Every rollup running today has a sequencer - a single entity that decides transaction ordering before anything hits the base layer. For most of the rollup era, that sequencer has been controlled by the same team that built the rollup. Arbitrum’s sequencer is run by Offchain Labs. Optimism’s by OP Labs. It works, but the centralization is obvious and the failure mode is real: if the sequencer goes down or censors transactions, users are stuck.
Shared sequencing attacks this problem from an angle that’s different from just decentralizing each rollup’s own sequencer. The idea is to have a single, neutral sequencer - or a decentralized sequencer network - that accepts transactions from multiple rollups simultaneously, orders them, and produces a commitment that all participating rollups agree to honor.
Why Ordering Is the Actual Problem
Most discussions of rollup centralization focus on fraud proofs or withdrawal delays. Those matter, but ordering power is more immediate. Whoever controls ordering controls MEV extraction, can front-run users, and can impose soft censorship without any on-chain evidence. A shared sequencer doesn’t eliminate MEV, but it changes who captures it and introduces competitive pressure that a monopoly sequencer never faces.
Espresso Systems has been the most visible project building in this space, developing a sequencer network that rollups can plug into rather than operating their own. Astria takes a related approach, treating sequencing as a modular layer that rollup stacks can adopt independently of their execution environment.
The Atomicity Angle
The less-discussed benefit is cross-rollup atomicity. If two rollups share the same sequencer, a transaction bundle involving both chains can be ordered atomically - either both legs execute or neither does. That’s something today’s bridge-heavy ecosystem fundamentally cannot guarantee. A user bridging from one rollup to another currently faces a gap where one leg can succeed and the other fail.
This matters most for DeFi protocols that want to operate across multiple execution environments without exposing users to partial-fill risk.
Still Early, Still Unproven
No shared sequencer has reached significant production adoption as of late 2026. The coordination overhead is real - rollup teams have to agree on a common sequencing interface, and the liveness assumptions of the shared layer become a new dependency for every rollup that opts in.
That said, the trend toward modular blockchain design makes shared sequencing more plausible than it would have seemed two years ago. Rollup teams are already accustomed to outsourcing data availability to Celestia or EigenDA. Outsourcing sequencing is the same logic applied one layer closer to execution.