Layer 2 rollups are fast because they don't process every transaction on Ethereum. A user sends a transaction, gets confirmation in under a second, and pays a fraction of a cent in gas. What makes this possible is a single actor sitting between the user and Ethereum — the sequencer.
That speed comes with a tradeoff that gets less attention than it deserves.
The Role of the Sequencer
When you submit a transaction on an L2 rollup — Arbitrum, Optimism, Base, zkSync, Scroll — it doesn't go directly to Ethereum. It goes to the sequencer.
The sequencer's job is to receive user transactions, order them, issue a soft confirmation (near-instant acknowledgment that the transaction will be included), batch transactions together periodically, and post the compressed batch to Ethereum L1.
Once the batch lands on Ethereum, the ordering is permanent and inherits L1 security. Before that, users are trusting the sequencer's soft confirmation — which in practice is almost always honored, but isn't backed by the same cryptographic finality as an on-chain confirmation.
The gap between "sequencer accepted your transaction" and "Ethereum confirmed the batch" is where most of the interesting risk lives.
Ordering Control and MEV
The sequencer doesn't just receive transactions — it decides the order.
On Ethereum mainnet, transaction ordering is largely determined by gas price. Validators choose which transactions to include based on fees, creating the familiar priority fee market. On an L2, that control belongs to the sequencer. Within a batch, the sequencer determines which transactions execute first.
This matters because order determines outcomes in DeFi. If a large swap is pending, whoever trades before it gets a better price; whoever trades after it doesn't. The entity controlling order can exploit this — inserting its own transactions ahead of users, reordering to capture arbitrage, or frontrunning pending operations.
Whether specific sequencers do this varies by architecture and policy. Arbitrum's sequencer uses a first-in-first-out rule, which constrains this significantly. Others don't make the same guarantee. The important point is structural: the sequencer role carries MEV authority by design, and how that authority gets exercised depends on who's running it.
The Centralization Problem
Every major rollup currently runs a centralized sequencer operated by a single company: Arbitrum by Offchain Labs, Optimism by OP Labs, Base by Coinbase, zkSync by Matter Labs, Scroll by the Scroll Foundation.
A single operator controlling the sequencer creates two risks.
Liveness. If the sequencer goes offline, users can't get soft confirmations and normal transaction flow halts. The mitigation is a "force inclusion" mechanism — users can submit transactions directly to the L1 smart contract, bypassing the sequencer entirely. Most major rollups have this. The downside is the delay: typically hours rather than seconds.
Censorship. A centralized sequencer can refuse to include specific transactions — from specific addresses, for specific protocols, or in response to regulatory pressure. This is different from an Ethereum validator refusing a transaction; here, a single company makes the call. Force inclusion is the escape hatch, but again, it's slow.
The rollup builders are aware of this. The typical framing is that centralization is a temporary design choice — pragmatic during early development — and that decentralized sequencing is on the roadmap. Whether that roadmap materializes is genuinely uncertain.
What's Changing
Decentralized sequencing is being built, but it isn't live at meaningful scale on the major rollups yet.
Shared sequencer networks — Espresso Systems, Astria, and others — are building infrastructure that multiple rollups could plug into, replacing the single-operator model with a validator committee. Besides decentralization, shared sequencers enable a capability that's otherwise very hard: cross-rollup atomic transactions, where a single operation touches two different L2s simultaneously.
Based rollups use a different approach. Instead of a separate sequencer, they let Ethereum's own validators sequence L2 transactions. This removes the sequencer as an independent role and inherits L1 decentralization natively. The tradeoff is speed — Ethereum block times are roughly 12 seconds, which constrains soft confirmation latency. Taiko is the main implementation.
The honest state as of mid-2026: production rollups use centralized sequencers. Decentralized alternatives are in testnet and early deployment. The major L2 teams have committed to decentralizing their own sequencers, but without specific timelines.
What Would Confirm Progress
A major production rollup — Arbitrum, Base, OP mainnet — transitioning to a decentralized sequencer committee and operating without incident for a sustained period. A shared sequencer network handling meaningful transaction volume across multiple rollups. The force inclusion delay dropping to minutes rather than hours.
What Would Complicate or Invalidate the Thesis
Regulatory action requiring rollup operators to maintain centralized control for compliance purposes would constrain decentralization regardless of technical progress. Shared sequencer networks shipping but with committee sets small enough to be functionally centralized — technically distributed, actually concentrated — would fulfill the letter of decentralization without the substance. If the major rollup teams simply don't prioritize sequencer decentralization, the roadmap commitment stays a statement.
The "decentralization is coming" framing is worth treating as a hypothesis rather than a given.
Timing
Now: Centralized sequencers are the production reality. Soft confirmations are fast and reliable in practice. Force inclusion exists as a backstop. MEV extraction potential is real, constrained by operator policy. For most users, the day-to-day experience is fine — the risk is structural, not operational.
Next: First production decentralized sequencers from shared networks are expected in the 2025-2026 window. Based rollup activity is growing.
Later: Major rollup sequencer decentralization is a stated goal without a hard timeline. Treat as a multi-year question.
Boundary
This covers the sequencer's role in rollup architecture — ordering, batching, and soft confirmation. It doesn't address the separate role of provers in ZK rollups (proving validity is a different function from sequencing), the economics of running a sequencer, or the mechanics of any specific rollup's force inclusion contract. Whether centralized sequencing constitutes an acceptable risk in any context is outside scope.
