Research Brief — 2026-05-21
Abstract: This brief examines how Maximal Extractable Value (MEV) manifests on Layer 2 rollup architectures, focusing on sequencer-extracted MEV, shared sequencer models, and protocol-level mitigation strategies including Flashbots SUAVE, Arbitrum Timeboost, and OP Stack sequencer revenue designs.
Maximal Extractable Value (MEV) — the profit block producers can extract by reordering, including, or excluding transactions — has evolved significantly from its Ethereum L1 origins. On Layer 2 rollups, the MEV landscape differs fundamentally because L2s use sequencers rather than validators to order transactions. This architectural distinction creates both new MEV opportunities and new mitigation possibilities unavailable on L1.
As of Q2 2026, the total MEV extracted across major L2s is estimated at $380M+ annually (source: EigenPhi MEV dashboard, 2026), with approximately 65% flowing to sequencers and 35% to searchers and builders. Understanding these dynamics is critical for L2 designers, DeFi protocols, and validators.
Unlike Ethereum L1 where MEV is captured by validators through PBS (Proposer-Builder Separation), L2 sequencers hold a monopoly on transaction ordering within their rollup. This gives them unique extraction capabilities:
- Front-running: The sequencer observes all pending transactions in its mempool and can insert its own transactions before user transactions.
- Censorship for profit: A sequencer can delay or exclude transactions that would reduce its own profits (e.g., liquidation transactions that compete with the sequencer's own positions).
- Back-running: Inserting transactions after high-value user transactions (e.g., sandwich attacks).
- Bundle insertion: Creating and inserting custom transaction bundles that extract maximum value.
Most L2s in production (Arbitrum, Optimism, Base, Blast) use a single, centralized sequencer. While this provides fast user confirmations (sub-second), it creates a central point for MEV extraction:
"The centralized sequencer model gives the sequencer operator full discretion over transaction ordering — effectively a license to print MEV." — Flashbots Research, 2025
Empirical data from Dune Analytics (2025-2026) shows that centralized sequencers on major L2s generate $50-120M annually in MEV-related revenue, most of which is retained by the sequencer operator rather than returned to users or dApps.
Several projects are working on decentralized sequencer designs:
| Approach | Project | Mechanism | MEV Mitigation |
|---|---|---|---|
| Threshold consensus | Espresso | HotShot consensus with fair ordering | Threshold encryption |
| Shared sequencer | Astria | Shared sequencing layer | Commit-reveal ordering |
| Rotating leader | Radius | zk-proof of ordering | Sequential consistency |
| Auction-based | Skip (formerly Skip Protocol) | MEV-aware sequencing | Builder separation |
Shared sequencers provide a single ordering service across multiple rollups, creating a unified sequencing layer that can enable atomic cross-rollup composability while reducing MEV extraction.
Astria implements a shared sequencer that:
- Accepts transactions from multiple rollups
- Orders them using a BFT consensus
- Submits the ordered batch to each rollup's settlement layer
Key design decisions for MEV mitigation:
- Commit-reveal: Users submit encrypted transactions that are only revealed after ordering is finalized, preventing front-running.
- No transaction reordering: Once a batch is committed, the sequencer cannot reorder individual transactions.
- Cross-rollup atomicity: Enables atomic MEV arbitrage across rollups without trust assumptions.
Shared sequencers face several challenges:
- Trust assumptions: Users must trust the shared sequencer set not to collude.
- Latency: Decentralized ordering introduces 2-5 second latency compared to centralized sequencer sub-second confirmations.
- Economic security: The shared sequencer's economic security must exceed the value it sequences (similar to L1 security budget considerations).
- MEV redistribution: Who gets the MEV? Fair distribution among participating rollups remains unsolved.
Flashbots SUAVE (Single Unified Auction for Value Expression) represents a paradigm shift in MEV handling. SUAVE is a specialized rollup designed to be the mempool and execution environment for MEV across all blockchains, including L2s.
On L2s, SUAVE provides:
- Confidential compute: MEV transactions are executed inside SUAVE's TEE (Trusted Execution Environment), preventing front-running and sandwich attacks at the L2 level.
- Cross-domain MEV: A searcher can submit a single bundle that extracts MEV across multiple L2s simultaneously — something impossible with siloed sequencing.
- Pre-confirmations: SUAVE provides pre-confirmation guarantees for bundles, improving UX for MEV-aware protocols.
SUAVE introduces key smart contract primitives:
MevBundle: A bundle of transactions with execution constraints across domains.ConfidentialStore: Encrypted storage for private order flow.PermissionedExecutor: Selective execution rights for different MEV roles.
As of May 2026, SUAVE is in testnet with several L2s conducting integration trials. The Flashbots team estimates mainnet readiness for Q3 2026, with initial support for Arbitrum, Optimism, and Base.
Arbitrum Timeboost, proposed in Arbitrum Improvement Proposal (AIP) in early 2026, introduces a time-based ordering auction for the Arbitrum sequencer's transaction ordering rights.
- Express Lane Auction: Users bid for the right to have their transactions included in a specific time slot (250ms window).
- Time-based Priority: Transactions within a time slot are ordered by bid amount (highest bidder gets earliest position).
- Revenue Redistribution: Auction proceeds are redistributed to ARB stakers.
- Optional Express Lane: Users can bypass the auction by accepting a 1-block delay (free lane).
- Sequencer revenue: Timeboost converts the sequencer's MEV extraction monopoly into a transparent auction market.
- Searcher competition: Searchers compete on bid price rather than latency, reducing the advantage of geographically close searchers.
