Why the network uses energy
Miners repeatedly hash block headers in search of a result below the difficulty target. This competition makes producing history costly while full nodes can verify the result cheaply. Electricity is therefore not a fee for moving one particular bitcoin; it is an input to the mechanism that orders blocks and makes rewriting recent history expensive.
Consumption is not a fixed command in the protocol. Miners compare expected rewards with electricity, hardware and operating costs; higher revenue can attract more computing power, after which difficulty adjusts. Adding one transaction to a block does not by itself switch on a proportional number of new machines.
What is measured and what is estimated
Hashrate can be inferred from block arrival and difficulty, but the network does not record every ASIC model, its power draw or its electricity supplier. Cambridge therefore uses modelled ranges and surveys. Its 2025 report estimated annual consumption at about 138 TWh, roughly 0.54% of global electricity use; the sample represented about 48% of hashrate and the authors themselves warn of geographic bias.
That number is a snapshot, not a permanent property of Bitcoin. Hashrate, machine efficiency, BTC price, fees and mining locations change. A responsible comparison states the period, source, central estimate and uncertainty; a number without a date soon becomes stale.
- electricity use is modelled from indirect observations
- a survey may not represent every miner
- more efficient machines do not guarantee lower total consumption
Electricity, emissions and local harm differ
The same megawatt-hour can have a very different carbon footprint depending on whether it came from coal, gas, nuclear, hydro, wind or a mixture. For the same period, Cambridge reported about 39.8 MtCO₂e under its survey-based model and 69.6 MtCO₂e under an IP-location model. The gap shows how strongly the answer depends on location and energy-mix assumptions.
Carbon is not the only impact. Noise, local air pollution, cooling water, electronic waste and grid congestion may matter. Conversely, interruptible mining can sometimes use surplus supply or turn off during peaks. That is a property of a particular project, market and contract—not proof that every mining operation automatically helps its grid.
The fair question is value relative to cost
Saying that Bitcoin uses less electricity than an entire other industry does not settle whether its use is justified. The service, alternatives, marginal emissions and local effects all matter. Critics may reasonably value the service below its environmental cost; users may value global settlement and verification without a central operator.
Proof of stake demonstrates that digital consensus can be designed with far less direct energy consumption, but it uses a different security model. Bitcoin could not adopt it without a fundamental rule change accepted by users. The dispute is therefore not whether consumption exists, but whether this security model justifies its costs and externalities.