Ethereum's rollup-centric roadmap was supposed to deliver both scalability and decentralization. Three years in, the scaling arrived — but the decentralization didn't. As of March 2026, no major general-purpose Layer 2 has reached Stage 2 on L2BEAT's maturity framework. The three dominant rollups...
"If you create a 10,000 TPS EVM where its connection to L1 is mediated by a multisig bridge, then you are not scaling Ethereum." — Vitalik Buterin, Ethereum Co-Founder
Ethereum's rollup-centric roadmap was supposed to deliver both scalability and decentralization. Three years in, the scaling arrived — but the decentralization didn't. As of March 2026, no major general-purpose Layer 2 has reached Stage 2 on L2BEAT's maturity framework. The three dominant rollups — Base, Arbitrum, and Optimism — process approximately 90% of all L2 transactions while operating centralized sequencers controlled by single corporate entities. Together they secure over $38 billion in user assets behind trust assumptions that look uncomfortably similar to the traditional financial intermediaries blockchain was designed to replace.
The contradiction is now impossible to ignore. In February 2026, Vitalik Buterin himself declared the rollup-centric roadmap "no longer makes sense," arguing that L2s have decentralized "far slower" than expected while Ethereum's base layer has advanced rapidly. Meanwhile, a 21Shares report warns that most smaller L2s will not survive 2026, as Base, Arbitrum, and Optimism tighten their oligopoly. What's emerging isn't decentralized infrastructure — it's a three-player toll booth extracting rent from Ethereum's security guarantees while delivering centralized block production.
This report examines the economic structure of L2 sequencer centralization, the revenue dynamics driving consolidation, and the emerging technical responses — including native rollups — that could reshape the landscape.
L2BEAT's Stages framework provides the most rigorous assessment of rollup maturity. Stage 0 means users trust operators entirely. Stage 1 means smart contract-enforced security with training wheels. Stage 2 means fully decentralized — no trusted parties, permissionless proving, and users can always exit without sequencer cooperation. As of March 2026, the landscape looks like this:
| Network | Stage | Sequencer | Proving System | |---------|-------|-----------|----------------| | Arbitrum One | Stage 1 | Single operator (Offchain Labs) | Fraud proofs (permissioned) | | Base | Stage 1 | Single operator (Coinbase) | Fault proofs via OP Stack | | OP Mainnet | Stage 1 | Single operator (Optimism Foundation) | Fault proofs | | Starknet | Stage 0 | 3 operator-controlled nodes | Validity proofs (centralized prover) | | ZKsync Era | Stage 0 | Single operator (Matter Labs) | Validity proofs (centralized prover) | | Linea | Stage 0 | Single operator (Consensys) | Validity proofs (centralized prover) |
No major rollup has reached Stage 2. The implications are significant: centralized sequencers control transaction ordering, can theoretically censor transactions, extract MEV, and represent single points of failure. The "decentralization roadmaps" published by each team remain aspirational documents with no binding timelines.
Starknet's Grinta upgrade (v0.14.0, September 2025) introduced three sequencers running Tendermint consensus, reducing block times from 30 seconds to 4 seconds. The team marketed this as "decentralized sequencing." In practice, all three sequencers remain operator-controlled — this is redundancy engineering, not decentralization. The distinction matters enormously when $593 million in user assets depends on it.
The economic incentive to remain centralized is the story few want to tell. Centralized sequencers capture 100% of transaction ordering revenue, including priority fees and MEV opportunities. There is no protocol-level mechanism forcing L2 operators to share this value with validators, users, or Ethereum L1.
The numbers tell the story:
The Dencun upgrade (March 2024) reduced L1 data posting costs by approximately 90% through blob transactions. This was supposed to benefit users through lower fees. Instead, it primarily expanded sequencer profit margins. L2 operators reduced user-facing fees modestly but captured the majority of the cost savings as profit. This dynamic mirrors the classic toll booth problem: infrastructure operators extract rent at chokepoints rather than passing efficiency gains to users.
The total addressable sequencer revenue across all Ethereum L2s is estimated at approximately $161 million annually. If this revenue were redirected to L1 validators through decentralized sequencing, it would meaningfully boost ETH staking yield. The fact that it isn't reveals where the real economic incentives lie.
21Shares published a sobering assessment in early 2026: most Ethereum L2s will not survive the year. The data supports this conclusion. Base alone processes over 60% of all L2 transactions. The top three networks — Base, Arbitrum, and Optimism — capture roughly 90%. Everyone else is fighting over a shrinking 10% of activity.
Smaller rollups are becoming "zombie chains" — technically operational but with usage dropping by 61% or more. The Dencun fee reduction, intended to democratize the L2 space, instead triggered fee wars that pushed most rollups into operating losses. Only Base, subsidized by Coinbase's distribution engine and exchange integration, managed to turn a profit.
21Shares expects the surviving L2 landscape to consolidate around three categories:
The pattern is clear: L2s are becoming extensions of centralized exchanges rather than independent decentralized networks. The irony — that crypto's scaling layer is being won by the same corporate entities it was meant to disintermediate — deserves serious examination.
Base is the most commercially successful L2 and the most instructive case study in centralization economics. Launched by Coinbase in August 2023, Base operates a single sequencer controlled entirely by Coinbase. Every transaction on Base — ordering, inclusion, timing — flows through Coinbase's infrastructure.