- User protection: Users who don't want to participate in the MEV game can use the free lane, accepting delayed inclusion but avoiding extraction.
- Revenue distribution: Auction proceeds directly benefit ARB token holders, aligning incentives.
Critics argue that Timeboost institutionalizes MEV rather than eliminating it. Key concerns:
- Bid transparency: All bids are public, enabling sophisticated MEV bots to calculate exact extraction profit and bid accordingly.
- Centralization risk: Only well-capitalized searchers can compete in the express lane, potentially centralizing MEV extraction.
- Implementation complexity: The 250ms time window requires precise clock synchronization, which is technically challenging in a decentralized context.
The OP Stack (Optimism's modular rollup framework) provides a configurable sequencer model where each L2 instance can choose its own sequencing strategy. This design philosophy differs from Arbitrum's unified approach.
OP Stack sequencers generate revenue from multiple sources:
- Sequencing fees: Users pay a small fee per transaction for ordering.
- MEV tips: MEV-aware users can include tips for preferred ordering.
- Priority fees: Similar to Ethereum EIP-1559-style priority fees.
- Data availability fees: Fees paid for posting transaction data to L1 (Ethereum) or alternative DA layers (EigenDA, Celestia).
The OP Stack includes a SequencerFeeVault contract that captures a portion of sequencer revenue. Key parameters:
- Sequencer share: Configurable percentage (typical: 80-100%) retained by the sequencer operator.
- L1 data fee share: Pass-through to L1 for data availability costs (typically 10-20% of total revenue).
- Protocol fee share: Optional fee sent to the Optimism Collective treasury (configurable, typically 0-5%).
The OP Stack's Superchain vision includes:
- Shared ordering: A future Superchain sequencer could order transactions across all OP Stack chains atomically.
- Cross-chain MEV bundles: Users could submit bundles that include transactions on multiple OP Stack chains.
- Sequencer neutrality: Open-source sequencer implementation enables community audits and alternative client implementations.
Based on on-chain data from 2025-2026 (sources: Optimism Dune Dashboard, L2Beat):
| L2 | Sequencer Revenue (Annual) | MEV Share | Fee Distribution |
|---|---|---|---|
| Optimism | $45-65M | 55-70% | 80% sequencer / 20% L1 fees |
| Base | $35-50M | 50-65% | 85% sequencer / 15% L1 fees |
| OP Stack testnets | <$1M | N/A | 100% to testnet operators |
| Approach | Front-running Protection | Censorship Resistance | Cross-chain MEV | Implementation Complexity |
|---|---|---|---|---|
| Centralized sequencer | None | None | Limited | Low |
| Timeboost | Partial (bidding) | Partial | None | Medium |
| Shared sequencer (Astria) | Good (encryption) | Good | Excellent | High |
| SUAVE | Excellent (TEE) | Good | Excellent | Very High |
| OP Stack configurable | Configurable | Configurable | Future | Low-Medium |
The fundamental trade-off across all approaches is latency vs. decentralization:
- Centralized sequencers offer 100-500ms confirmations but maximal MEV extraction.
- Shared sequencers offer 2-5s confirmations with strong MEV protection.
- SUAVE offers sub-second pre-confirmations but requires TEE hardware trust.
Several trends are shaping the L2 MEV landscape in 2026:
- MEV redistribution: Growing pressure from L2 communities for sequencer MEV to be redistributed to token holders or ecosystem funds.
- Cross-domain arbitrage: Increasing MEV opportunities as L2 liquidity fragments across 40+ rollups.
- Regulatory attention: Regulatory bodies (SEC, ESMA) beginning to examine sequencer MEV practices.
- Standardization efforts: The L2 MEV working group (formed 2025) working on standardized MEV reporting.
MEV on L2 rollups represents both a challenge and an opportunity. The centralized sequencer model that powers today's leading L2s creates significant MEV extraction potential — but also provides a clean architectural point for intervention that doesn't exist on permissionless L1s.
Shared sequencers, Flashbots SUAVE, and mechanisms like Arbitrum Timeboost each offer different trade-offs between latency, decentralization, and MEV resistance. The OP Stack's modular approach allows L2 operators to choose their preferred configuration.
The medium-term outlook (2026-2027) suggests a convergence toward hybrid models: centralized sequencers for fast pre-confirmations, with fallback to decentralized ordering for censorship resistance and MEV protection. SUAVE's confidential compute layer may emerge as the dominant cross-domain MEV execution environment.
- Ethereum Foundation Research. "MEV on Layer 2: A Taxonomy." ethresear.ch, 2025.
- Flashbots. "SUAVE: An MEV-Centric Rollup." flashbots.net, 2025-2026.
- Arbitrum Foundation. "AIP: Timeboost — Time-Based Sequencing Auction." forum.arbitrum.foundation, 2026.
- Optimism Collective. "OP Stack Sequencer Specification." github.com/ethereum-optimism/optimism, 2026.
- Astria. "Shared Sequencing: Architecture and Design." docs.astria.org, 2025-2026.
- EigenPhi. "MEV Dashboard: L2 Extractable Value." eigenphi.io, 2026.
- Dune Analytics. "Optimism and Arbitrum Sequencer Revenue." dune.com, 2025-2026.
- L2Beat. "Risk Analysis: Sequencer Centralization." l2beat.com, 2026.
- Espresso Systems. "HotShot Consensus for Decentralized Sequencing." docs.espressosys.com, 2025.
- Radius. "zk-proof of Sequential Consistency." radius.xyz, 2025-2026.