A Cambridge Centre for Alternative Finance report estimates that Ethereum’s post-Merge network consumes about 7.87 gigawatt-hours of electricity per year. The figure is an annualized model, not a reading from one global meter. Its value comes from the way the researchers connect measured hardware demand to an estimate of the network’s physical footprint.
From machines to a network estimate
The report starts with approximately 8,522 discoverable full nodes. It separates the hardware into two broad operating profiles: residential systems and workstation-class or hosted systems. Researchers measured power at the wall and tested 20 combinations of consensus and execution clients. The reported median demand was about 18 watts for a residential configuration and about 153 watts for a workstation-class configuration.
After weighting those profiles by the observed hosting distribution, the report arrives at an average near 105 watts per node. Scaling that figure across the mapped node population produces a continuous network demand of roughly 0.90 megawatts, equivalent to about 7.87 GWh over a year.
This method also defines the estimate’s limits. Discoverable nodes are not a perfect census, hardware configurations vary, and node location cannot be known with certainty. Validators and full nodes are different units: one node can hold keys for many validators. Counting validator credentials as if each represented a separate machine would overstate the physical equipment drawing power.
Why the Merge changed the calculation
Before the Merge, proof-of-work miners competed through computation, making electricity consumption part of block production. Proof of stake replaced that contest with validators committing capital. Ethereum still depends on servers, networking equipment, storage, and client software, but its consensus process no longer requires a mining fleet to perform repeated hashing.
The Cambridge report describes the change as a drop of roughly 3.5 orders of magnitude from a pre-Merge demand near 2.4 gigawatts to the modeled post-Merge baseline near 0.90 megawatts. That comparison reflects a structural protocol change rather than gradual gains in chip efficiency.
Electricity use and emissions are separate
Electricity consumption does not by itself determine greenhouse-gas emissions. The report maps node locations to the carbon intensity of their host jurisdictions. It estimates a node-weighted electricity mix of about 56.4% sustainable sources, defined as renewables and nuclear, and 43.6% fossil fuels.
Applying that geographic mix to the power estimate yields an annual footprint of approximately 2.37 kilotonnes of carbon-dioxide equivalent. Cambridge calculates that as roughly 99.98% below Ethereum’s final pre-Merge annualized emissions. The remaining result depends more on the grids serving nodes than on the consensus mechanism.
What the number can and cannot establish
The 7.87 GWh estimate quantifies the power demand of Ethereum’s mapped full-node infrastructure under the report’s assumptions. It does not measure transaction efficiency, application demand, decentralization, or financial adoption. Nor does a lower energy estimate resolve questions about client diversity, hosting concentration, validator structure, fees, or network capacity.
Cambridge provides enough detail to treat the number as a modeled baseline with stated inputs rather than a standalone headline. Future estimates may move as node counts, hardware, hosting patterns, and electricity grids change.
Source: BTCUSA.