Ethereum drew about 2,601 megawatt-hours of electricity across the entire global network, or 0.0026 terawatt-hours per year. That represents a cut of more than 99.988% versus the network’s pre-Merge draw and an approximately 99.992% cut in carbon footprint, from 11,016,000 tonnes CO2e down to 870 tonnes CO2e.
Key Takeaways
- Post-Merge Ethereum’s global annualized electricity consumption sits at 0.0026 TWh per year (2,601 MWh), per CCRI’s bottom-up estimate published on ethereum.org.
- An independent Cambridge Centre for Alternative Finance estimate puts the figure higher at roughly 7.34 GWh per year.
- The Merge was executed on September 15, 2022, completing Ethereum’s transition to proof-of-stake and deprecating proof-of-work consensus.
- The carbon footprint of a single Ethereum transaction fell from 109.71 kg CO2e to 0.01 kg CO2e, recorded on September 20, 2022.
- Ethereum’s post-Merge draw is roughly 57,000 times lower than Bitcoin at 149 TWh per year, and 176 times lower than Netflix at 0.457 TWh.
- As of June 15, 2026, the consensus layer carried 1,239,795 active validators, a net gain of 96,462 in the first five and a half months of 2026.
- Ethereum Foundation projections put a home staker’s full stack, including a consensus-layer node, at around 100 watts continuous.
Editor’s Choice
- Yearly carbon emissions across the entire Ethereum network total 870 tonnes CO2e, applying regional carbon-intensity factors.
- The European Union Blockchain Observatory reported that the Merge cut network-wide electricity demand by more than 99.9% relative to the proof-of-work baseline.
- Under proof-of-stake, Ethereum runs at roughly 2.62 megawatts steady-state, on par with powering 2,100 American homes.
- Digiconomist estimated Ethereum drew 44.49 TWh per year before the Merge, with roughly 264 kWh per single transaction at peak proof-of-work usage.
- The proof-of-work-era Ethereum figure benchmarked on the same ethereum.org table sat at 21 TWh per year.
- Approximately 35.7 million ETH was staked across the consensus layer by September 2025, held by 1,060,332 active validators and representing roughly 30% of circulating supply.
- The Pectra upgrade activated on May 7, 2025, and introduced compounding validators that can consolidate up to 2,048 ETH per key, replacing the original 32-ETH cap.
Ethereum’s Annual Energy Footprint After the Merge
- 2,601 MWh per year: Post-Merge Ethereum’s global annualized electricity consumption per CCRI.
- 870 tonnes CO2e per year: Total network-wide carbon footprint after applying regional carbon-intensity factors.
- 99.988% cut: Electricity reduction against the pre-Merge baseline; 99.992% cut in carbon.
- From 11,016,000 to 870 tonnes CO2e: the absolute drop in annual emissions.
By the numbers: Post-Merge Ethereum draws roughly 2,601 megawatt-hours across the entire global network per year, versus a pre-Merge baseline near 23 million megawatt-hours in the same units per EY, a cut of more than 99.988% in electricity and a matching 99.992% cut in annual carbon output from 11,016,000 tonnes CO2e down to 870 tonnes CO2e per CCRI’s bottom-up estimates.
| Metric | Post-Merge Ethereum | Pre-Merge Ethereum | Reduction |
|---|---|---|---|
| Annual electricity | 2,601 MWh | ~23,000,000 MWh | 99.988% |
| Annual carbon | 870 tCO2e | ~11,016,000 tCO2e | 99.992% |
| Per-transaction carbon | 0.01 kg CO2e | 109.71 kg CO2e | ~99.99% |
| Per-transaction energy | ~0.03 kWh | ~264 kWh | ~99.99% |
Source: ethereum.org and CCRI, 2024; Digiconomist Ethereum Energy Index, 2024; EY Switzerland, 2023.
The canonical post-Merge number published by the Ethereum Foundation tracks a bottom-up estimate from the Crypto Carbon Ratings Institute. CCRI measured electricity consumption across Ethereum clients and applied it to the observed node distribution, arriving at 2,601 megawatt-hours per year for the entire global network. Decrypt’s coverage of the CCRI report matched that number and characterized it as a 99.988% reduction in electricity use versus the pre-Merge proof-of-work network.
