Dencun went live 18 months ago. TPS across Ethereum L2s surged past 500. Gas fees on Arbitrum hit $0.001. Yet here's the uncomfortable truth I keep circling back to: every major L2 still runs a single sequencer. We didn't get decentralized sequencing. We got cheaper centralized sequencing.
I've been tracking this since 2023, when I first mapped out sequencer revenue models for my graduate thesis. Back then, the narrative was clear: rollups scale Ethereum, but the sequencer is a single point of failure. By 2024, the narrative shifted to "blob economics." By 2025, everyone was drunk on L2 TVL numbers. Now in mid-2026, with total L2 TVL north of $120B, the structural risk is larger than ever. Alpha isn't in which L2 has the coolest UX; it's in understanding that the narrative of decentralization has been quietly replaced by the narrative of efficiency.
Let me give you the hard data. I pulled on-chain metrics for the top 10 rollups by TVL over the past 90 days. Only one — Fuel — has a permissionless sequencer in production. The rest: Arbitrum, Optimism, Base, zkSync, StarkNet, Linea, Scroll, Blast, and Mantle — all operate a single sequencer run by their respective teams or a designated entity. The justification is always the same: it's a temporary measure. But temporary has become permanent. History doesn't repeat, but the incentives do — keep the sequencer centralized to capture MEV, maintain upgrade control, and keep the token narrative alive.
The core insight here is hidden in the collective belief system. Dencun solved data availability by introducing blobs. It slashed L2 costs by 90%. But it did nothing to the sequencer's monopoly. Why? Because decentralization of the sequencer is a governance and economic problem, not a protocol one. The L2 teams have no incentive to give up that revenue stream. I estimated that Arbitrum's sequencer captured roughly $240M in MEV and ordering fees in 2025. That's a line item no team wants to see zeroed out by a shared sequencing layer.
Now, the contrarian angle: maybe centralized sequencing is actually the optimal model for the next cycle. This is heresy in Ethereum maximalist circles, but hear me out. LUNA didn't collapse because it had a single sequencer; it collapsed because the algorithmic stablecoin had no real backing. Centralization of the sequencer doesn't automatically mean insecurity. It means a larger attack surface, but it also means faster upgrades, better UX, and lower latency. The tradeoff between security and speed is real. For mass adoption, you might actually want centralized sequencing with strong fraud proofs or validium-like exits. The real risk isn't the sequencer being evil; it's the sequencer being compromised by a state actor or a massive exploit. We saw that with the Ronin bridge — a centrally controlled validator set got hacked for $600M. The same principle applies.
But here's what most analysts miss: the narrative shift from “we need decentralized sequencing” to “centralized sequencing is fine” is itself a market signal. When the community stops demanding a core decentralization feature, it means the market is pricing in a different risk premium. In this case, the risk premium on L2 security is dropping relative to the risk premium on throughput. This is typical in late-cycle bullish narratives: everyone focuses on what's improving (speed, fees) and ignores the static risk (centralized control).
Based on my experience surviving the 2022 LUNA collapse, I recognize this pattern. We ignored the mechanism design flaws because the narrative of “algorithmic dollar” was too seductive. Today, we ignored the sequencer problem because the narrative of “blob-based scaling” is too seductive. The numbers tell the story: since Dencun, total L2 revenue from sequencer fees increased 4x, while the number of sequencer nodes per network remained at 1. That's a warning sign.
We need to look at the regulatory angle. MiCA, which came into full effect in 2025, didn't explicitly address sequencer centralization. But its CASP (Crypto Asset Service Provider) classification requirements indirectly penalize any protocol that doesn't have clear governance accountability. A centralized sequencer is actually more compliant under MiCA because there's a legal entity to sue if something goes wrong. That's a perverse incentive: regulation is pushing L2s further into centralization, not away from it. Small projects can't afford the legal overhead of running a decentralized sequencer network with anonymous nodes, so they stick with a single sequencer under a registered entity. This kills innovation exactly where it's needed most — in novel sequencing mechanisms like shared sequencing or based rollups.
Now, let me connect this to my recent work. In early 2026, I led a project to design a compliant tokenization framework for RWA in Southeast Asia. We had to choose between Optimism's OP Stack and Arbitrum's Nitro. Both had centralized sequencers. The regulators in Singapore were okay with it because they could demand downtime or freeze assets via the sequencer operator. That's a feature, not a bug, for them. But it means the “permissionless” promise of Ethereum is being abandoned in favor of regulatory convenience. The ETF inflow we saw in 2024 wasn't bullish for decentralization — it was bullish for institutional-grade centralization. Bitcoin ETFs demand custodians. Ethereum L2s demand sequencer operators. Same pattern.
What does this mean for the next 12 months? I see three possible paths. First, a major L2 gets exploited via its sequencer — either a bug in the mempool or a compromised key. That will trigger a massive de-rating of all L2 tokens and a rush toward truly decentralized alternatives (like Fuel or perhaps a yet-unlaunched zk-rollup with native sequencing). Second, the Ethereum foundation finally forces the issue by mandating sequencer decentralization as part of the next upgrade (unlikely, given political dynamics). Third, nothing changes — the market continues to value speed over security, and the centralized sequencer becomes the de facto standard. I'm betting on the third, but positioning for the first.
Takeaway: The next narrative shift in L2s will not be about higher throughput. It will be about who controls the sequencer. Watch for governance votes on shared sequencers; that's where the real alpha is.
I realize this sounds bearish. It's not meant to be. It's meant to be honest. The crypto industry constantly hides structural risks behind shiny new features. We did it with LUNA. We did it with FTX. We did it with Celsius. We're doing it now with L2s. The narratives change, but the underlying incentives don't. Centralization is efficient. Decentralization is resilient. The market is currently pricing efficiency over resilience. That's a bet that works until it doesn't. And when it doesn't, only those who understand the narrative mechanics will survive.
