Over the past 14 months, the total value locked across Ethereum Layer2s has surged past $45 billion. Yet during that same period, the number of sequencers running the most popular rollups has barely budged. Arbitrum One, the largest optimistic rollup by TVL, operates with a single sequencer. Optimism’s OP Mainnet? Also one. Base, Coinbase’s incubated chain, runs a single sequencer controlled by the company. When I audit these networks for my educational platform, the same question emerges from students: "If it’s decentralized, why does one node control the entire transaction flow?"
That question sits at the heart of a tension that the industry has papered over with white papers and roadmaps. For two years, teams have promised "decentralized sequencing" as the next milestone. For two years, those promises have remained slides in investor decks. The gap between rhetoric and reality is not a bug — it is a feature of a market that prizes speed to TVL over architectural integrity.
Context: The Layer2 Promise vs. The Sequencer Reality
Layer2 rollups exist to scale Ethereum by moving execution off-chain while posting compressed transaction data to the mainnet. The sequencer is the node that orders transactions, constructs batches, and submits them to L1. In a fully decentralized system, anyone should be able to propose and order transactions, preventing censorship and single-point failure. But today, nearly every major rollup uses a single, permissioned sequencer operated by the core team or a trusted entity.
Arbitrum’s "AnyTrust" model assumes a committee of validators, but the sequencer itself remains centralized. Optimism’s "Bedrock" upgrade improved data availability but did not touch sequencer decentralization. zkSync Era uses a centralized sequencer with a plan to decentralize "in phases." Scroll, Linea, StarkNet — the story repeats. The only exception is Metis, which launched a decentralized sequencer in late 2023, but its TVL remains a fraction of the leaders.
I remember sitting in a developer workshop in 2022 where a prominent rollup engineer said, "Sequencer decentralization is a UX problem first." That framing has become the industry’s default excuse. But after auditing the actual node requirements for these networks, I see a deeper issue: economic incentives, not technology, are the bottleneck.
Core: The Technical Anatomy of a Centralized Sequencer
To understand why decentralization is stalled, we have to look at the sequencer’s revenue model. A sequencer earns two types of fees: priority fees from users who pay extra for faster inclusion, and a portion of the L1 data posting cost (often passed to users as "L2 gas"). When a single sequencer operates, it captures 100% of these fees. In a decentralized model, those fees must be shared among a set of sequencers, diluting the incentive for the core team to give up control.
Let me use real numbers. Arbitrum processed roughly 1.5 million transactions per day in Q1 2026. At an average priority fee of 0.001 ETH per transaction, that’s 1,500 ETH daily — over $3 million at current prices. The sequencer operator, Offchain Labs, pockets that revenue. If they decentralized to 10 sequencers, each would earn $300,000 per day, still lucrative, but the core team loses the dominant share. More importantly, they lose the ability to extract MEV (maximal extractable value) by reordering transactions — a practice that, while controversial, generates substantial income.
The MEV layer is the hidden reason for centralization. In a single-sequencer model, the operator can see the entire mempool and order transactions to capture arbitrage and liquidations. Decentralized sequencer proposals like "sequencer auctions" or "shared sequencer sets" attempt to distribute this value, but they introduce latency and complexity. Based on my experience auditing three rollup codebases, I can say that the technical challenge is not the consensus algorithm — it’s the economic design that prevents rent extraction.
Consider the "fair ordering" problem. If multiple sequencers propose blocks simultaneously, how do you ensure fair transaction ordering without a leader? Solutions like "threshold decryption" or "timelock encryption" exist in research papers but have never been deployed at scale on a mainnet rollup. The only production-tested approach is Metis’s "decentralized sequencer pool," which uses a Proof-of-Stake committee to rotate sequencing rights. Yet Metis handles only 50,000 transactions per day — 3% of Arbitrum’s volume. Scaling that model to millions of daily transactions while maintaining low latency is an unsolved problem.
Contrarian: Maybe Centralized Sequencing Is the Pragmatic Choice
I’ve spent years arguing that decentralization is non-negotiable for trustless systems. But the data forces me to confront an uncomfortable truth: the market has voted with its feet. Users and developers have flocked to chains with centralized sequencers because they offer predictable fees, fast finality, and no front-running (at least from external actors). Base, despite being fully controlled by Coinbase, has grown to $8 billion TVL in two years. The user base does not seem to care about sequencer centralization — they care about cost and speed.
This is the contrarian angle that the Ethereum maximalist echo chamber rarely acknowledges. For most DeFi applications, a centralized sequencer that is accountable to a corporate entity (like Coinbase) may be more trustworthy than a poorly designed decentralized sequencer that can be gamed by MEV bots. The real risk is not centralization per se, but the lack of transparency and exit options. If a sequencer censors transactions, users should be able to force-include them via L1 — a feature that both Arbitrum and Optimism support through "forced transaction" mechanisms. In practice, this escape hatch has never been used at scale, but its existence provides a credible threat.
Yet I remain uneasy. The forced transaction mechanism requires users to submit directly to L1, which costs 10–50x more than L2 fees. For retail users, that’s not a realistic option. Community is not a user base; it is a shared soul. When a single entity controls the sequencer, that entity holds the power to decide which transactions live and die. In a bear market or a regulatory crackdown, that power can be weaponized.
Takeaway: The Road Ahead
The industry is at a crossroads. Either Layer2 teams deliver on their decentralization roadmaps within the next 12 months, or the narrative of "Ethereum scaling" will become synonymous with "corporate-run sidechains." The irony is that the technology is almost ready — shared sequencer networks like Espresso and Radius are in testnet, and new cryptographic primitives like "atomic composability across sequencers" are being researched. But the incentives are misaligned.
We build not for the token, but for the tribe. If the tribe — the developers and users who truly value sovereignty — demands decentralized sequencers, they must start voting with their wallets. Stop depositing into chains that promise decentralization but deliver a single point of failure. Until then, the PowerPoints will keep cycling, and the sequencers will remain centralized. The question is whether we are building for the next bull run or for the next decade.