Carbon follows the electricity number closely. EY’s post-Merge analysis reported that Ethereum’s overall electricity draw fell to just 2,600 megawatt-hours per year against a pre-Merge baseline of 23 million megawatt-hours. The gap in orders of magnitude explains why regulators and ESG-focused investors quietly stopped citing Ethereum in the same breath as Bitcoin after the Merge shipped.
Two Independent Estimates Compared
- 2,601 MWh per year (CCRI, ethereum.org): bottom-up, four Ethereum clients weighted by observed node share.
- 7,340 MWh per year (Cambridge CBNSI): wider node population, three hardware configurations, seven-day moving average.
- ~2.8× multiple between the two: same order of magnitude, both dramatically below any pre-Merge baseline.
- ~2,500 MWh per year (Digiconomist post-Merge): per-gas-unit modelling; not per-transaction.
- ~22,950 MWh per year: Ethereum Foundation’s 2021 ex-ante projection (2.62 MW steady-state × 8,760 hours).
- 25% of nodes at 16.56 to 37.42 watts: Cambridge lower-bound configuration.
- 50% of nodes at 36.82 to 59.95 watts: Cambridge mid-range best-guess configuration.
- 25% of nodes at 139.90 to 186.76 watts: Cambridge upper-bound configuration.
The single best-known number for Ethereum’s post-Merge draw is CCRI’s, but it is not the only serious estimate. The Cambridge Blockchain Network Sustainability Index publishes its own best-guess figure of roughly 7.34 GWh of annualized electricity, updated daily with a seven-day moving average, and using the Armiarma crawler to identify active consensus-layer nodes. That is roughly 2.8 times the CCRI headline number.
The Cambridge figure runs higher partly because it counts every consensus-layer node the crawler sees, including nodes running heavier hardware profiles. The Cambridge best-guess distribution assumes 25% of nodes run a lower-bound configuration at 16.56 to 37.42 watts per node, 50% run the mid-range at 36.82 to 59.95 watts, and 25% run an upper-bound configuration at 139.90 to 186.76 watts. In practice, both estimates land at the same qualitative conclusion: the network runs on the power budget of a small town, not a country.
Recent Developments
- June 2026: The consensus layer reached 1,239,795 active validators, adding 96,462 net validators in the first five and a half months of 2026.
- May 2025: The Pectra upgrade activated on May 7, 2025, and introduced compounding validators that can consolidate up to 2,048 ETH per key, allowing them to hold a materially larger share of staked ETH.
- September 2025: Staked ETH reached 35.7 million across 1,060,332 active validators, representing roughly 30% of circulating supply.
- 2023-2025: Lido Finance’s liquid-staking share fell from more than 32% to approximately 24.4%, indicating a diversification of the staking landscape.
- October 2022: The CCRI post-Merge report established the 2,601 MWh baseline and the 99.988% reduction figure now published on the ethereum.org energy page.
Pre-Merge Baseline: What the Network Consumed Under Proof-of-Work
- 44.49 TWh per year: Digiconomist’s conservative pre-Merge estimate at peak PoW usage.
- 112 TWh per year: EY’s higher pre-Merge estimate using a different modelling frame.
- 21 TWh per year: ethereum.org’s benchmark PoW-era figure now used for post-Merge comparisons.
- 264 kWh per transaction: Digiconomist’s peak pre-Merge per-transaction footprint.
- 23 million MWh per year: EY’s aggregate baseline figure for full-network PoW-era electricity draw.
The headline reduction figure only lands if the pre-Merge baseline is defensible. Digiconomist’s Ethereum Energy Consumption Index estimated an annualized network draw of roughly 44.49 terawatt-hours immediately before the Merge, with a per-transaction footprint on the order of 264 kilowatt-hours. EY’s independent framing put the same pre-Merge number at approximately 112 TWh per year.
