A sequencer is the component of a rollup that decides what order incoming transactions execute in, builds blocks from those transactions, and posts the resulting batches and state roots to the rollup's settlement layer. On every Ethereum layer-2 in production today, that work is performed by a single piece of software run by a single team. The sequencer is the part of a rollup users actually interact with second to second.
​

This article covers what a sequencer does on every block, why all production L2s still run a single sequencer in 2026, the failure modes that design exposes, and the decentralization paths now being attempted.
​

A sequencer is the rollup component that receives transactions from users, orders them, executes them against the rollup's state, packs the results into a batch, and submits that batch to L1. It is both the rollup's transaction mempool and its block producer. Without a sequencer, a rollup has no live ordering and produces no new state.
​

The sequencer sits between two surfaces. On one side, users and dapps send transactions to a sequencer RPC endpoint. On the other, the sequencer publishes ordered batches and proof data back to an L1 like Ethereum. The Arbitrum documentation describes this dual role: the sequencer is the only party that can produce a soft confirmation of a transaction, and the only party that controls inclusion until that batch is posted to L1.
​

This is different from a validator on a monolithic L1. An Ethereum validator participates in a permissionless rotation. A rollup sequencer is a singleton service: across Arbitrum One, Base, OP Mainnet, zkSync Era, Linea, and Scroll, that singleton is run by the rollup team itself. For more on the broader L2 sequencer landscape, see our deeper piece on Ethereum L2 sequencers.
​

On every block the sequencer performs four jobs in sequence: accept incoming transactions, choose an order, execute them against current state to produce new state, and batch the result for submission to L1. The first three jobs happen in milliseconds. The L1 submission step happens on a slower cadence measured in seconds or minutes.
​

Step one is the mempool. The sequencer exposes an RPC endpoint that wallets and dapps connect to. Transactions arrive, get basic validity checks against fees and nonces, and queue up.
​

Step two is ordering. The default policy on most rollups is first-come-first-served by arrival at the sequencer endpoint, with priority-fee bumping in some implementations. Arbitrum's Timeboost auction, live since April 2025, is one alternative: searchers bid for the right to place their transactions at the front of an upcoming block.
​

Step three is execution and soft confirmation. The sequencer runs the ordered transactions through its EVM and produces a new state root. Users typically receive a soft confirmation within roughly 250 milliseconds, well before the batch has touched L1.
​

Step four is L1 submission. The sequencer batches a window of transactions, compresses them, and posts the compressed data plus a state commitment to L1. On Ethereum, post-Dencun, this happens through EIP-4844 blobs. Base typically posts every few minutes; Arbitrum every minute or less under load. Once the batch reaches L1 and the rollup's settlement contract accepts it, the transactions are no longer reversible by the sequencer.
​

Every production Ethereum L2 in 2026 runs a single sequencer operated by one entity. Arbitrum, Base, OP Mainnet, zkSync Era, Linea, Scroll, Blast, Mantle, and Mode each rely on one operator. The reason is operational, not ideological: a single sequencer is faster and cheaper to run, and it was good enough to ship the rollup category in the first place.
​

A centralized sequencer can soft-confirm in under a second because no consensus round is required. It also captures sequencing revenue (the gap between user-paid L2 fees and L1 data costs) without splitting it across a validator set. Public Dune Analytics dashboards have shown that revenue running into the tens of millions of dollars annually for the largest rollups since 2024.
​

The other factor is correctness pressure. Sequencer code has not been battle-tested across many independent operators. Most rollup teams have publicly stated they want to decentralize once the surface area stabilizes; as of 2026 production designs are still rolling out. Arbitrum's decentralized Timeboost specification, published jointly with Espresso Systems, is one example of that work.
​

A centralized sequencer concentrates three failure modes in one piece of software: downtime, censorship, and MEV extraction. None are hypothetical. Each has been observed on a production rollup at least once between 2022 and 2026. The mechanism in each case stems from a single operator holding both the ordering decision and the soft-confirmation key.
​

Downtime. Since the Nitro upgrade in August 2022, Arbitrum has seen multiple sequencer outages typically ranging from about 30 minutes to a couple of hours, including a roughly 78-minute pause during the December 2023 Inscriptions traffic surge documented by CoinDesk. Base and other L2s have experienced similar incidents. User funds remain safe during an outage because the L1 state is unchanged, but no new transactions can be soft-confirmed.
​

Censorship. A centralized sequencer can refuse to include a specific address or transaction. Most rollups counter this with a force-inclusion escape hatch on L1: a user submits a transaction directly to the rollup's L1 inbox contract, and the sequencer must include it within a fixed window (around 24 hours on Arbitrum, around 12 hours on Optimism). Force inclusion exists; it is also slow and unfamiliar to most users.
​

