Most L2s, however, still opt for centralized sequencers. The reasons are practical. Centralized sequencers offer fast inclusion times, low-latency confirmations, and simplified operations. They also bring financial benefits. For instance, in 2023, Arbitrum's sequencer generated over $20 million in revenue through user fees and MEV, and those funds were directed back into the Arbitrum DAO for grants, development, and ecosystem growth. This revenue model enables rapid funding of ecosystem initiatives and compensates the teams that maintain critical infrastructure. For emerging L2s, this centralization represents a crucial growth lever as it lowers operational complexity while enabling agility in execution and reinvestment. Moreover, centralized control allows for rapid decision-making, which is especially valuable in the early stages of protocol development.

The financial logic of centralized sequencers also appeals to projects pursuing sustainable funding models. Instead of relying on external grants or token emissions, sequencer fees can create a circular economy within the ecosystem. In this model, value flows from users to the protocol and back into developer grants, hackathons, liquidity mining, and long-term research. This feedback loop creates a powerful incentive for projects to maintain control over the sequencer, especially if decentralization introduces new attack surfaces or inefficiencies.

But that doesn’t mean decentralization should be ignored. In fact, there are clear and compelling reasons to decentralize sequencers over time. At the core is the principle of censorship resistance. If one entity controls the sequencer, it can block or delay any transaction, undermining user trust and potentially falling victim to regulatory capture. A decentralized sequencer ensures liveness, where if one participant fails or censors, others continue operating. Incentive alignment is another benefit. By opening participation to many, rewards from MEV and fees can be distributed more broadly, attracting diverse operators and reducing the risk of collusion. This also guards against geographic and political centralization, where infrastructure may be vulnerable to legal pressure or nation-state interference. In the long run, users and developers are likely to favor L2s that are resilient, credibly neutral, and aligned with Ethereum’s core values.

Decentralization also enhances economic fairness and composability. With shared decentralized sequencing, multiple rollups could coordinate transaction ordering, unlocking atomic cross-rollup execution and reducing arbitrage inefficiencies. This could lead to the emergence of unified liquidity layers and interconnected ecosystems, where applications on different rollups interact natively.

Still, decentralization comes with real challenges. Technically, it’s hard. Consensus among sequencers introduces latency. Data availability becomes a shared concern. Spam resistance and DoS prevention require new mechanisms. Economically, incentives need to be redesigned to determine how sequencers are compensated, and how rewards are distributed. Operationally, governance becomes more complex. Who joins the sequencer set? Who coordinates upgrades? These are non-trivial problems. Introducing more parties into the sequencing process also increases coordination costs and risks of failure, particularly during network congestion or attacks.

An important nuance to this debate is highlighted by Paradigm’s Dan Robinson, who argues that decentralization alone does not ensure trustlessness or credibility. The real anchor of user trust is whether the sequencer adheres to clear, transparent, and enforceable rules. Decentralization can serve as a means to reinforce these rules by making them harder to arbitrarily change or subvert operations, but without robust constraints on behavior, even a decentralized sequencer can act against user interests. In this view, decentralization is not a goal in itself, but a tool to preserve rule integrity. Rollups should prioritize creating rule-bound systems with transparent sequencing logic and mechanisms that ensure compliance, regardless of whether the sequencer is centralized or distributed. The credibility of the system, then, is derived not solely from who runs it, but from how reliably and verifiably it operates. For example, a sequencer that is decentralized but lacks enforcement of ordering rules may still engage in harmful MEV extraction or censorship.

Today, most L2s have centralized sequencers, as shown below:

The path to decentralization is not clear, as each rollup is experimenting with different models. What makes decentralizing sequencers so hard is that it’s not just about removing a single operator but about replacing it with an entire distributed system. One that can maintain performance, resist manipulation, and fairly distribute rewards. The trade-offs are steep: latency goes up, coordination becomes harder, and throughput may suffer.

Several structural patterns are emerging in decentralized sequencer proposals. Most start with permissionless or permissioned participation where operators who meet technical and economic requirements can join the sequencer set. To reach agreement on transaction order, these systems often use Byzantine Fault Tolerant (BFT) consensus protocols, such as HotStuff, QBFT, or Narwhal/Bullshark. Many include fallback mechanisms to resist censorship such as cryptographic inclusion lists, time-locked commitments, or fraud-proof-like challenges that ensure user transactions are eventually included. Reward structures vary, but often include transaction fees, token emissions, and MEV redistribution that are sometimes split between proposers and builders, similar to Ethereum’s Proposer-Builder Separation (PBS).

Performance-wise, the impact is real. Multi-party consensus adds latency as transactions that were once confirmed in milliseconds may now take seconds. Throughput may drop due to communication overhead, especially if the network spans diverse geographies. But the trade-off is resilience and neutrality. With engineering effort and optimization through things like batching, off-chain coordination, or optimistic confirmation rules, decentralized sequencers can get close to centralized performance. Systems like Astria and Espresso are exploring such optimizations to minimize the UX trade-offs.