The proof-of-work-era Ethereum figure the Ethereum Foundation itself now benchmarks against sits between those two. On the current energy-consumption comparison table, ethereum.org lists proof-of-work Ethereum at 21 TWh per year, roughly 8,100 times higher than post-Merge draw. For a sense of what that number bought, the Ethereum Foundation’s pre-Merge projection cited Digiconomist’s 44.49 TWh figure and characterized the post-Merge network as approximately 2,000 times more energy efficient.
Per-Transaction Energy: Then and Now
- 0.02 to 0.03 kWh per transaction post-Merge (Digiconomist, per gas unit).
- 35 Wh per transaction in the Ethereum Foundation’s pre-Merge projection.
- 0.01 kg CO2e per transaction post-Merge (EY, September 20, 2022).
- 109.71 kg CO2e per transaction the day before the Merge.
- 264 kWh per transaction at the pre-Merge peak (Digiconomist).
| Per-transaction metric | Pre-Merge | Post-Merge | Change |
|---|---|---|---|
| Energy (kWh) | ~264 | ~0.03 | 99.99% cut |
| Carbon (kg CO2e) | 109.71 | 0.01 | 99.99% cut |
| Foundation projection (Wh) | n/a | ~35 | reference figure |
Source: Digiconomist Ethereum Energy Index, 2024; EY Switzerland, 2023; Ethereum Foundation Blog, May 2021.
Per-transaction estimates are where the numbers get politically loaded, because they can be sliced by gas units, by output transactions, or by an “average” of the two. Digiconomist notes that its figures are measured per unit of gas consumed rather than per output transaction, which means the exact per-transaction figure can vary depending on the gas profile of a given block, and lands in a range of roughly 0.02 to 0.03 kWh per transaction. The projection sat in the same order of magnitude, at roughly 35 Wh per transaction, comparable to twenty minutes of television.
EY’s post-Merge number for the carbon component fell to 0.01 kg CO2e per transaction on September 20, 2022, versus 109.71 kg pre-Merge. That is the single largest ESG improvement in the history of the PoW vs PoS debate.
Validator Hardware and Power Draw
- ~100 watts continuous: a home staker’s full stack (Ethereum Foundation projection).
- 16.56 to 37.42 watts: Cambridge lower-bound hardware configuration (~25% of nodes).
- 36.82 to 59.95 watts: Cambridge best-guess mid-range configuration (~50% of nodes).
- 139.90 to 186.76 watts: Cambridge upper-bound server-grade configuration (~25% of nodes).
- 2.62 megawatts steady-state: total network-wide draw, on par with powering 2,100 American homes.
The energy footprint of an Ethereum validator is set almost entirely by the always-on hardware running the consensus-layer node and execution client, not by any peak “mining” load. The pre-Merge projection assumed a home staker running a consensus-layer node, 5.4 validator clients, and a full Ethereum 1 node draws roughly 100 watts continuous. Cambridge’s higher upper-bound configuration models a small subset of nodes running at 139.90 to 186.76 watts, while the majority run under 60 watts.
Aggregated up to the whole network, the Foundation projected a steady-state draw of roughly 2.62 megawatts total across proof-of-stake Ethereum, on par with powering 2,100 American homes. That figure predated the Merge but has held up as an order-of-magnitude anchor in every independent measurement since.
Ethereum vs Bitcoin, Data Centers, and Netflix
- 190 TWh: global data centers (~73,000× Ethereum).
- 149 TWh: Bitcoin (~57,000× Ethereum).
- 131 TWh: gold mining (~50,000× Ethereum).
- 34 TWh: gaming in the USA (~13,000× Ethereum).
- 0.457 TWh: Netflix (~176× Ethereum).
- 0.0026 TWh: post-Merge Ethereum (baseline).
| Reference load | Annual energy (TWh) | Multiple of Ethereum |
|---|---|---|
| Global data centers | 190 | ~73,000× |
| Bitcoin | 149 | ~57,000× |
| Gold mining | 131 | ~50,000× |
| Gaming in the USA | 34 | ~13,000× |
| Proof-of-work Ethereum (2021) | 21 | ~8,100× |
| 19 | ~7,300× | |
| Netflix | 0.457 | ~176× |
| Post-Merge Ethereum | 0.0026 | 1× |
Source: ethereum.org energy consumption page, 2024.
