What Is MEV? Maximal Extractable Value on Ethereum Explained

What MEV Actually Is

Every Ethereum block is an ordered list. Someone has to decide what goes in it and in what sequence, and that decision is worth money. Maximal extractable value, or MEV, is the name for that money.

Ethereum own developer documentation defines it as the maximum value that can be extracted from block production in excess of the standard block reward and gas fees, by including, excluding, and changing the order of transactions in a block. That definition is deliberately broad, because MEV is not a flaw in one smart contract that a patch could fix. It is a structural consequence of three facts about public blockchains: pending transactions are visible before they settle, execution is fully deterministic so anyone can simulate the outcome in advance, and somebody always has discretion over ordering.

Put those three together and you get an entire industry of bots, block builders and infrastructure companies whose business is capturing the difference.

Why the Name Changed from Miner to Maximal

The original term was miner extractable value, coined while Ethereum still ran on proof of work and miners assembled blocks directly. After the Merge in September 2022, validators propose blocks instead, so the acronym was rebased to maximal extractable value. Ethereum documentation records exactly this shift.

The rename was more than cosmetic. Under proof of stake, the validator who wins a slot rarely builds the block themselves. In practice they buy a finished block from a specialist. So the party with real ordering power is a professional block builder, and the validator is closer to a landlord collecting rent on blockspace. Maximal, rather than miner, captures that the value accrues across a whole supply chain.

Where MEV Comes From

The public mempool. When you send a normal Ethereum transaction it is broadcast to a peer-to-peer network of pending transactions before it is included in a block. Anyone running a node can read it, including your exact trade size and your slippage tolerance. That visibility window is the raw material for most retail-facing MEV. It is also why a stuck or pending transaction behaves the way it does: until a block includes it, it is public and unconfirmed.

Deterministic execution. A bot can simulate your pending swap against current pool state and know, to the wei, what price you will get and what price the pool will be at afterwards. There is no uncertainty to price in.

Ordering discretion. Priority fees create a bidding market for position within a block, and whoever assembles the block can reorder freely. If you have ever wondered why gas fees spike during volatile moments, competitive MEV bidding is part of the answer.

The Main MEV Strategies

DEX arbitrage

The canonical example. Two decentralised exchanges briefly show different prices for the same pair. A searcher buys on the cheaper venue and sells on the dearer one inside a single atomic transaction, so if the second leg fails the whole thing reverts and no capital is at risk. This is riskless arbitrage in the textbook sense, and it is what keeps quoted prices roughly consistent across venues. The cost is borne by the liquidity providers whose pool was mispriced, not by an identifiable victim.

Liquidations

When a borrower on a lending protocol falls below its collateral threshold, anyone can repay the debt and claim a liquidation bonus. Searchers compete to be first. Slow liquidations are how lending markets accumulate bad debt, so this competition is genuinely load-bearing for DeFi solvency. The cost falls on the borrower who was already in breach of terms they agreed to.

Sandwich attacks

Here the framing changes. A searcher watches the mempool for a large pending swap, buys the same asset immediately before it to push the price up, allows the victim transaction to execute at the degraded price, then sells immediately after. The profit is manufactured from the victim, not from a pre-existing inefficiency.

The encouraging news is that this activity has been shrinking. Data from MEV analytics firm EigenPhi, covering more than 95,000 sandwich attacks between November 2024 and October 2025 and reported by Cointelegraph Research, shows monthly sandwich extraction on Ethereum falling from close to 10 million dollars in late 2024 to roughly 2.5 million dollars by October 2025, with average monthly net profit after gas around 260,000 dollars across 2025 — the gap between the two reflects how much of gross extraction is consumed by gas and builder payments. That decline happened while DEX volumes were rising, which suggests protective tooling is genuinely working rather than the opportunity simply drying up.

Just-in-time (JIT) liquidity

A liquidity provider mints a tightly concentrated position immediately before a large swap, collects the trading fees from it, and burns the position immediately after. Uniswap Labs published an analysis of the practice covering May 2021 to July 2022, finding it supplied just over 2 billion dollars of liquidity across 8,287 attempts, amounting to roughly 0.3 percent of all liquidity demand, and that over 95 percent of it came from a single account. Crucially, their finding was that JIT liquidity gave those traders better execution, not worse. The contested part is the effect on passive liquidity providers, who see their fee share diluted on exactly the largest and most profitable trades. Note that this dataset is now several years old and should be read as directional.

CEX-DEX arbitrage

Likely the largest and certainly the least visible category. Prices move first on high-throughput centralised exchanges, then someone arbitrages the lagging on-chain pool. Unlike DEX-to-DEX arbitrage this is not atomic, because one leg sits off-chain, so it requires inventory and carries real risk. A 2025 study by Wu, Sui, Thiery and Pai estimated 233.8 million dollars extracted across 7,203,560 identified CEX-DEX arbitrages between August 2023 and March 2025, and found that three searchers captured roughly three quarters of both volume and value. The same paper found searcher profitability tracks how tightly integrated a searcher is with a block builder, which is the clearest evidence available that vertical integration is where the advantage now lives.