We’re also seeing a diversity of approaches to decentralized sequencing. Some rollups aim to build their own native decentralized sequencers. Starknet, for example, is working toward a Cairo-native decentralized network, with the ultimate goal of proving that a fully decentralized, high-performance ZK-rollup can exist. zkSync’s roadmap includes decentralization through its L3 and zkPorter system, leveraging ZK proofs to enforce transaction correctness without trusting any single party. Other teams are betting on shared infrastructure. Projects like Espresso, Astria, and Radius are building decentralized sequencer networks that multiple rollups can plug into. This has the benefit of inter-rollup composability and shared MEV infrastructure, but it also introduces dependencies and coordination challenges. A failure in the shared sequencer could ripple across multiple rollups.

Then there’s the Ethereum connection. With Proposer-Builder Separation (PBS) and enshrined proposer commitments coming to L1 via future Ethereum upgrades (e.g., EIP-4844 and beyond), there’s an opportunity for L2s to adopt similar separation. PBS could enable L2 builders to handle transaction ordering while proposers handle inclusion. Some projects are exploring adapting MEV-Boost-style infrastructure for L2s, allowing a marketplace for blockspace that reduces MEV extraction by central sequencers. Finally, DA layers like Celestia and EigenDA are blurring the lines where some proposals envision DA providers serving as decentralized sequencers as well, streaming blobs and ordering transactions in parallel. This convergence suggests that the modular future of rollups will involve deep integrations between DA, sequencing, and execution layers.

The implications of decentralizing sequencers go beyond just technical trade-offs. Security improves as reliance on single operators drops. Composability across rollups becomes possible with shared sequencers. MEV extraction becomes more manageable with fair ordering mechanisms. And user trust, which is arguably the most important asset in Ethereum’s ecosystem, deepens when decentralization is real, not just claimed.

Looking forward, the timelines will vary. Espresso plans to launch a shared sequencer in 2025. Optimism’s Superchain strategy includes shared sequencer research. Starknet and zkSync are playing the long game. The big questions are still open: Will L2s converge on a shared sequencer infrastructure, or fragment into silos? How will Ethereum’s L1 roadmap influence L2 design? And, can decentralized sequencers ever match the UX of centralized ones?

The sequencer layer may be the most overlooked point of centralization in rollups. But its decentralization is vital. This is about more than performance, it’s about credibility. Because in the end, whoever controls the sequencer controls the story.

Stay curious, stay building. The best infrastructure is the kind you never notice.

Notable L2 Onchain Proposals (May)

Optimism Governance

Maintenance Upgrade: Absolute Prestate Updates for Isthmus Activation
This maintenance upgrade sets the activation time for the Isthmus hard fork to May 9, 2025, at 16:00:01 UTC, and includes a critical fix for the blob preimages bug.

ZkSync Governance

[GAP-3] Security Council to Convert Recovered ETH into ZK [Executed]
This proposal authorizes the ZKsync Security Council to convert ETH, recovered from the April 2025 exploit of unclaimed airdrop tokens, back into ZK tokens.

[TPP-3] ZIP Audit Reimbursement Program (ZARP) [Executed]
An annual program valued at $5m USD (89,285,714 ZK) to reimburse security audit costs for successfully executed ZKsync Improvement Proposals (ZIPs) in 2025, ensuring high security standards for ZKsync protocol development.

[TPP-4] Deactivate Capped Minters for TPP-1 Ignite Program [Executed]
This proposal removes the minter role from all Ignite Program capped minters.

[ZIP-11] V28 Precompile Upgrade [Executed]
ZIP-11 proposes the V28 upgrade for ZKsync.

[ZIP-10] Activate ZK Gateway as a Settlement Layer [Executed]
This proposal aims to whitelist ZK Gateway chain as a settlement layer for the Elastic Network, paving the way to facilitate fast interop by providing a layer for cheap batch settlement and as well as quick communication between ZK Chains.

Arbitrum Governance

The Watchdog: Arbitrum DAO’s Grant Misuse Bounty Program
This proposal establishes a grant misuse bounty program dubbed “The Watchdog” to incentivize the identification and reporting of misused DAO-allocated funds. The program would utilize an incentive mechanism to reward community contributors and investigators who submit verifiable reports of misappropriation.

[CONSTITUTIONAL] AIP: ArbOS Version 40 Callisto
This AIP proposes to upgrade Arbitrum One and Arbitrum Nova to ArbOS 40 “Callisto.” Callisto adds support for relevant Ethereum Execution Layer (EL) changes from Ethereum’s recent Pectra network upgrade and includes a minor fix related to Arbitrum Stylus

Polygon Governance

[PIP-65] Economic Model for VEBloP Architecture
This proposal defines an economic model designed to function alongside the Validator-Elected Block Producer (VEBloP) architecture proposed in PIP-64. Recognizing that VEBloP alters how transaction fees accrue, this PIP introduces a fee redistribution mechanism.

[PIP-66] Allow early block announcements
This proposal seeks to implement an optimisation to the consensus rules to allow the primary block producers in the network to announce their block early (as soon as it's built) for better block propagation and reducing the chance of reorgs.

Scroll Governance

Scroll DAO Delegate Accelerator Proposal
This proposal requests $87,000 + 100,000 SCR (371,875 SCR total) for a DAO-funded Delegate Accelerator (D/Acc) program that will develop a competent and diverse set of delegates capable of actively governing the Scroll DAO.