How we estimate bitcoin's electricity use and emissions: our method, inputs and sources

Our estimate beside the others

Our electricity figure sits between the two. The gap to Cambridge is mostly the machines: we assume the whole network runs on the machines in our list, while a real fleet includes older, less efficient ones we can't see. The gap to Digiconomist is mostly the method: it works back from what miners earn and assumes a larger share goes on power. On emissions we use the world's average grid, a cleaner mix than Digiconomist assumes (558 g per kWh) and a dirtier one than Cambridge's survey found (288 g per kWh), so our emissions land between theirs as well.

The method in five steps

1. Hashrate. The network was running at about 1,048 exahashes a second (EH/s) on 2026-10-30, from mempool.space. Block subsidy is 3.125 BTC, and fees averaged 0.0198 BTC a block over the last year (1,042 BTC over 52,558 blocks).
2. What a terahash earns. (3.125 + 0.0198) BTC × 144 blocks × $77,472 ÷ (1,048 EH/s × 1,000,000 TH per EH) = $0.0335 per terahash per day. The bitcoin price is a public quote read from mempool.space prices API (read 31 October 2026).
3. Which machines are worth running. A machine using J joules per terahash burns J × 24 ÷ 1,000 kWh per terahash per day. It is worth running while that costs less than it earns, so the most joules per terahash it can use is 0.0335 ÷ (power price × 24 ÷ 1,000). At $0.03 a kWh that is46.5 J/TH; at $0.05, 27.9 J/TH; at $0.07, 19.9 J/TH.
4. Three bands. Low: every hash done on the most efficient machine in our list (11.0 J/TH). Central: the average of the machines first shipped in 2024 or later that are worth running at $0.05 (15.4 J/TH). High: the average of every machine worth running at $0.03, which includes the older ones (18.4 J/TH). Each is multiplied by 1.1 for cooling, power supply and site overheads, our assumption.
5. Electricity and emissions. Terawatt hours a year = 1,048 EH/s × 1,000,000 × J/TH × 1.1 × 8,760 hours ÷ 10¹². Emissions = that × a grid intensity.

Emissions on four grid scenarios

We don't know where the machines are, so we show the central electricity figure (155 TWh) on four grids rather than picking one. Intensities are Ember's lifecycle figures for electricity generated in each place.

The central figure on this page uses the world average. The hydro and coal rows are bounds, not estimates of where bitcoin is mined: no real fleet runs entirely on either.

The machines

20 SHA-256 mining machines, hashrate and joules per terahash as published by the seller or tester named, read on 31 October 2026. Power is hashrate × joules per terahash. The year is approximate and only weights the central band. A tick means the machine is worth running at $0.05 a kWh.

Specifications for the same model differ between sellers, and some pages here are resellers because the manufacturer's was not reachable. Where they disagreed we used the figure the seller or tester lists for that exact variant. We copy facts only, never a manufacturer's text or images.

What this estimate can't do

• No miner survey. Cambridge's 2025 figures come from firms covering 48% of the network. We publish scenarios instead.
• No location data. We can't say how much mining is in any country, which is why the grid is a scenario.
• The machine list is a sample. It leaves out many older and less common models, and a real fleet is more mixed than any average of 20 models.
• Not peer reviewed. We publish the method so others can check and challenge it.
• Bitcoin only. Other coins are not covered yet.

Who owns the tonnes

A network estimate becomes a company's figure through the hybrid method from South Pole and the Crypto Carbon Ratings Institute: the network's emissions are split between holders and transactions by how miners are paid (new coins against fees), and a holder takes its share of the coins. Six rules keep that honest:

1. Every kWh is counted once, for the year it was used, then split. It is never copied forward or counted again.
2. Minting, security and transactions are one cost; new coins and fees are how miners are paid for it.
3. The footprint stays with the chain that used the energy.
4. It is a yearly flow, not a stock: emissions do not travel with a coin when it is sold.
5. Offsets are reported separately and never deducted from the gross figure.
6. A miner's own reported figure beats the network average for its share of the hashrate. Historical figures are research, not a reportable number today.

On our central estimate, one bitcoin held for a year carries about 3.53 tonnes CO2e. The crypto carbon calculator works out any amount, and the holdings guide explains whether you have to report it.

Licences

• Cambridge and Digiconomist: only their published headline totals are quoted, credited and dated. None of their datasets or charts is used.
• Ember Yearly Electricity Data: CC BY 4.0, contains data from Ember (ember-energy.org).
• Blockchain figures and the bitcoin price: public facts read through mempool.space.
• Machine specifications: facts only, with the page each was read from.

Estimates for reporting and research, not financial or tax advice. No agency endorses them.

Changelog

• 31 October 2026: first version. 20 machines, bitcoin at $77,472, hashrate 1,048 EH/s, Ember grid data retrieved 20 October 2026.

How Carbon Recycling helps

Carbon Recycling works out the rest of your footprint, Scope 1, 2 and 3, from your bills, mileage, travel and spending with the UK government factors, with the method behind every line.