Ethereum’s switch to proof of stake sharply reduced the electricity used to secure the network, but a percentage alone does not explain what is being measured. A 2026 study from the Cambridge Centre for Alternative Finance offers a bottom-up estimate based on hardware, hosting and electricity-grid data. Its findings are useful precisely because they separate physical nodes from validators and electricity demand from carbon emissions.
The headline estimate
Cambridge estimates Ethereum’s annual electricity consumption at about 7.87 gigawatt-hours, equivalent to continuous demand of roughly 0.90 megawatts. That is more than 99.9% below the study’s final pre-Merge baseline of about 2.4 gigawatts.
The comparison reflects a change in consensus design. Under proof of work, miners competed with energy-intensive computing equipment. Under proof of stake, validators commit capital and run conventional servers or home hardware. The estimate does not mean Ethereum uses no electricity; it means the network no longer ties security to large-scale computational work.
How Cambridge built the number
The researchers modelled approximately 8,522 physical nodes, which should not be confused with the much larger validator count. One machine can operate multiple validators, so treating every validator as a separate computer would overstate hardware demand.
The study used wall-plug measurements from 20 client combinations and scaled them against observed hosting patterns. It arrived at a network-weighted average of roughly 105 watts per node. Residential setups had a median draw near 18 watts, while workstation-class enterprise deployments were estimated around 152 watts. These assumptions make the estimate auditable, but they also mean future changes in client software, hardware and node geography can move the result.
Location affects emissions
Electricity use and emissions are related, not interchangeable. Cambridge estimates annual emissions of about 2.37 kilotonnes of carbon-dioxide equivalent. Its model attributes 56.4% of the network’s electricity to sustainable sources: 39.4% renewables and 17% nuclear power. Natural gas was the largest single source at 27.7%.
Geography therefore matters. The United States, Germany, Finland and France hosted about 62% of the full nodes in the study. Roughly 64% of nodes were in cloud or enterprise data centres, and Hetzner, Amazon Web Services and OVH together hosted around 40% of the estimated node count. Moving the same machines between grids could change emissions without changing protocol rules or total electricity demand.
What the report does not establish
A smaller energy footprint does not settle questions about decentralization, client diversity, uptime, fees or validator concentration. Nor does it prove that institutions will adopt Ethereum. Those are separate technical, market and governance questions.
The figures also should not be transferred to proof-of-work networks. Bitcoin and Ethereum use different security designs, so Ethereum’s post-Merge result is not evidence that another network could make the same transition without changing its own consensus model.
How to read future updates
When a new estimate appears, check its date, node sample, power measurements, hosting assumptions and grid mix before comparing it with this baseline. Ask whether the number describes instantaneous power, annual electricity consumption or greenhouse-gas emissions. Keeping those units separate turns a dramatic percentage into a measurement that can be tested over time.
Adapted from Cambridge Report Confirms Ethereum’s Energy Use Fell to 7.87 GWh After The Merge.