Ethereum’s post-Merge draw is small enough that the honest comparisons are no longer to countries, but to individual services and industries. On the ethereum.org comparison table, global data centres consume approximately 190 TWh per year, Bitcoin at 149 TWh, gold mining at 131 TWh, gaming in the USA at 34 TWh, Google at 19 TWh, and Netflix at 0.457 TWh.
The Netflix comparison is the most useful public-facing benchmark; a single streaming service consumes 176 times more electricity than every Ethereum node on the planet combined.
Layer-2 Rollups Push the Number Lower Again
- ~0.03 kWh per mainnet Ethereum transaction (Digiconomist post-Merge).
- ~0.03 kWh mainnet per transaction, divided further across users as rollup batching rises.
- ~2 more orders of magnitude below mainnet on a per-user basis.
- Zero further protocol change required for the effect: the architecture layer, not consensus, does the work.
Ethereum’s mainnet numbers already sit far below every meaningful comparator, but activity is increasingly moving to Layer-2 rollups that batch many logical transactions into a single mainnet settlement. As an illustration, when a rollup batches many logical transactions into a single mainnet settlement, each user’s share of that settlement’s energy falls well below the sourced ~0.03 kWh mainnet per-transaction figure, the exact reduction tracking whatever batching ratio a given rollup runs. See our Layer-2 gas-market breakdown for the current mainnet-vs-rollup share of activity.
Key finding: Rollup batching compounds Ethereum’s post-Merge efficiency by roughly two more orders of magnitude on a per-user basis, dropping each user’s share of settlement energy further below the ~0.03 kWh mainnet per-transaction figure as batching rises, without any further protocol change on the base layer or extra hardware from validators.
For adjacent grid-side deployments, our blockchain energy trading data set tracks how the settlement layer itself is being used in energy markets.
How much energy does Ethereum use per year?
Ethereum’s post-Merge network draws approximately 0.0026 TWh per year, equivalent to 2,601 megawatt-hours, according to CCRI figures published on ethereum.org. A separate estimate from the Cambridge Centre for Alternative Finance places the figure at roughly 7.34 GWh per year. Both estimates land within the same order of magnitude and dramatically below any pre-Merge baseline.
How much power does one Ethereum validator use?
A single Ethereum home staker running a consensus-layer node, 5.4 validator clients, and a full execution client draws roughly 100 watts continuous, per the Ethereum Foundation’s projections. Cambridge’s node-distribution assumptions place half of validators between 36 and 60 watts, with a small upper-bound cohort running 140 to 187 watts on heavier hardware. That footprint is comparable to a single laptop or a small home server left on.
Is Ethereum better than Bitcoin for the environment?
On ethereum.org’s benchmark table, Bitcoin’s annual electricity draw sits at 149 TWh versus Ethereum’s 0.0026 TWh, making Ethereum roughly 57,000 times lower. The two networks are no longer in the same energy category after the Merge. The comparison that reads more honestly today is against streaming services and data-centre workloads rather than against Bitcoin.
Conclusion
Ethereum’s post-Merge energy footprint sits between the CCRI figure of 2,601 megawatt-hours per year and the Cambridge best-guess of 7.34 gigawatt-hours; either way, an order of magnitude smaller than Netflix’s streaming operation and roughly 57,000 times lower than Bitcoin. The 99.988% cut in electricity and 99.992% cut in carbon are now settled parts of the network’s identity rather than unproven projections: three years of independent measurement across CCRI, Cambridge, Digiconomist, and the EU Blockchain Observatory have all landed in the same order of magnitude, and the underlying validator hardware trajectory is trending toward lower per-node draw rather than higher.
Watching the number for 2026 and beyond, three things matter. The current validator population above 1.24 million, the growing Pectra-era compounding-validator cohort, and the rising share of activity moving to Layer-2 rollups will all keep the per-user carbon footprint on a downward path even without further protocol changes, a compounding efficiency story that Ethereum’s critics have been slow to price in.
