The MEV Market in 2026

1. Where it started

MEV began when the first decentralized exchanges went live on Ethereum, around 2017. The moment there were pools with publicly visible pending transactions and a competitive field of bots, the extraction was already happening. It just did not have a name yet.

In April 2019, a group of researchers gave it one. Flash Boys 2.0 was the first systematic empirical study of transaction ordering exploitation on decentralized exchanges, and it introduced a term the field had been missing: Miner Extractable Value, the maximum value extractable by a miner controlling transaction ordering in a Proof of Work block. The term was later renamed Maximal Extractable Value as the ecosystem moved to Proof of Stake validators. The concept stayed the same.

The paper documented something that had been visible on-chain for months but not formally analyzed. Bots were competing for priority transaction ordering, bidding up gas fees in what the authors called Priority Gas Auctions (PGAs). The strategies they observed were varied: arbitrage between DEX pools, liquidations in lending protocols, frontrunning of large trades, sandwich attacks around predictable price-moving transactions. In each case, the bot's edge came from the same source: controlling where its transaction landed relative to others in the block.

What they measured was striking for the time. From a subset of DEX transactions, they identified a lower bound of $6 million in pure arbitrage revenue captured by these bots. The market was already oligopolistic: a handful of addresses dominated the pure revenue space for extended periods before being displaced by competitors.

But the most important finding was not the size of the extraction market. It was the systemic risk MEV creates at the consensus layer. When the value extractable in a single block exceeds the block reward itself, producers have an incentive to attempt time-bandit attacks: rewriting recent history to recapture past opportunities. MEV as a consensus security problem, not just an extraction opportunity, is what made the paper foundational.

Any system executing ordered transactions at scale will generate extractable value. That value creates incentives that can undermine the system itself.

2. The specialization of the market

Flash Boys 2.0 described a market where individual bots competed openly in the mempool. By 2021, that market had changed structurally.

As DeFi grew, so did the economic value of ordering control. Arbitrage, liquidations, sandwich attacks, and backrunning all scaled with TVL. The fees bots were willing to pay in PGAs rose accordingly. And at a certain point, the competition for block space became expensive enough that it made sense to specialize.

Four distinct roles emerged:

Searchers identify MEV opportunities and construct the transactions or bundles that capture them. They range from individual developers running custom bots to institutional operations with proprietary simulation infrastructure. The edge is algorithmic and latency-sensitive.

Builders aggregate bundles from multiple searchers and construct full candidate blocks optimized for total extracted value. They compete not on speed but on the quality of their simulation and the breadth of their orderflow access.

Relays act as trusted intermediaries between builders and validators, communicating block bids while keeping the block contents confidential until a validator commits. They became a critical trust layer in the system, and a centralization point.

Validators select the highest-paying block from the available bids. Under MEV-Boost, a middleware developed by Flashbots that implemented this separation, most Ethereum validators began accepting blocks from external builders rather than constructing their own.

This architecture, known as Proposer-Builder Separation (PBS), solved a real problem: it prevented validators from needing to run MEV extraction infrastructure themselves, democratizing access to MEV revenue. But it introduced a new one. The builder market concentrated rapidly. A handful of builders came to process the majority of Ethereum blocks, creating chokepoints where censorship, orderflow favoritism, and rent extraction became structural concerns rather than hypothetical ones.

The competitive advantage in this market was no longer primarily algorithmic. It was relational and infrastructural: access to exclusive orderflow, co-location with key nodes, and integration with protocols willing to route transactions privately. The open mempool war of 2019 had evolved into a closed infrastructure race.

3. MEV goes cross-chain

As DeFi fragmented across multiple chains and L2s, MEV followed. But cross-domain MEV is not simply regular MEV applied across chains. The research had to develop a new framework to reason about it.

SoK: Cross-Domain MEV, published in August 2023, provided that framework. It introduced a distinction that matters for understanding how value is extracted across domains: ordering extraction versus signal extraction. Ordering extraction derives value purely from controlling transaction sequence using on-chain data, the same dynamic Flash Boys 2.0 identified. Signal extraction derives value from off-chain information: prices on centralized exchanges, events on other blockchains, real-world data.

An extractor who knows the CEX price before the on-chain AMM price has caught up can profitably trade against the spread without any mempool visibility, purely from the signal. This is CEX-DEX arbitrage, and it is the dominant form of cross-domain MEV by value.

The paper also introduced the concept of time-extractable value: unlike intrinsic MEV that must be captured immediately, some opportunities grow in value the longer an extractor waits, analogous to options pricing. This optionality makes time-extractable MEV harder to eliminate through standard mitigations.

