Cambridge Maps Where Ethereum’s Power Really Lives After the Merge
• July 11, 2026 9:38 am • CommentsEthereum’s move away from mining slashed its electricity use. A new Cambridge study shows where the network’s physical and economic power now sits.
The Cambridge Centre for Alternative Finance estimates that 64% of Ethereum’s discoverable full nodes run in cloud or enterprise hosting. The United States alone accounts for 31% of node activity.
Those figures matter because the machines running Ethereum still live in real data centers, homes, countries, and legal jurisdictions, even when the network itself has no headquarters.
The new Cambridge Centre for Alternative Finance report estimates roughly 8,522 discoverable full nodes. The researchers describe that number as a floor because private deployments, restrictive firewalls, disabled discovery, and machines that reject crawler connections can remain invisible.
Cambridge divided the measured population into two broad hardware profiles. Residential configurations drew a median of about 18 watts, while workstation-class setups drew about 153 watts across 20 tested combinations of Ethereum execution and consensus clients.
After weighting those profiles by the network’s hosting mix, the report arrived at a reference draw of approximately 105 watts per node. Cloud and enterprise facilities also carry cooling and networking overhead that home operators do not.
The report’s lead researcher introduced the work as a return to physical measurement:
1/
How much electricity does Ethereum use today? We went back to the wall plug.
'Ethereum after the Merge – A Change in Power' – our updated assessment of the network's energy and climate footprint – is out today from the Cambridge Centre for Alternative Finance. 🧵— Alexander Neumüller (@alexneumueller) July 10, 2026
The hosting map is concentrated. Hetzner accounts for about 15% of discoverable nodes, Amazon Web Services 13%, and OVH 12%, putting roughly 40% of the measured population with three providers.
Another 36% runs through residential connections. That home-operator share keeps the physical layer from collapsing into a pure cloud service, while the provider totals still create obvious outage and policy pressure points.
The Cambridge Blockchain Network Sustainability Index methodology explains why the estimate starts with nodes. Each full node pairs execution software, which handles transactions and smart contracts, with consensus software, which follows proof-of-stake agreement and routes network data.
Validator keys represent economic participation. A single physical node can operate hundreds of validators, so Cambridge counted roughly 8,522 discoverable nodes beside approximately 894,000 validators.
The methodology is designed for continuing revision as Ethereum’s software, hardware requirements, node population, and electricity grids change. Cambridge says its index updates power-related data daily and treats the network count as an estimate shaped by what its discovery tools can actually see.
The geography adds a second layer of concentration. Cambridge found 31% of discoverable nodes in the United States, 16% in Germany, 8% in Finland, and 6% in France, with the underlying measurements putting the four-country total at roughly 62%.
The U.S. share has already appeared in a legal fight. The Cambridge report notes that the Securities and Exchange Commission previously cited the density of U.S.-based Ethereum nodes while arguing for jurisdiction in a token-enforcement case.
The Block summarized the headline numbers after speaking with the research team:
THE BLOCK: Cambridge's Alexander Neumüller finds Ethereum nodes clustered on Hetzner, AWS and OVH, with 31% of node activity in the U.S.
More than one-third of validators going offline simultaneously could stop checkpoints from finalizing, he says. The SEC argued for U.S.… pic.twitter.com/MYqKL3NPwZ
— The Block (@TheBlockCo) July 10, 2026
The Block focused on the security consequences of that concentration. A regional cloud outage, a provider policy change, or a common software failure can affect many validators at once when operators depend on the same hosting and client stack.
The finality threshold is measured by staked ETH. Cambridge estimated about 39 million ETH staked at its reference point, securing the consensus layer with roughly $79 billion at the prices used in the report.
If stake representing one-third of the total stops attesting, the network can lose the two-thirds supermajority needed to finalize new checkpoints. A simple count of one-third of servers or validator keys does not establish that outcome because stake is distributed unevenly.
The report also maps software concentration. Lighthouse represented about 54% of consensus clients, followed by Prysm at 23%, Teku at 9%, and Caplin at 6%.
On the execution side, Geth held about 41%, Nethermind 38%, Besu 16%, Erigon 3%, and Reth roughly 2%. Independent clients provide a safety layer when one implementation encounters a serious bug.
The security tradeoffs were also the subject of a current interview with Cambridge’s research lead:
THE STARTING BLOCK:
What is more secure? Proof of Work or Proof of Stake?
Cambridge research lead @alexneumueller weighs up the security guarantees of Bitcoin vs Ethereum after unpacking @CambridgeAltFin latest Ethereum report with @gazza_jenks.
Link to the broadcast 👇 pic.twitter.com/Uxd2szLDwS
— The Block (@TheBlockCo) July 10, 2026
Ethereum.org’s proof-of-stake documentation confirms that checkpoint finality requires votes representing at least two-thirds of total staked ETH. Stake representing one-third can delay finality by preventing that supermajority from forming.
Ethereum includes an inactivity-leak mechanism for that situation. After the chain fails to finalize for more than four epochs, stake held by validators that are not attesting with the majority gradually loses weight until the active majority can regain two-thirds and resume finalization.
That recovery path does not erase the operational risk. It is designed to restore liveness over time while imposing an economic penalty on inactive stake.
Cambridge’s environmental numbers show the scale of the Merge’s achievement. The report estimates continuous power demand fell from roughly 2.4 gigawatts immediately before the September 2022 transition to about 0.90 megawatts today, equal to approximately 7.87 gigawatt-hours per year.
The estimated annual climate footprint is about 2.37 kilotonnes of carbon-dioxide equivalent, roughly 99.98% below the final pre-Merge level. Cambridge estimates that 56.4% of the electricity mix serving the measured node population comes from sustainable sources, including renewables and nuclear power.
The Merge delivered a dramatic energy reduction. Cambridge’s new map puts the next pressure points in plain view: cloud concentration, jurisdiction, client diversity, and the distribution of the stake that keeps Ethereum finalizing.
Join the conversation!
We have no tolerance for comments containing violence, racism, profanity, vulgarity, doxing, or discourteous behavior. If a comment is spam, instead of replying to it please click the icon below and to the right of that comment. Thank you for partnering with us to maintain fruitful conversation.
