TL;DR: After the merge, Ethereum will utilize approximately ~99.95% less energy.
In the coming months, Ethereum is set to finalize its transition to Proof-of-Stake, introducing various enhancements that have been anticipated for years. Now that the Beacon chain has been operational for several months, it’s time to examine the concrete figures. One area we’re particularly eager to analyze is the new energy consumption estimates as we cease consuming the energy equivalent of a small nation for consensus.
Currently, there are no definitive statistics on energy usage (or the specific hardware being utilized), so what follows is a rough approximation of Ethereum’s future energy consumption.
Since many individuals operate multiple validators, I opted to use the number of unique addresses that have made deposits as a proxy for evaluating how many servers are currently in use. While many stakers may have employed multiple eth1 addresses, this largely balances out against those with redundant configurations.
As of this writing, there are 140,592 validators associated with 16,405 unique addresses. This figure is significantly influenced by exchanges and staking services; removing these entities leaves us with 87,897 validators presumed to be staking from home. This suggests that the average home-staker operates around 5.4 validators, which seems like a plausible estimate.
Power Requirements
What is the power requirement for running a beacon node (BN), 5.4 validator clients (VC), and an eth1 full-node? Based on my personal configuration, it’s roughly 15 watts. Joe Clapis (a Rocket Pool developer) recently operated 10 VCs, a Nimbus BN, and a Geth full node from a 10Ah USB battery bank over 10 hours, meaning this setup averaged about 5W. It’s improbable that the average staker would have such an optimized setup, so let’s estimate it at 100W total.
If we multiply this by the 87k validators mentioned earlier, home-stakers are estimated to consume around ~1.64 megawatts. Estimating the power use of custodial stakers is more complex, as they manage tens of thousands of validator clients with redundancy and backups.
To simplify, let’s assume they utilize 100W per 5.5 validators. From discussions with staking infrastructure teams, this is a significant overestimate. The actual number is likely closer to 50 times less (And if you’re part of a custodial staking team using more than 5W/validator, I’d love to connect and offer assistance).
In total, a Proof-of-Stake Ethereum thus consumes approximately 2.62 megawatts. This is not on the scale of nations, provinces, or even large cities, but rather that of a small town (around 2100 American households).
For context, the current energy consumption of Ethereum’s Proof-of-Work (PoW) consensus is akin to that of a medium-sized country, which is necessary to maintain a PoW chain’s security. As the term implies, PoW arrives at consensus through the fork that has completed the most “work.” Two primary methods exist to enhance the “work” rate: increasing mining hardware efficiency and deploying more hardware concurrently. To stave off potential attacks, miners must perform “work” at a rate that exceeds any adversary’s capabilities. Because attackers typically use similar hardware, miners must continuously operate large amounts of efficient hardware to prevent being out-mined, all of which consumes substantial power.
Under PoW, the price of ETH and the hashrate are positively correlated. Hence, as the price spikes, so does the power consumption by the network. In contrast, under Proof-of-Stake, when the value of ETH rises, the network’s security also strengthens (due to the increased value of the staked ETH), while the energy requirements remain stable.
Some Comparisons
Digiconomist estimates that Ethereum miners currently consume about 44.49 TWh annually, which translates to 5.13 gigawatts continuously. This indicates that PoS is roughly 2000 times more energy-efficient based on these conservative estimates, equating to a reduction of at least 99.95% in total energy consumption.
If you prefer a per-transaction energy consumption perspective, that’s approximately ~35Wh/tx (averaging ~60K gas/tx), or equivalent to around 20 minutes of television viewing. Conversely, Ethereum’s PoW operates on the energy equivalent of a household for 2.8 days per transaction, while Bitcoin uses 38 house-days of energy.
Looking Forward
While Ethereum is still utilizing PoW for the moment, this will not last much longer. Recently, we have witnessed the introduction of the first testnets for The Merge, the title given to Ethereum’s transition from PoW to PoS. Numerous engineering teams are diligently working to ensure that The Merge occurs as swiftly as possible, without compromising safety.
Scaling solutions (such as rollups and sharding) will contribute to further reducing the energy consumed per transaction by utilizing economies of scale.
The era of Ethereum’s energy-intensive operations is approaching its conclusion, and I hope this is a trend that the entire industry will follow.
Special thanks to Joseph Schweitzer, Danny Ryan, Sacha Yves Saint-Leger, Dankrad Feist, and @phil_eth for their input.
