We didn't realize we were trading one gatekeeper for another until the Optimism governance vote in March 2025. A proposal to upgrade the sequencer on OP Mainnet passed with 98% approval, but the actual code change was pushed by a single infrastructure provider. No on-chain vote halted the upgrade. No L1 reorg was triggered. The sequencer just... updated. That moment crystallized something I had been feeling for months but couldn't articulate: the Layer2 we celebrated as "Ethereum scaled" is still running on a permissioned train track.
Let me rewind. I first heard the term "sequencer" in mid-2021 while auditing the Arbitrum One launch. Back then, the narrative was simple: rollups need a temporary centralized sequencer to order transactions cheaply, and eventually we'll decentralize it. The word "temporary" did heavy lifting. Three years later, nearly every major rollup—Arbitrum, Optimism, Base, zkSync Era, Scroll—still relies on a single sequencer operated by the core team or a third-party operator. The decentralization roadmaps published in 2022 promised shared sequencing networks by 2023. We're now in 2025, and not a single production rollup has live, trustless, decentralized sequencing. The industry has effectively normalized a two-tier security model: Ethereum's L1 provides immutable settlement, but the sequencer controls which transactions get included and in what order.
The technical architecture of a Layer2 today looks more like a corporate server than a trustless network. The sequencer is a single node that receives transactions, orders them, batches them, and submits the batch to L1. It can censor transactions, reorder them for front-running (MEV extraction), or simply go offline. Users have no alternative route to get their transactions included unless the rollup implements a forced inclusion mechanism—like an escape hatch. Most do, but the delays and gas costs are punitive. For example, on Arbitrum One, a forced transaction via L1 requires a 7-day delay and costs roughly 10x the normal L2 fee. That's not a bug; it's a feature designed to discourage use. The sequencer's power is absolute unless users are willing to pay a steep price for sovereignty.
I've seen this pattern before. In 2020, I reverse-engineered the exploit of a yield farming protocol that had a single point of failure in its admin key. The team claimed it was decentralized, but the admin key controlled the entire contract. The sequencer is the same: a single point of control dressed in the language of "temporary centralization." The difference is that the community has bought into a timeline that keeps extending. Every quarter, the decentralization update is "coming soon" or "in testnet." We accepted this because it enabled cheap, fast transactions. We forgot that trustlessness is not a feature toggle—it's a property of the system's architecture.
To understand why decentralized sequencing is stalled, you need to look at the economic incentives. Running a sequencer is profitable. It captures MEV from transaction ordering, often extracting value that would otherwise go to L1 validators or users. On Arbitrum, the sequencer has generated an estimated $120 million in MEV revenue since launch. That revenue is currently captured by Offchain Labs, the core development team. In a decentralized sequencing network, that revenue would be distributed among multiple operators, diluting the core team's share. The rational economic choice is to delay decentralization as long as possible. Meanwhile, projects like Espresso, Astria, and Radius have built shared sequencing layers that promise to solve this, but adoption has been slow because existing rollups have no incentive to integrate.
Here's the contrarian angle: maybe permanent centralized sequencing isn't a bug—it's a feature of the market's demand for efficiency. Users overwhelmingly choose rollups with cheap, fast UX over those with decentralized sequencing. No rollup has lost significant market share because its sequencer was centralized. In fact, the top three rollups by TVL—Arbitrum, Optimism, and Base—all have centralized sequencers. The market is voting with its feet, and it's voting for speed over sovereignty. This challenges the core ideological premise that decentralization is a necessary end state. Perhaps, for most users, the sequencer's centralization is an acceptable trade-off for a working product. We didn't want to admit this because it undermines the foundational narrative of "blockchain as censorship-resistant infrastructure."
But accepting permanent centralized sequencing comes with real risks. A centralized sequencer can collude with a government to freeze accounts, blacklist addresses, or reorder transactions for friendly parties. In regimes where financial censorship is routine, the sequencer becomes a choke point. The workaround for a user in such a jurisdiction is to use the L1 forced inclusion—but that defeats the purpose of being on L2. The promised "universal access" is conditional on the sequencer's permission. This is the contradiction we are living with: we built a system that scales Ethereum's decentralization into a walled garden.
Truth in blockchain isn't found in white papers or roadmap slide decks. It's found in the production code and the incentives that govern decisions. The truth today is that Layer2 sequencers are centralized by design, and the industry lacks both the economic motivation and user demand to change that. The next bull run will likely paper over this issue with optimistic roadmaps and hype around "decentralized sequencing testnets." But I've learned not to confuse a testnet with a solution. The real question isn't when sequencing will be decentralized. It's whether we are willing to sacrifice the efficiency we love for the sovereignty we claim to value. If we aren't, then we should stop calling Layer2 rollups "Ethereum's scaling solution" and start calling them what they are: fast, cheap, but permissioned networks bolted onto a chain that isn't.