MEV Extraction on Layer 2s: Differences & Opportunities
Maximal Extractable Value (MEV) has long been a defining—and controversial—feature of Ethereum’s mempool. On Layer 1, searchers compete in a public, transparent auction for block space, leading to frontrunning, sandwich attacks, and billions in extracted value. But as activity migrates to Layer 2 rollups, the MEV landscape undergoes a fundamental transformation. The shift from a permissionless block-building market to sequencer-controlled ordering creates both new constraints and unexploited opportunities.
This guide examines how MEV manifests differently on optimistic and zero-knowledge (ZK) rollups, the unique extraction strategies that emerge, and what these dynamics mean for everyday users. You’ll learn why sandwich attacks become harder on some L2s, how private order flow creates a two-tiered system, and why cross-rollup arbitrage is the next frontier. Whether you’re a searcher seeking alpha or a user wanting to avoid unnecessary slippage, understanding MEV on Layer 2 is essential.
- MEV on L2 is fundamentally different from L1 due to sequencer-controlled ordering and private mempools, not just a scaled-down version.
- Optimistic rollups currently host more traditional MEV (arbitrage, liquidations) than ZK-rollups, which tend to suppress transaction-level extraction but enable sequencer-level auctions.
- Sandwich attacks are nearly impossible on fully private L2s, but sequencers themselves can extract MEV, creating centralization risks.
- Searchers must pivot from public mempool competition to securing private order flow and building relationships with sequencers.
- Cross-rollup and cross-domain MEV is a high-growth frontier, driven by fragmented liquidity and latency between different L2s.
- User safety improves for individual trades (less frontrunning) but users face potential censorship and sequencer extraction unless protocols enforce fair ordering.
What Makes MEV on Layer 2 Different?
At its core, MEV arises from the ability to reorder, include, or censor transactions within a block. On Ethereum L1, this power is distributed among validators and builders competing in a public mempool. Layer 2 rollups, however, introduce a central gatekeeper: the sequencer. Sequencers are responsible for ordering transactions and submitting batches to the L1. Most rollups initially launch with a single, centralized sequencer, giving that entity sole control over transaction ordering.
This structural shift has several consequences:
- No public mempool by default: Many L2s route transactions directly to the sequencer, not to a gossip network. This eliminates the transparent auction that fuels most L1 MEV extraction.
- Private order flow dominance: Without a public mempool, searchers must obtain order flow directly from users or through RPC endpoints, creating information asymmetry.
- Sequencer as monopoly: Unless forced by protocol rules (e.g., fair ordering), the sequencer can extract all available MEV itself or auction the ordering rights.
Understanding these fundamentals is crucial: L2 MEV is not a scaled-down version of L1 MEV—it’s a different game with different players and rules.
Optimistic vs ZK-Rollups: The MEV Landscape Compared
While all rollups share a sequencer architecture, the technical differences between optimistic and ZK rollups create distinct MEV environments.
| Factor | Optimistic Rollups (e.g., Arbitrum, Optimism) | ZK-Rollups (e.g., zkSync Era, Starknet) |
|---|---|---|
| Transaction finality | Delayed (~7 days for challenge period) | Near-instant via validity proofs |
| Sequencer centralization | Often single sequencer initially; plans for decentralization | Single sequencer currently; some have multiple proposers |
| Mempool visibility | Often private; some expose delayed mempools | Typically fully private to sequencer |
| Cross-domain complexity | Low (EVM compatible, shared state via L1) | Higher (non-EVM, different execution environments) |
| Typical MEV activity | Arbitrage, liquidations (within L2 and cross-L2 via bridges) | Less arbitrage due to fragmented liquidity; more protocol-level auction extraction |
The key takeaway: Optimistic rollups currently host more traditional MEV because they mirror EVM mechanics and have more mature DeFi ecosystems. ZK-rollups, with faster finality and often different VM designs (e.g., Cairo, Zinc), naturally suppress certain MEV vectors while opening others—like sequencer auctions for block space.
“On a ZK-rollup, the sequencer can prove valid state transitions instantly. This makes frontrunning less profitable because the target transaction cannot be ‘unconfirmed’ during the challenge window. The MEV extraction moves from the transaction level to the sequencing level.” — adapted from research by Barabander et al.
MEV Extraction Strategies on L2s: What Still Works?
Despite the private ordering, some L1 strategies survive—with modifications:
- Sandwich attacks: These become rare on fully private L2s because the attacker cannot see the victim’s transaction before the sequencer includes it. However, if the sequencer leaks order flow (e.g., via a public mempool alternative) or if search bots watch pending batches, sandwiches remain possible, albeit with more delay.
- Arbitrage: Still viable, especially between L2 DEXs and between L2 and L1. Because L2 block times are faster, arbitrage opportunities are shorter-lived but more frequent. Searchers with low-latency access to the sequencer can capture them.
- Liquidations: On L2s with lending protocols, liquidations are still profitable. The sequencer can prioritize liquidation transactions, effectively collecting the liquidation reward as MEV. Some L2s allow third-party liquidators to compete via prioritize mechanisms.
- Cross-domain MEV: Unique to L2s. Arbitrage between two L2s, or between L2 and L1, requires bridging—but finality delays create temporal arbitrage windows. This is an area where MEV is actually higher than on L1 due to fragmented liquidity and slower bridges.
For searchers, the primary skill shifts from competing in a public auction to securing private order flow and maintaining close relationships with sequencers or their RPC endpoints.