Pandora's Box, published in January 2025, provided the first large-scale empirical measurement of cross-chain arbitrage. Analyzing nine blockchains over one year, the researchers identified 260,808 cross-chain arbitrages generating a lower-bound profit of \(9.5 million on \)465 million in traded volume, with an average profit margin of 66.92%, significantly higher than the sub-10% typical of single-chain atomic arbitrage.

The most counterintuitive finding: 67.63% of cross-chain arbitrages do not use a bridge. The dominant strategy is Sequence-Independent Arbitrage (SIA), covered in depth in How to Design a Cross-Chain Arbitrage Strategy: the arbitrageur holds inventory on both chains and executes both legs independently, without transferring assets between them.

The reason is purely economic: SIA completes in a median of 10 seconds. The bridge-dependent alternative (SDA) takes 246 seconds. At that speed difference, price windows close before SDA can execute.

The data also shows a market maturing in the same direction as single-chain MEV: increasing concentration among a small number of arbitrageurs, and a measurable rise in private mempool adoption. Sophisticated actors are hiding their transactions from competitors, routing them directly to builders rather than broadcasting to the public mempool. Private mempools reduce reactive MEV exposure — frontrunning and sandwich attacks require visibility into pending transactions. But they address only one layer of the MEV stack. Structural and anticipatory MEV survive encryption entirely, as analyzed in detail in MEV Layer Cake.

The second paper, Bunny Hops (October 2025), asked a different question: can cross-chain arbitrage scale vertically through multiple hops? The answer from the data is unambiguous. Across 2.4 billion transactions and 34.8 million bridge interactions on 12 chains, the researchers found exactly 10 multihop arbitrage instances: 8 three-hop and 2 four-hop. The distribution follows a power law: every additional hop multiplies execution cost, bridge latency, and price exposure. A three-hop arbitrage takes an average of 7.2 minutes to complete. A four-hop takes 11.6 minutes. Only 60% of identified multihop transactions generated positive returns.

Cross-chain MEV scales in breadth, not depth. More chains, more pairs, more opportunities, but no meaningful move toward complex multi-hop paths. The structural constraints are too severe.

4. Where it goes from here

PBS solved the immediate problem of validator centralization, but created a new one: a handful of builders now control the majority of block production on Ethereum. The open mempool competition of 2019 became a closed infrastructure oligopoly. The question the industry is now trying to answer is whether that centralization is a temporary phase or a structural endpoint.

Flashbots, the organization that built MEV-Boost and much of the PBS tooling, has been the most explicit about this problem and the most public about their answer. In their 2023 SUAVE vision and their 2026 roadmap, they propose a phased path toward decentralized block building.

Phase 0 is where most of the market sits today: monolithic systems running on company-controlled cloud infrastructure. Builders, relays, and orderflow routers operate as centralized intermediaries. The technical architecture is a server, not a network.

Phase 1 introduces replicated privacy through trusted execution environments (TEEs). Multiple parties can share economically sensitive data and achieve crash fault tolerance without trusting a single operator. Flashbots is building toward this with BuilderNet. The test: any single entity going down produces no change in block output.

Phase 2 moves to co-built blocks, where no single entity can be attributed authorship of a block. Construction is distributed across cooperating nodes through purely technical means, without legal or political arrangements as the trust substrate.

Phase 3 is the long-term objective: a globally distributed building network where meaningful economic activity occurs on nodes worldwide, and any machine can permissionlessly enter and exit. Geographic distribution becomes the measure of political decentralization.

The SoK paper on cross-domain MEV, published two years before this roadmap, had already mapped the protocol-level responses: shared sequencers and orderflow auctions were identified as the approaches with the greatest potential to mitigate cross-domain MEV. Both face significant technical barriers (atomic cross-chain guarantees, censorship resistance, latency requirements), but the research direction aligns with where Flashbots is building. The infrastructure question and the protocol question are converging on the same answer.

One ecosystem worth noting separately is Substrate-based chains. While Ethereum is solving the block building decentralization problem through external infrastructure like PBS and TEEs, Substrate chains can approach it differently at the protocol level. The architecture allows pallets to execute logic before any user transaction at every block — the chain itself acts as the first mover, capturing liquidations, correcting price discrepancies, redistributing value to users or treasury. No external builder infrastructure required. It is a native capability that most MEV research has not examined in depth.

The ambition behind the Flashbots roadmap is larger than Ethereum. Flashbots explicitly frames it as a blueprint for any system that sequences transactions: L2s, payment chains, cross-chain infrastructure. Block building, in this vision, becomes the foundation layer for blockchain interoperability. The same primitives required to coordinate builders across trust boundaries are the same primitives required to coordinate execution across chains.

MEV started as a symptom of open mempools. It became an infrastructure industry. It is now being redesigned as a public layer, programmable, verifiable, and distributed, that any protocol or chain can plug into. Whether that vision materializes is an open question. That the market has evolved far enough to ask it is not.