MEV Strategies Compared

The MEV Supply Chain

Modern Ethereum MEV flows through a specific pipeline: user, then wallet or RPC endpoint, then either the public mempool or a private order flow channel, then searcher, then builder, then relay, then proposer.

Searchers run bots that spot opportunities and package them into bundles, which Flashbots documentation describes as arrays of signed transactions plus metadata specifying the conditions under which they may be included. Bundles are conditional: if the opportunity evaporates, nothing lands and nothing is paid for.

Builders collect bundles and mempool transactions and assemble the most valuable block they can. They see everything that reaches them, which is why builder neutrality matters so much.

Relays sit between builders and validators. They hold the full block, hand the validator only a header plus a bid, and release the body once the validator has signed a commitment. This stops validators from stealing a builder strategy after seeing it.

Proposers, meaning validators, simply pick the highest bid.

PBS and MEV-Boost

Proposer-builder separation (PBS) is the design principle that whoever proposes a block should not need to be the one who builds it. MEV-Boost is Flashbots implementation of that idea, running outside the Ethereum protocol as optional validator software. It has become the default way blocks get made: the large majority of Ethereum blocks are now sourced from the builder market rather than assembled locally by validators.

That market is uncomfortably concentrated. Per relayscan.io, over the seven days to 20 July 2026, Titan built 51.22 percent of MEV-Boost blocks, Quasar 23.09 percent, BuilderNet 10.78 percent and Eureka 10.27 percent — four operations covering roughly 95 percent of block production. On the relay side, relay.ultrasound.money delivered 27.82 percent of payloads and titanrelay.xyz 21.70 percent, with two bloXroute relays together accounting for another 37 percent or so.

Concentration matters because of censorship. MEV Watch, which reads relay payload data from relayscan.io, put the share of blocks delivered by censoring relays at roughly 42 percent as of 19 July 2026, up meaningfully over the preceding month.

Ethereum plans to bring this in-protocol. Enshrined PBS (EIP-7732) is a headline proposal for the Glamsterdam upgrade, which ethereum.org lists as planned for the second half of 2026. The censorship-resistance counterpart, fork-choice enforced inclusion lists (FOCIL, EIP-7805), was selected as a consensus-layer headliner for the following upgrade rather than being bundled into Glamsterdam, per the Ethereum Foundation April 2026 checkpoint.

How to Protect Yourself From MEV

You cannot opt out of MEV existing. You can make yourself an unprofitable target.

• Switch your wallet RPC to a protected endpoint. Flashbots Protect routes transactions to a private mempool instead of broadcasting them publicly, which removes the visibility that front-running and sandwiching depend on. It also adds revert protection so failed transactions do not cost you gas, and pays MEV refunds — the Flashbots settings documentation states that by default 90 percent of the MEV refund goes to the transaction sender, with the remaining 10 percent to the validator. MEV Blocker is the main alternative, running a back-running auction and rebating the bulk of the winning bid to the sender; note that per CoW DAO’s own announcement its operation transferred to Consensys’ Special Mechanisms Group in January 2026.
• Tighten your slippage tolerance. A sandwich attack is only profitable up to the point where the victim transaction would revert. Loose slippage is an open invitation. Set it as tight as the pool volatility allows rather than accepting a wallet default.
• Use intent-based or batch-auction venues for large trades. CoW Protocol settles orders in batches using uniform directed clearing prices, so every order for the same pair in a batch clears at the same price. When ordering within a batch cannot change your price, ordering-based attacks stop working.
• Split large orders. A single very large swap is a far better target than several smaller ones, and the profitability of an attack scales with the price impact you create.
• Prefer deep liquidity. Thin pools produce large price impact from modest size, which is precisely the condition sandwich bots look for.
• Do not simply raise your gas price to escape. Bidding higher does not hide your transaction; it just makes you more expensive to outbid while remaining fully visible.

MEV on Layer 2s

Rollup MEV looks quite different, mostly because almost every major layer 2 network still runs a single centralised sequencer. There is generally no public mempool for outside searchers to watch and no competitive builder auction, so value that would be shared across a supply chain on mainnet tends to be internalised by the sequencer operator or auctioned explicitly.

Arbitrum is the clearest worked example. Its Timeboost mechanism replaces first-come-first-served ordering with a sealed-bid second-price auction for an express lane. Per Arbitrum documentation, rounds last 60 seconds, bidding closes 15 seconds before each round begins, and transactions not sent through the express lane receive a deliberate 200-millisecond delay on their arrival timestamp. The winner pays the second-highest bid to a beneficiary designated by the chain owner. Notably, the express lane controller cannot reorder other transactions or guarantee top-of-block placement, so it is a latency advantage rather than a licence to sandwich.