Unique Opportunities for Searchers
The L2 environment, while less chaotic than L1, offers distinct avenues for profit:
- Sequencer auctions: Several rollups (notably Optimism’s OP Stack, and Arbitrum) plan to auction block space or ordering rights. Searchers can bid for the right to include their bundles, much like PBS (Proposer-Builder Separation) on L1, but with simpler execution.
- Delayed transaction inclusion: Some L2s expose a delayed mempool (e.g., Arbitrum’s public mempool for delayed inbox). Searchers can extract MEV from those transactions before they are forced into upcoming batches.
- Cross-rollup arbitrage bots: With multiple L2s (Arbitrum, Optimism, Base, zkSync, etc.) and each having different liquidity depths, price disparities of 1–2% are common. Bots that can simultaneously trade on two L2s via fast bridges can capture this spread. The catch: bridge latency and costs.
- MEV from forced transactions: On optimistic rollups, users can force transactions through the L1 inbox. Searchers watching the L1 can predict upcoming L2 state changes and preemptively trade on the L2.
These opportunities require deep technical knowledge of each rollup’s sequencing mechanism and its interaction with the base layer. The barrier to entry is higher than on L1, but the rewards can be more consistent due to less competition.
Implications for Users: Are You Safer on L2?
For the average user, moving to an L2 rollup changes the MEV risk profile:
- Less frontrunning: In a private-mempool L2, you won’t be sandwiched or frontrun by bots for common swaps. The sequencer acts as a trusted ordering party (though it could theoretically frontrun you).
- New risks: The sequencer itself becomes a powerful actor. If it is centralized (as most are today), it can censor transactions or extract value from users. Protocol-enforced “fair ordering” rules (like first-come-first-served) mitigate this, but not all L2s implement them.
- Slippage from sequencer fees: Unlike L1 where you pay gas based on network demand, L2 sequencers often charge a fixed fee or a premium for priority. You may face higher realized slippage if the sequencer places your trade unfavorably within a batch.
- Cross-domain risks: When moving funds between L2s or to L1, the bridging step is an MEV vector. Liquidity providers on bridges can sandwich your transfer, especially if the bridge uses an AMM model.
Overall, L2s reduce the most visible MEV (sandwiches) but centralize the power. As L2s decentralize their sequencers, the MEV landscape may become more like L1—but with better user protection built into the protocol design (e.g., threshold encrypted mempools).
The Role of Sequencers and Decentralization
Sequencers are the lynchpin of L2 MEV. Today, most rollups operate a single, team-run sequencer. This centralization is a double-edged sword: it enables faster, cheaper transactions and predictable ordering, but it also grants the sequencer enormous power. The sequencer can:
- Extract MEV by placing its own trades before user transactions.
- Sell ordering rights to third parties via “MEV auctions.”
- Censor specific addresses or transactions.
The solution being explored by many rollups is decentralized sequencing—spreading the ordering power among multiple nodes using consensus mechanisms like round-robin, leader election, or threshold-based ordering. Protocols like Espresso, shared sequencing networks (e.g., Astria), and EigenLayer’s restaking are building infrastructure for this.
For searchers, a decentralized sequencer set reintroduces the competition for block space that exists on L1, potentially leading to more MEV opportunities (and more complex extraction). For users, decentralized sequencing reduces censorship risk and levels the playing field, but may reintroduce some of the frontrunning they sought to escape. The eventual equilibrium will likely be similar to L1 PBS, but tailored for faster block times and lower costs.
Cross-Rollup MEV: The Next Frontier
As the number of L2s grows, MEV that spans multiple rollups becomes increasingly lucrative. This is not just arbitrage between two DEXs; it includes:
- Bridge exploits: If a bridge pauses or a liquidity pool becomes unbalanced, MEV bots can drain it across chains.
- State-dependent arbitrage: An event on one L2 (e.g., a large liquidation) may create price discrepancies on another L2. Bots that can monitor multiple chains and execute trades quickly capture this.
- Sequencer latency arbitrage: If Rollup A’s sequencer includes a transaction faster than Rollup B’s, a bot can trade on B before the information arrives.
This domain is still nascent because cross-rollup communication is slow and costly (bridging takes minutes to hours). However, with the rise of fast bridges (e.g., Across, Stargate) and shared settlement layers (e.g., zkSync’s Hyperchain, Polygon CDK), cross-rollup MEV will grow. Searchers who build multi-chain infrastructure now will have a first-mover advantage.
“Cross-rollup MEV is today where L1 MEV was in 2020—underexplored, underappreciated, and extremely profitable for the few who understand the mechanics.” — pseudonymous researcher ‘beefy’
Frequently asked questions
Is MEV on Layer 2 the same as on Ethereum L1?
No. On L2, a single sequencer typically orders transactions privately, eliminating the public mempool that enables most L1 MEV. Sandwich attacks become rare, but the sequencer itself can extract value—creating a different set of risks and opportunities.
Can users avoid MEV on L2?
Partially. Using a rollup with fair ordering (e.g., first-come-first-served) or a private mempool reduces the chance of being frontrun. However, the sequencer still controls ultimate ordering, and cross-bridge movements can be exploited. Users can also use MEV-protected RPCs (like Flashbots Protect) on some L2s.
What is cross-rollup MEV and why is it unique?
Cross-rollup MEV involves extracting value from price or state differences between two or more L2s (or between an L2 and L1). It is unique because it requires fast, reliable bridging between chains, and the latency differences between sequencers create temporal arbitrage windows not present within a single chain.
Related reading
Track the entities behind the concepts
DeFi Intel maps 11,000+ protocols, tokens and companies to a typed knowledge graph — with live data, incidents and regulation.