Decentralization Index
7.2out of 10
| Measure | Weight | Score |
|---|---|---|
| Nakamoto coefficient | 25% | 3.5 |
| Full nodes | 15% | 8.4 |
| Client diversity | 10% | 6.4 |
| Hardware to run a node | 10% | 10.0 |
| Token concentration | 15% | 10.0 |
| Insider allocation at launch | 15% | 8.5 |
| Control over upgrades | 10% | 6.0 |
How the index works · method v1.0
Insider allocation at launch
8.5Widely distributed
Insiders 8.3% Insider-controlled 8.3% Public 83%
| Group | Counted as | Share |
|---|---|---|
| Ether sale (42-day public sale for BTC) | Public | 83% |
| Early contributors endowment (0.099x ether sold) | Insider | 8.3% |
| Long-term endowment to the non-profit foundation (0.099x ether sold) | Insider-controlled | 8.3% |
Shares of: Genesis supply of about 72.0 million ETH (72,009,990.5 per Etherscan): about 60 million sold in the 2014 ether sale, plus two endowment pools each equal to 0.099x the ether sold
Insider lockups: Not disclosed
Shares come from the official sale terms: two pools, 'each 0.099x the initial quantity of ether sold', so sale = 1/1.198 and each pool = 0.099/1.198 of genesis supply. This is close to Etherscan's genesis split of 60,000,000 crowdsale and 12,009,990 other ETH (83.3% and 16.7%). The sale counts as public because anyone could buy during the 42 days on one published schedule that applied to every buyer alike (blog.ethereum.org, July 22, 2014). The foundation endowment counts as insider-controlled because the foundation decides how to spend it. Judgement call: The Block (December 2018) reports that the 12 million premined ETH were split into 6 million for 85 early contributors, 3 million for the Ethereum Foundation, and 3 million for a 'developer purchase program' that let about 50 foundation workers buy on crowdsale terms. If that split is used, the 3 million would count as insider (a sale open only to selected buyers). We kept the official two-pool split because The Block links a spreadsheet for the 85 contributors but gives no source for splitting the foundation pool into 3 million for the Ethereum Foundation and 3 million for a developer purchase program, and we found no project document describing that program. Vesting: the official post says ether was not usable until the genesis block. We found no lockup terms for either pool. Coins issued later as mining or staking rewards are not part of the launch supply.
All measures
| Measure | Value | As of | How we got it |
|---|---|---|---|
| Nakamoto coefficient Fewest block producers who could together disrupt the network. | 4 Counted by the company that runs the validators, using ETH already staking on November 3, 2026. That was 44.3 million ETH, so one third is 14.77 million (validatorqueue.com: https://www.validatorqueue.com/). Lido is a staking pool whose validators are run by many separate operators (Rated counts 566), so we split it among them instead of counting it as one. The largest companies we could size: Coinbase ran 12.16% of staked ETH on its own validators, averaged over April to June 2026. That is its latest report, and it was 12.17% the quarter before (https://www.coinbase.com/blog/ethereum-validator-performance-report-q2-2026). We use that share for November. Coinbase's 4.76 million ETH average would be 10.7% of today's total, so the share is the cautious choice unless Coinbase has grown faster than the network since June. Rated also lists Coinbase Cloud at 0.38% (https://explorer.rated.network/o/Coinbase%20Cloud?network=mainnet&timeWindow=1d&idType=nodeOperator). On Rated's explorer on November 3, 2026, Binance's own pool held 7.97% (3.543 million ETH across 76,925 validators), which we count as run by Binance (https://explorer.rated.network/o/Binance?network=mainnet&timeWindow=1d&idType=pool). Kraken's pool held 6.76% (pools view: https://explorer.rated.network/?network=mainnet&view=pool&timeWindow=1d&page=1&pageSize=15&poolType=all). Bitmine reported 5,227,037 ETH staked as of November 1, 2026, 11.8% of staked ETH. It says part of it is staked on MAVAN, its own staking platform, and that it uses staking partners it doesn't name (https://prnewswire.com/news-releases/bitmine-immersion-technologies-bmnr-announces-eth-holdings-reach-6-06-million-tokens-and-total-crypto-and-total-cash--marketable-securities-holdings-of-14-5-billion-302926341.html). MAVAN also runs 0.61% of staked ETH for Lido (https://explorer.rated.network/o/MAVAN?network=mainnet&timeWindow=1d&idType=nodeOperator). Pier Two, which Bitmine bought (https://research.lido.fi/t/pier-two-joins-mavan-a-bitmine-immersion-technologies-company/11380), runs 0.09% (https://explorer.rated.network/o/Pier%20Two?network=mainnet&timeWindow=1d&idType=nodeOperator). Take the worst case: MAVAN runs all of Bitmine's ETH, and Coinbase Cloud's stake isn't already in Coinbase's report. Then MAVAN, Coinbase and Binance together run at most 33.0%, just under one third (33.3%). We found no other company running more than Binance's 7.97%. So at least four of the companies we can size are needed, and we record 4. Public data can't link about a third of staked ETH to the company that runs it. That includes the 16.3% Rated lists for SSV. SSV splits each validator's key among several operators, and stakers choose their own operators (https://docs.ssv.network/learn/introduction/). Rated doesn't show which companies run them. The true figure may be higher, or 3 if a company we can't see runs more than about 8.3%. Chainspect showed 5 on November 3, 2026, without explaining its method (https://chainspect.app/chain/ethereum). | November 3, 2026 | Researched |