The direction of travel is towards decentralised and shared sequencing, where a distributed validator set orders transactions for one or several rollups. That work is real and progressing, but production-grade decentralised sequencing is not the default on any major rollup today, and anyone claiming otherwise is ahead of the evidence.

What Is Still Unsolved

Encrypted mempools are the most-discussed structural fix. Shutter Network and Primev have been building a threshold-encrypted mempool for Ethereum PBS pathway, in which transactions can only be decrypted once a threshold number of independent keypers cooperate to release key shares, with builder commitments backed by slashable stake. Shutter puts cumulative MEV drained from Ethereum users since 2020 at more than 1.8 billion dollars — a figure from a party building a solution, so treat it as vendor-reported rather than independent.

Even if encryption ships broadly, several problems remain open. Metadata such as transaction size and timing still leaks. Decryption adds latency in a system already fighting for milliseconds. Threshold schemes replace one trust assumption with a different one spread across keypers, and hardware-based approaches such as trusted execution environments introduce their own vendor dependency. Cross-domain MEV spanning multiple chains and centralised exchanges sits outside any single protocol jurisdiction. And builder-searcher vertical integration, which the CEX-DEX research identifies as the current source of edge, is an economic dynamic that no cryptographic primitive addresses.

Risks and Honest Framing

The single most important thing to internalise is this: MEV cannot be eliminated, only redistributed or minimised. Every proposal on the table — PBS, order flow auctions, encrypted mempools, batch auctions, inclusion lists — changes who captures the value and how visible the capture is. None of them make the value stop existing.

That has practical consequences. Protected RPCs shift trust to the RPC operator rather than removing trust. Order flow auctions that rebate value to users also concentrate order flow with whoever runs the auction. Batch auctions depend on a healthy competitive solver set. Every mitigation carries a trade-off, and treating any of them as a complete solution is how people get hurt.

Nothing here is financial or investment advice. Interacting with DeFi protocols carries risk of total loss from smart contract bugs, oracle failures, market volatility and operator failure, entirely separate from MEV. Do your own research and never commit funds you cannot afford to lose.

Which Protection Should You Choose?

If you make occasional swaps of modest size on deep pools, switching your wallet to a protected RPC and tightening slippage covers the overwhelming majority of your realistic exposure, and costs you nothing.

If you regularly move size large enough to create visible price impact, add a batch-auction or intent-based venue for those trades specifically, and split orders rather than sending one large swap.

If you are a builder or protocol team, the relevant question is order flow: where your users transactions go by default determines who captures the value they create, and that default is a design decision you are making whether you think about it or not.

If you run a validator, the concentration and censorship figures above are the argument for relay diversity, including neutral relays, rather than optimising purely for the highest bid.

How This Guide Was Built

This guide is an editorial synthesis of primary documentation, public on-chain data and published academic research. Definitions and protocol mechanics were taken from Ethereum’s official developer documentation and roadmap pages, Flashbots documentation for the auction, Protect and MEV-Share components, Arbitrum documentation for Timeboost, and CoW Protocol documentation for batch auctions. Block builder and relay market share figures come from relayscan.io and censorship share from MEV Watch, both dated in the report. Sandwich attack data is EigenPhi analysis as reported by Cointelegraph Research; JIT liquidity data is Uniswap Labs published analysis; CEX-DEX figures come from a 2025 academic paper by Wu, Sui, Thiery and Pai.

We did not run our own measurements or benchmarks, and we have deliberately avoided quoting aggregate all-time MEV totals, because published estimates vary widely depending on methodology and which strategies are counted. Where a number could not be tied to a named source it has been described qualitatively instead. Market share and censorship figures change daily; check the live dashboards before relying on them.

Conclusion

MEV is what happens when ordering has value and someone gets to choose the order. On Ethereum that reality has been industrialised into a supply chain of searchers, builders, relays and proposers that most users never see but all users pay into. Some of it — arbitrage, liquidations — does genuine work keeping markets honest and lending protocols solvent. Some of it, particularly sandwiching, is a straight transfer from traders to bots, though the measured trend there is downward as protective tooling spreads.

The realistic goal is not a world without MEV. It is a world where extraction is competitive rather than monopolised, where users capture a share of the value their own transactions create, and where no small group of builders can quietly decide what gets included. Ethereum roadmap is moving that way with enshrined PBS and inclusion lists. Until it arrives, a protected RPC and disciplined slippage settings do most of the work for most people.

This is an editorial synthesis of documentation, public data, and community reports; see our [methodology](/methodology). Verify current details with each project.

Sources: ethereum.org MEV documentation, ethereum.org Glamsterdam roadmap, Ethereum Foundation Checkpoint 9, Flashbots Auction overview, Flashbots Protect settings guide, relayscan.io, MEV Watch, Uniswap Labs on JIT liquidity, Measuring CEX-DEX Extracted Value and Searcher Profitability, Arbitrum Timeboost documentation, CoW Protocol batch auctions, MEV Blocker, BuilderNet, Shutter Network on encrypted mempools