MEV extraction. Because the sequencer controls ordering, it can insert, reorder, or sandwich user transactions for profit. Most major rollups have committed publicly not to extract MEV directly, but the capability sits in the same process that signs batches. Designs like Timeboost auction MEV opportunities to external searchers instead. Encrypted-mempool approaches like Radius's PVDE delay encryption aim to make ordering blind to transaction contents altogether.
​

Decentralized sequencing replaces the single-operator design with a set of operators that produce blocks under a consensus protocol. Shared sequencing extends this: one operator set sequences transactions for multiple rollups at once, enabling atomic cross-rollup composition. Based rollups push further still and delegate sequencing to Ethereum L1 validators directly.
​

Espresso. The Espresso Network moved from devnet to a production proof-of-stake network with its Mainnet 1.0 launch in late 2025, and by early 2026 was publishing confirmations for integrated chains including ApeChain, Celo, and others across the Arbitrum and Polygon ecosystems. Espresso pairs the HotShot consensus protocol with a data availability layer called Tiramisu.
​

Astria. Astria launched a Celestia-based shared sequencer mainnet in October 2024, raised roughly $18 million, and recorded its last block in December 2025 before the team wound the network down. The Block covered the shutdown, with limited rollup adoption cited as a factor. The Astria timeline is the clearest data point on the difficulty of bootstrapping a shared sequencing market.
​

Radius. Radius focuses on MEV mitigation through encrypted mempools. The sequencer collects and orders verified transactions without decrypting them, using Practical Verifiable Delay Encryption to conceal contents until ordering is final. In 2026 Radius partnered with Avail to enable synchronous atomic transactions across rollups via a permissionless searcher auction called Lighthouse.
​

Based rollups. A based rollup delegates sequencing to the L1 itself. Ethereum validators, already proposing L1 blocks in a permissionless rotation, also propose the rollup's blocks; there is no separate sequencer operator. Taiko Alethia is the most prominent based rollup in production. MEV flows to L1 validators, which strengthens Ethereum's economic security. The trade-off is latency: based rollups inherit Ethereum's 12-second slot time for hard ordering, though preconfirmation designs aim to reduce perceived latency.
​

The three live sequencer designs in 2026 make different trade-offs across operator count, MEV ownership, latency, and cross-rollup composition. The table below summarizes how each shapes the same dimensions, reflecting production state rather than roadmap claims.
​

None of these designs is strictly dominant. Centralized sequencing remains the production default because it is fast and simple. Shared sequencing offers cross-rollup atomicity at a latency cost. Based rollups maximize alignment with Ethereum at the cost of slot-bound finality.
​

For a dapp or wallet developer, sequencer decentralization changes four numbers: soft confirmation latency rises by hundreds of milliseconds to seconds, censorship resistance improves because no single operator can refuse a user, MEV exposure shifts toward explicit auctions or encrypted mempools, and cross-rollup atomic composition becomes possible only when two rollups share the same sequencer set.
​

The bundler setup in ERC-4337 account abstraction is a useful analogy. A bundler aggregates user operations off-chain, orders them, and submits them on-chain, much as a sequencer orders transactions for a rollup. Decentralizing 4337 bundlers and decentralizing rollup sequencers are structurally similar: an off-chain ordering role that needs liveness, censorship resistance, and MEV controls. A 2024 arXiv survey catalogs the mitigation approaches that apply to both.
​

For cross-chain routing, moving value between Arbitrum and Base today requires bridging through L1 or a fast-bridge protocol because the two rollups have independent sequencers and no shared block. Two rollups on the same shared sequencer can execute paired transactions in the same block, opening atomic cross-rollup swaps and intent settlement. Eco Routes treats each L2's sequencer as a black box and uses partner rails like Circle's CCTP and Hyperlane for cross-chain settlement. As shared sequencing matures across rollup families, those settlement paths shorten.
​

Ethereum L2 sequencers deep dive · What is ERC-4337 account abstraction · What is a blockchain wallet · What is a smart wallet · What is tokenization · What is a wrapped token · What is a DEX aggregator · Stablecoin pillar · Bridging and wrapped tokens · Best DEX aggregators
​

Sources and methodology. Sequencer mechanics drawn from Arbitrum and Optimism documentation; outage history from CoinDesk and Dedaub root-cause analysis; shared sequencer status from Espresso Systems and The Block's Astria coverage; based rollup specs from Taiko Labs; sequencer revenue figures from Dune Analytics dashboards. Figures refresh as the decentralized-sequencing landscape evolves.
​