Who really controls a proof-of-stake chain? Nobody alone, and a few clerks more than most
No single party controls Ethereum, but control isn't spread evenly either. Large staking providers, the most popular client software and the relays that pass blocks to validators each hold a lever. The last word belongs to the people who run nodes and choose which chain to follow.
Grandma Rosa has a simple question about the deposit box by the well: who's in charge of it? There's no mayor. Nobody signs off on the notebook. So who decides?
The honest answer has four parts. A few clerks hold many villagers' coins in the box. Most houses write with the same make of pen. A handful of messengers carry finished pages to the signers, and some won't carry certain letters. And behind them all stand the villagers, who decide which notebook they keep copying.
Nobody holds every lever. But some hold more than others, and it pays to know who. This article closes the proof-of-stake section by counting them.
What it is
Control of a proof-of-stake chain means the power to decide which blocks get written, which get finalised, and which rules the chain follows. On Ethereum it sits in several places at once: the stake, the client software, the block supply chain and the people who run nodes.
The rules themselves aren't voted on chain. ethereum.org says "Ethereum governance happens offchain", and gives the last check to the people running the software: node operators can refuse to upgrade their client, and a change nobody runs never happens (ethereum.org).
How it works: four levers, counted
1. The stake. ethereum.org sets out what share of stake buys what (ethereum.org). More than a third can "prevent the chain from finalizing". More than half "could dominate the fork choice algorithm". Two-thirds "can finalize their preferred chain". Now the sums, with 44,339,549 ETH of active stake read on 19 November 2026 from beaconcha.in:
- One-third: about 14.78 million ETH. Two-thirds: about 29.56 million ETH.
- The liquid-staking protocol Lido showed 9,759,632 ETH staked on its site the same day (Lido). That's 22.0%, about 5.02 million ETH short of one-third. The site lists more than 700 node operators running that stake.
- For comparison, Rocket Pool's rETH in circulation came to about 355,700 ETH at Etherscan's price (Etherscan): 0.8%. Examples, not recommendations.
2. The pen. Every validator runs client software, and a bug in one client hits everyone using it. clientdiversity.org explains the line: "If a client with >33% marketshare forks, it will prevent the chain from finalizing." Above two-thirds, a buggy client could finalise the wrong chain, and the page says its validators couldn't return to the real chain without being slashed (clientdiversity.org). Read on 19 November 2026, its trackers disagreed on who leads: Miga Labs had Lighthouse at 50.78% of consensus clients, Rated had Teku at 53.86%. Different counts, same warning: one client above a third.
3. The messengers. Many Ethereum blocks are built by specialist builders and passed to validators through relays, a business that "MEV explained" covers. Some relays filter out transactions touching sanctioned addresses. For 17 November 2026, MEV Watch counted 13.3% of MEV-Boost block deliveries through such relays, using data from the Flashbots-run relayscan.io (MEV Watch). It warns this leaves out blocks validators build themselves, so it "may overstate" censorship (MEV Watch methodology). In plain numbers: if a filtering relay skips your transaction, about 87% of deliveries come through neutral relays, so it can still be included in a later block. Delay, not denial.
4. The villagers. When stake fails, people decide. ethereum.org: "The Ethereum social layer is the ultimate backstop" (ethereum.org). It has happened before. On 20 July 2016, at block 1,920,000, Ethereum forked to move over 3.6 million ETH drained from the DAO contract into a refund contract. "Some miners refused to fork", and their chain became Ethereum Classic (ethereum.org).
Why it matters
- Big piles are big targets. ethereum.org warns that "Large amounts of staked ETH under the control of a few organizations create conditions for censorship" (ethereum.org). A pile under one-third can't stop finality alone. Two piles together might.
- Software monoculture is the quiet risk. No attacker is needed, just one bug in a client most validators share.
- The backstop is human. The social layer can reject an attacker's chain, but it means messy coordination and possibly a chain split. It's a fire brigade, not a lock.
Where people get it wrong
- "Lido controls Ethereum." On the 19 November 2026 readings, it held 22.0% of stake, spread over many operators. That's a lot of weight, not a majority, and below the one-third line.
- "Validators vote on Ethereum's rules." They don't. Rule changes come from EIPs, client developers and node operators who choose to upgrade, off chain.
- "Censorship means your transaction is banned." A filtering relay can delay it. Neutral relays and self-built blocks still include it.
- "Counting validators shows how decentralised it is." Since May 2025 one validator can hold up to 2,048 ETH, and one firm can run thousands. Count the stake and who runs it.
Ask whose pile you're adding to
If you stake through any provider, look up its share of all stake on a dated reading before you start. Then ask: does my deposit make the biggest pile in the village bigger?
The notebook stays safe while no clerk's pile reaches a third of the box. Ben is waiting patiently for the day one does.
Frequently asked questions
Who controls Ethereum?
No single party. Influence is spread across stakers and staking providers, client software teams, block builders and relays, and the node operators who choose which software and chain to run. Protocol rules change through an off-chain process, not an on-chain vote.
How much stake would an attacker need on Ethereum?
According to ethereum.org, more than a third of all stake can stop the chain finalising, more than half can dominate which chain is followed, and two-thirds can finalise a chain alone. With 44,339,549 ETH staked on 19 November 2026, one-third was about 14.78 million ETH.
What share of Ethereum stake does Lido have?
On 19 November 2026, Lido's site showed 9,759,632 ETH staked, about 22.0% of the 44,339,549 ETH of active stake on beaconcha.in that day. That stake is run by many separate node operators.
Why does client diversity matter?
Because a bug in one client affects every validator running it. If a client used by more than a third of validators fails, the chain stops finalising; above two-thirds, a faulty client could finalise the wrong chain.
Can Ethereum validators censor transactions?
Individual block builders and relays can leave transactions out, and some filter transactions touching sanctioned addresses. MEV Watch counted 13.3% of MEV-Boost block deliveries through such relays for 17 November 2026. Other blocks still include those transactions, so the usual effect is a delay.
Sources
- ethereum.org, Proof-of-stake attack and defense
- ethereum.org, Ethereum governance
- ethereum.org, The history of Ethereum (DAO fork)
- ethereum.org, Pooled staking
- beaconcha.in, finalised epoch 490,260: active stake and validators (read on 19 November 2026)
- Lido, home page with total ETH staked and node operators (read on 19 November 2026)
- Etherscan, Rocket Pool ETH (rETH) token page (read on 19 November 2026)
- clientdiversity.org, Ethereum client diversity (read on 19 November 2026)
- MEV Watch, share of MEV-Boost blocks via OFAC-filtering relays, data from relayscan.io (reading for 17 November 2026)
- MEV Watch, Methodology