The financial results are striking. Base generated $55 million in net profit in 2025, with application-layer revenue of $369.9 million (Aerodrome alone contributed $160.5 million, or 43% of the total). Coinbase's overall 2025 revenue reached $7.18 billion, with Base becoming a material contributor to the company's infrastructure services segment.
Base reached L2BEAT Stage 1 in January 2026, meaning users can technically exit without sequencer cooperation through the fault proof system. But Stage 1 is not decentralization — it's an emergency escape hatch. During normal operations, Coinbase retains full control over block production, transaction ordering, fee pricing, and all strategic decisions.
Coinbase's Chief Legal Officer has argued that the sequencer's role is "limited to ordering transactions" and that matching is handled by decentralized smart contracts. This is technically accurate and strategically misleading. Transaction ordering is the value capture layer. Controlling which transactions appear first, which are delayed, and how blocks are constructed is precisely where MEV extraction and rent-seeking occur.
The Optimism Superchain model — a shared tech stack (OP Stack) with revenue sharing — was meant to create a cooperative L2 ecosystem. Base was its crown jewel, generating the lion's share of Superchain sequencer fees. In January 2026, the Optimism Collective approved a governance proposal to allocate 50% of net Superchain revenue to monthly OP token buybacks, deploying approximately $8 million annually.
Then Base announced it was leaving.
Coinbase's L2 is transitioning from the OP Stack to its own self-managed tech stack called "base/base," ending a three-year dependency. This is the highest-revenue chain in the Superchain ecosystem severing its revenue-sharing relationship with the Optimism treasury. The timing — weeks after the buyback proposal passed with 84.4% approval — created an immediate credibility crisis for the Superchain economic model.
The OP buyback program's effectiveness is now directly threatened. The $8 million annual deployment was calculated using historical revenue that included Base's contributions. Without Base, the Superchain's sequencer revenue drops precipitously, undermining the tokenomic mechanism designed to create sustainable demand for OP.
This fracture illustrates a fundamental problem with cooperative L2 economic models: when the dominant player's incentive to extract value individually exceeds the collective benefit of participation, the cooperative breaks down. Game theory predicted this outcome; the Superchain delivered it.
Ethereum's response to the L2 centralization crisis is native rollups — a proposal (EIP-8079) that would embed L2 verification directly into Ethereum's base layer. The Ethrex client team, collaborating with the Ethereum Foundation and L2BEAT, completed a proof-of-concept in early 2026.
The mechanism is elegant: instead of requiring separate fraud proof or validity proof systems (which introduce their own centralization vectors), native rollups use a new EXECUTE precompile that allows Ethereum to re-execute L2 blocks directly on L1. No ZK circuits, no fraud proof games, no complex proof infrastructure to maintain.
The implications for the sequencer centralization problem are significant:
Native rollups are currently a research milestone, not a deployment decision. But they signal that Ethereum's base layer is being redesigned with the explicit goal of reclaiming the security guarantees that centralized L2 sequencers have eroded. The Ethereum Foundation's 2026 roadmap includes the Glamsterdam and Hegotá upgrades, both of which prioritize L1 scaling and L2 alignment — a direct response to the fragmentation crisis.
No major L2 has reached Stage 2 decentralization. After three years of the rollup-centric roadmap, every major rollup still operates centralized sequencers controlled by single corporate entities.
Sequencer centralization is economically rational. Base's $55 million profit in 2025 demonstrates that controlling the sequencer is the primary value capture mechanism for L2 operators. The incentive to decentralize is negative absent external pressure.
The L2 landscape is consolidating into an oligopoly. Three networks process 90% of transactions. 21Shares projects most smaller L2s won't survive 2026. Exchange-backed chains are winning through distribution, not decentralization.
The Superchain cooperative model is fracturing. Base's departure from the OP Stack undermines the revenue-sharing model and exposes the fragility of cooperative L2 economics.
Native rollups could restructure the incentive landscape. EIP-8079 would eliminate the technical justification for centralized proof systems, but deployment timelines remain uncertain.
$38+ billion in user assets sit behind centralized trust assumptions. This is the core risk: L2s market themselves on Ethereum's security while operating infrastructure that doesn't inherit it.
Ethereum's Layer 2 ecosystem has produced impressive scaling — lower fees, higher throughput, meaningful user adoption. But it has not produced decentralization. The dominant L2s operate as corporate-controlled block producers, extracting sequencer revenue while deferring decentralization to roadmaps with no deadlines. Users holding $38 billion in L2 assets are trusting corporate entities, not cryptographic proofs.
Vitalik Buterin's February 2026 intervention — declaring the rollup-centric roadmap "no longer makes sense" — was a watershed acknowledgment that the L2 experiment has diverged from its original thesis. The native rollups proposal offers a technical path forward, but it arrives against a backdrop of deeply entrenched economic incentives favoring centralization.
The question is no longer whether L2s can decentralize — it's whether they will, given that centralization is where the money is. For institutional allocators, the distinction between "secured by Ethereum" and "built on top of Ethereum but operated by Coinbase" is not semantic — it's a risk model. The market is pricing L2 tokens as if they've delivered on decentralization promises they haven't kept. That gap between narrative and reality is where the risk lives.