Why it matters
In a conventional exchange the matching order is set by rule and enforced by the operator, and abusing it is a regulatory matter. On a public blockchain the order is the block producer's discretion, and there is no rule to abuse. Every pending transaction is visible before it executes, so the producer knows what traders are about to do and can insert its own transactions before, after or instead of theirs.
That discretion has a value large enough to have restructured Ethereum's block production into a supply chain of specialist firms, and it sets a floor on what users pay to trade that no fee schedule discloses. For an institution routing size onchain it is a component of execution cost, measurable like any other adverse selection.
How it works
ethereum.org 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". Three categories account for most observed activity. Arbitrage between venues, where a price difference between two pools is closed inside one atomic transaction and is, in the documentation's words, "true, riskless arbitrage". Liquidations, where searchers race to identify borrowers who can be liquidated and collect the fee. And sandwich trading, buying ahead of a large order and selling after it, which the same page notes is "riskier as it isn't atomic".
Since 2022 the extraction has been intermediated. Searchers identify opportunities and submit bundles; builders assemble full blocks from bundles and ordinary transactions; relays hold the blocks and forward the most profitable one to the proposer. The proposing validator runs MEV-Boost, which Flashbots describes as "open source middleware run by validators to access a competitive block-building market", commits to the header of the highest-bidding block without seeing its contents, and receives the bid. In practice the validator sells its ordering right rather than exercising it.
Solana has no in-protocol equivalent; a client performs the same function. Jito Labs describes Jito-Solana as "the first third-party, MEV-boosted validator client for Solana", where validators connect to a block engine to earn from traders' bundles and can share the proceeds with stakers to attract delegation.
Economic mechanism
Follow the money in one direction and the structure is legible. A trader submits an order that will move a price; a searcher observes it, computes the profit available from ordering around it, and bids most of that profit to a builder to win inclusion; the builder bids most of what it received to the proposer to win the block. The proposer keeps the bid, and the trader pays through a worse fill than the quoted price implied.
The competitive dynamic explains why the intermediaries earn thin margins and the validator earns the rent. Searchers compete on speed and modelling and their surplus is bid away; builders compete on block value and theirs is bid away; the proposer is a monopolist for one slot. This is why MEV shows up in staking returns rather than in the accounts of the firms doing the work. Flashbots estimates that "validators running MEV-Boost can increase staking rewards by over 60%", and an Ethereum research note modelling the issuance curve took total extractable value at 300,000 ether a year when computing validator income.
The uncomfortable implication is that ordering revenue subsidises consensus security: a protocol that eliminated extraction would reduce the return to staking, and therefore the capital securing it, unless it replaced the income. That is the substance of the burn proposals, which route the rent to the protocol instead.
Where it breaks is at the top of the supply chain. Two dependencies are load-bearing: the relay is trusted to pay the proposer and not to leak block contents to a builder, and the proposer is trusted to publish the block it committed to. Neither is enforced by consensus, so both are enforced by reputation. And because ordering revenue rises with scale (more flow, better models, amortised infrastructure), the mechanism pushes toward concentration in the layer whose diversity is the security assumption.
Participants
Searchers are trading firms, mostly small and undisclosed; builders are a handful of specialist operators; relays sit between them and proposers as trusted intermediaries. Validators and staking operators receive the bids and, where they pass MEV through, compete for delegation with it. Protocol developers set the rules: the Ethereum Foundation funds the research on proposer-builder separation and issuance, while Anza and the Solana Foundation determine Solana's client behaviour and so what discretion a leader has. Application developers respond at their own layer: Uniswap through fee tiers and routing, Aave through liquidation parameters. Venues built to remove the discretion are a distinct category: Lighter proves its own order matching and liquidations, so price-time priority can be checked rather than assumed.
Examples
The Peraire-Bueno prosecution is the only attempt so far to establish where extraction becomes a crime. The US Attorney for the Southern District of New York announced charges on 15 May 2024, alleging the brothers "fraudulently gained access to pending private transactions and used that access to alter certain transactions and obtain their victims' cryptocurrency", in an exploit "believed to be the very first of its kind". About $25m was taken, and the defence argued the conduct was fair play in a competitive market. After a four-week trial the case ended in a mistrial on 11 November 2025, the jury having failed to progress; prosecutors sought a retrial, and no verdict on whether the conduct was theft or competition had been reported by 25 July 2026.
Liquidation revenue is the second case, and it prices a service rather than extracting from a counterparty: a protocol sets a bonus high enough that a searcher will close an undercollateralised position within one block, the borrower pays it, and the same figure determines whether bad debt accumulates in a fast market. It is a governance parameter.
Risks and limitations
The first genuine dispute is whether extraction is a cost to users or a fee for a service. Arbitrage and liquidation close pricing errors and remove bad debt; their absence would leave stale prices and undercollateralised loans. Sandwich trading has no such defence: it transfers value from a trader to a searcher and worsens the price the trader receives. Practitioners who defend the activity argue from the first two categories; those who call it extraction argue from the third, and aggregate figures combining all three are evidence for neither.
The second is the legal question the mistrial left open. If exploiting the ordering system is theft, much routine searcher activity is exposed; if it is competition, so is the conduct the government charged. The distinction the government drew rests on deceiving the relay rather than on the ordering itself, and it has not been tested to verdict.
The third is centralisation. Proposer-builder separation kept block production accessible to small validators and created concentrated builder and relay layers instead. Relays are unremunerated, trusted and few; one that colludes with a builder, or fails, degrades the market for every proposer using it. On Solana the same pressure runs through delegation: a validator sharing bundle proceeds with stakers converts extraction into stake share. Whether the trade was worth making is unresolved.
Two measurement limitations temper any figure. Extraction inside private order flow is not observable in public data, so published totals are lower bounds. And a share of what is labelled MEV is ordinary market making that would exist on any venue, so cross-chain comparisons compare conventions as much as behaviour.
Key metrics
Payments from builders to proposers, per block and in aggregate, are the best available price of ordering discretion. Alongside them: the share of blocks built by the largest builder and relay; realised slippage on large onchain orders against the quoted price, which is what the cost looks like to a user; liquidation volume and the bonus paid; and ordering revenue as a share of validator income.
Related research
The ethereum.org documentation is the standard definition and taxonomy, and Flashbots' MEV-Boost overview describes the auction and its trust assumptions from the party that built it. The Ethereum research note on a capped issuance curve gives the scale of ordering revenue relative to issuance. The Southern District of New York charging documents and the November 2025 mistrial are the primary record of the legal question.