| Block producers Validators, miners or pools that produce blocks. | – Left blank: our definition counts validators that produced blocks in the last 30 days, and we found no beacon-chain explorer or dataset that publishes the number of distinct validators that proposed at least one block from October 4 to November 2, 2026. A 30-day window has only 216,000 twelve-second slots, so no more than 216,000 validators can propose in a month. For context, 813,754 validators were active on November 3, 2026 (validatorqueue.com: https://www.validatorqueue.com/), and Chainspect (https://chainspect.app/chain/ethereum) showed 815,294 the same day. Many validators are run by the same operators, so even a count of proposers overstates the number of independent block producers. | – | – |
| Full nodes Independent computers that check every rule. | 11.9K Nodes found by Etherscan's node tracker on November 3, 2026 (Etherscan describes the page as statistics of detected nodes; it doesn't say whether consensus-layer nodes are included). Top locations in the last 24 hours: United States 38.6%, Germany 18.3%, United Kingdom 9.4%. Crawlers only see reachable nodes, so the true number is likely higher. | November 3, 2026 | Researched |
| Largest client share Nodes running the most popular software. Lower is safer. | 48% Lighthouse at 48.39% of consensus-layer nodes, Miga Labs data shown on clientdiversity.org (updated daily). No reliable current node-level figure exists for execution clients. clientdiversity.org still charts Ethernodes at Geth 50.13% but says the Ethernodes data 'has been removed' and that it's unclear how it relates to real use. The self-reported supermajority.info survey (Geth and Nethermind 43% each, based on 59.4% network coverage) is updated manually and sits under clientdiversity.org's 'Data is stale' heading. Validator-based consensus estimates conflict: Rated shows Teku 53.86%, and Blockprint shows Teku 99.83%, far from every other source. Every source carries the note 'Data may not be 100% accurate.' | November 3, 2026 | Researched |
| Hardware to run a node What it takes to verify the network yourself. | Home computer ethereum.org says clients 'are able to run on consumer grade computers.' Minimum full-node requirements: CPU with 2+ cores, 16 GB RAM (32 GB recommended for stability), 2 TB NVMe SSD (likely exceeded by 2027), 25+ Mbit/s bandwidth. Recommended specs (EIP-7870) are 4+ cores, 32 GB RAM, a 4 TB NVMe SSD and 50/15 Mbit/s, which is closer to a powerful desktop. We chose the home-computer category based on the published minimum. | November 3, 2026 | Researched |
| Who can change the rules Whether one party can upgrade or pause it. | Core team coordinates Changes are proposed as Core EIPs, discussed on All Core Devs calls, built by several independent client teams and bundled into scheduled upgrades. ethereum.org: 'node operators can choose not to run the latest Ethereum client' and 'Protocol Developers have no way to force people to adopt network upgrades.' No single party can force an upgrade. Classed as coordinated because core developers plan and schedule upgrades that operators then choose to adopt. | November 3, 2026 | Researched |
| Held by top 100 addresses Excluding known exchanges and contracts where possible. | 7.6% The 100 largest addresses left after exclusions (raw ranks 11 to 143 on Etherscan's top accounts list, pages 1 and 2; ranks 101 to 143 are on page 2: https://etherscan.io/accounts/2?ps=100) hold 7.56% of Etherscan's 122,186,167 ETH supply. Excluded (43 addresses): contracts (Beacon Deposit Contract, Wrapped Ether, Tornado.Cash 100 ETH pool, WithdrawDAO, Liquity Active Pool); frozen funds (Polkadot MultiSig and Iconomi MultiSig 1, both labelled 'Parity Bug Funds frozen' by Etherscan); bridges and rollup escrows (Base Portal, Arbitrum Bridge, Optimism ETHLockbox, Mantle Proxy, Polygon Ether Bridge, Robinhood Bridge, Binance-Peg Tokens); and exchanges and custodians (Binance, Robinhood, Upbit, Bitfinex, Kraken, Gemini, QuadrigaCX, Deribit/Coinbase Prime Custody, Paxos, Crypto.com, Gate, OKX, Bybit, Ceffu). Kept: unlabelled addresses, named individuals, EF 2, and addresses tagged as hackers or exploiters. Vb 3 and EF 2 are kept although they are contract wallets (Etherscan marks both 'Smart Account by Safe'), because each belongs to one person or organization rather than holding pooled funds. Limits: about 44.3 million staked ETH (validatorqueue.com, November 3, 2026: https://www.validatorqueue.com/) sits in validator balances, not ordinary addresses, so it falls outside this measure. Unlabelled addresses may include more exchange, custodian or contract wallets. | November 3, 2026 | Researched |
