 240 terawatt-hours a year as of August 2022, with proof-of-work Ethereum at 20 to 39 percent of it. Cambridge's October 2026 reading for Bitcoin alone was about 176 terawatt-hours a year. Dates, models and scopes all differ, so these figures cannot be subtracted from one another to show a trend.