| Transactions per second 30-day average of user transactions, not consensus votes. | 21 53,436,105 transactions from October 4 to November 2, 2026 (UTC, Etherscan daily transaction chart CSV export) divided by 2,592,000 seconds. Ethereum's validator votes (attestations) are not recorded as execution-layer transactions (ethereum.org: https://ethereum.org/en/developers/docs/consensus-mechanisms/pos/attestations/), so nothing needed to be excluded. For comparison, Chainspect (https://chainspect.app/chain/ethereum) showed 24.29 TPS over the last hour on November 3, 2026; we don't use its theoretical figure. | November 2, 2026 | Researched |
| Daily active addresses 30-day average. Addresses are not people. | 566.4K Average of Etherscan's daily 'Unique Address Total Count' (unique addresses active as a sender or receiver) from October 4 to November 2, 2026 (UTC), from the chart's CSV export. Daily values ranged from 457,836 to 745,589. growthepie (https://www.growthepie.com/chains/ethereum) showed 495.9K for its latest day on November 3, 2026, using its own method. | November 2, 2026 | Researched |
| Typical fee Median for a simple transfer over 30 days. | – Left blank: our definition is the median fee for a simple transfer over 30 days, and we found no source that publishes a daily simple-transfer fee for Ethereum from October 4 to November 2, 2026. For context only, and not the figure we define: growthepie's median cost of all transaction types (txcosts_median_usd, Ethereum mainnet: https://api.growthepie.com/v1/fundamentals.json) gives a median of daily medians of $0.033 over those 30 days, with daily medians from $0.015 to $0.128. A simple transfer usually costs less: growthepie's hourly fees table showed a native-transfer median of $0.0202, against $0.0654 for all transactions, at 13:00 UTC on November 3, 2026 (https://api.growthepie.com/v1/fees/table.json). ethereum.org: 'A standard ETH transfer requires a gas limit of 21,000 units of gas' (https://ethereum.org/en/developers/docs/gas/). Chainspect's 1-day average fee was $0.142 on November 3, 2026 (https://chainspect.app/chain/ethereum), but an average is pulled up by large transactions. | – | – |
| Time to finality Until a payment practically cannot be reversed. | ~15 minutes ethereum.org: 'It takes about 15 minutes for an Ethereum block to finalize.' Finality needs votes from two-thirds of staked ETH on two checkpoints, which come at the start of each epoch of 32 twelve-second slots (ethereum.org: https://ethereum.org/en/developers/docs/consensus-mechanisms/pos/), so the minimum is two epochs. Chainspect (https://chainspect.app/chain/ethereum) shows 12m48s (768 seconds), which is the two-epoch minimum. A new block can be added in each 12-second slot, but it isn't final until finality is reached. | November 3, 2026 | Researched |
| Outages, last 24 months Times the network halted or stopped producing blocks. | 0 No halts or loss of finality on mainnet from November 2024 to November 2026 in the sources we checked. The largest disruption in the window: on December 4, 2025, after the Fusaka upgrade, a Prysm client bug cut participation to about 75% (finality needs two thirds). 248 of 1,344 slots were missed (epochs 411439 to 411480). Cointelegraph (December 5, 2025: https://cointelegraph.com/news/ethereum-prysm-bug-fusaka-client-diversity-risk) reported participation was under 9 percentage points above the two-thirds needed for finality. The last finality losses, on May 11 and 12, 2023, fall outside the window. | November 3, 2026 | Researched |
| New coins per year Yearly issuance as a share of supply. | 0.9% Gross issuance. Etherscan's total supply rose 86,325.8 ETH from October 3 to November 2, 2026, and 3,067.0 ETH was burned in fees over the same 30 days (https://etherscan.io/chart/dailyethburnt). That makes 89,392.8 ETH created, or about 1.09 million ETH a year: 0.89% of the 122.19 million supply. After burns, supply grew about 0.86% a year over those 30 days and 0.83% over the 12 months to November 2, 2026. Issuance rises with the amount of ETH staked, and burns rise with network demand. | November 2, 2026 | Researched |
See how Ethereum stacks up
Compare it side by side with up to three other networks.
Comparing networks isn't advice to buy their coins. Spotted an error? Email us with a source and we'll correct it.