Ethereum's rollup-centric roadmap, the scaling strategy that defined protocol development since 2020, has been formally abandoned by its architect. Vitalik Buterin declared on February 3, 2026, that the original vision for Layer 2s "no longer makes sense," citing two structural failures: L2 decen...
"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, the scaling strategy that defined protocol development since 2020, has been formally abandoned by its architect. Vitalik Buterin declared on February 3, 2026, that the original vision for Layer 2s "no longer makes sense," citing two structural failures: L2 decentralization progressed "far slower" than anticipated, and Ethereum L1 scaled faster than anyone projected. Average mainnet gas prices fell 93% year-over-year to 0.43–0.50 gwei in January 2026, eliminating the primary economic rationale for rollup migration.
The fallout has been swift. Coinbase's Base departed the OP Stack on February 18, 2026, triggering a 28% collapse in the OP token within 48 hours. OP Labs cut 20% of its workforce on March 12. ENS scrapped its planned L2 rollup (Namechain) and returned to mainnet deployment. Polygon announced the sunsetting of its $250 million zkEVM initiative. Of more than 50 tracked L2s, only two have reached Stage 2 decentralization. Three networks — Base, Arbitrum, and Optimism — process approximately 90% of all L2 transactions, while dozens of smaller rollups operate as "zombie chains," technically functional but economically irrelevant.
The replacement strategy: native rollups embedded directly into Ethereum, verified by L1 validators, eliminating the multisig-bridge trust model that Buterin now considers incompatible with Ethereum's security guarantees.
The rollup-centric roadmap was first articulated by Buterin in October 2020. The premise: Ethereum L1 would serve as a settlement and data availability layer, while execution would migrate to rollups that inherited mainnet security through fraud proofs or validity proofs. This model assumed two conditions would hold — that L2s would rapidly decentralize toward trustless operation, and that L1 would remain capacity-constrained, keeping fees elevated.
Neither condition held.
On February 3, 2026, Buterin published a post stating: "Both of these facts, for their own separate reasons, mean that the original vision of L2s and their role in Ethereum no longer makes sense, and we need a new path." He followed up two days later by blasting "copypasta" L2 chains that replicate standard EVM execution without meaningful differentiation.
The timeline of retreat:
Ethereum's base layer fee economics have transformed. The data:
| Metric | Peak (May 2021) | January 2025 | January 2026 | Change (YoY) | |--------|-----------------|--------------|--------------|---------------| | Average gas price (gwei) | 150+ | 7.14 | 0.43–0.50 | -93% | | Average tx fee (USD) | $53.16 | ~$5.00 | $0.44 | -91% | | Gas limit | 30M | 36M | 60M | +67% |
The Glamsterdam upgrade (scheduled H1 2026) will increase the gas limit from 60 million to 200 million — a 233% increase — while introducing parallel processing. EIP-4844 (proto-danksharding), implemented in March 2024, already reduced blob data costs to "near-zero marginal costs" for rollups. L2 average fees now range from $0.001 to $0.01 per transaction.
This creates a paradox: the infrastructure that was built to escape high L1 fees now operates in an environment where L1 fees are sustainably low. The economic moat that justified L2 existence has eroded.
At his April 20 Hong Kong keynote, Buterin confirmed a long-term target of 10,000+ TPS on L1 through ZK proof verification by validators, with finality achieved in 10–20 seconds across 1–3 block slots.
The rollup-centric thesis assumed L2s would progress through decentralization stages rapidly. L2Beat's stage framework defines maturity:
Current status as of April 2026:
| Network | Stage | Sequencer | Notes | |---------|-------|-----------|-------| | Arbitrum One | 1 | Centralized | Fraud proofs live | | OP Mainnet | 1 | Centralized | Fault proofs operational | | Base | 1 | Centralized | Permissionless fault proofs | | Taiko | 1 | L1-sequenced (based rollup) | Exception to centralized model | | Metis | 1 | Decentralized PoS pool | Exception since March 2024 | | All others (~45+) | 0 | Centralized | No permissionless proofs |
Only two of more than 50 major L2s have reached Stage 2. Nearly all major networks still operate single-operator centralized sequencers. Censorship resistance and neutrality depend entirely on the integrity of a single entity — precisely the trust model that Ethereum was designed to eliminate.
The critical distinction Buterin drew: a multisig-bridged L2 does not inherit Ethereum's security guarantees. Users of such systems have weaker protections than those transacting directly on L1.
On February 18, 2026, Coinbase announced Base would migrate from Optimism's OP Stack to a proprietary unified architecture maintained under a single base/base GitHub repository. The move was described as pursuing "independent development."
The financial impact on Optimism was immediate:
Base's economics justified the departure. In 2025, the network generated $82.6 million in annual revenue ($185,291 daily average) with approximately $55 million in net profit after L1 data costs. Under the prior arrangement, Base shared 2.5% of sequencer revenue or 15% of net profits with the Optimism Collective. Base was the only L2 achieving standalone profitability in 2025.
OP Labs CEO Jing Wang maintained on March 12 that the layoffs were "not related to finances," claiming the company has "enough funds and resources for several years ahead." The stated rationale: fewer tasks, faster execution, reduced coordination overhead.
The Superchain model — 30 L2s built on OP Stack sharing infrastructure and interoperability — now faces an existential question without its largest revenue contributor.
21Shares published a forecast stating most Ethereum L2s are unlikely to survive past 2026. The data supports this thesis:
Market concentration:
TVL distribution:
Casualties:
The surviving categories, per 21Shares:
The proposed replacement for the multisig-bridge model: native rollups that integrate verification logic directly into Ethereum's consensus protocol. Ethereum researchers presented a proof-of-concept in March 2026.
How native rollups differ:
| Property | Current L2 Model | Native Rollups | |----------|-----------------|----------------| | Verification | Fraud/validity proofs submitted by L2 operators | L1 validators re-execute L2 blocks | | Bridge trust | Multisig or security council | Ethereum consensus | | Security council | Required (can override) | Not needed | | Composability | Async (bridge delays) | Synchronous with L1 | | Security inheritance | Partial (depends on stage) | Full L1 security |
Technical prerequisites cited by Buterin at the Hong Kong keynote:
The January 19 tweet clarifying his position: ZK-EVM technology has improved sufficiently, and Ethereum's path to supporting ZK proofs at the base layer is "now realistic."
The rollup thesis created a specific value distribution: sequencer operators extracted fees, L2 tokens captured governance premiums, and Ethereum L1 received data availability payments. The new model redistributes this value.
Under the rollup-centric model:
Under the native rollup model:
This explains the market reaction: OP fell 97.5% from highs; L2-specific tokens face structural devaluation if their security assumptions are rendered obsolete by native rollup integration.
Solana's comparative revenue provides context: $1.03 million in daily chain fees versus ~$182,000 across all Ethereum L2s combined. The commoditization of L2 execution has collapsed fee revenue to a fraction of competing monolithic chains.
The rollup-centric thesis served its purpose during Ethereum's capacity-constrained period (2021–2024). With L1 fees at $0.44 and the gas limit tripling in the near term, the economic forcing function that drove users off-chain has dissipated. Simultaneously, the trust assumptions underlying most L2s — centralized sequencers, multisig bridges, security councils with override authority — proved incompatible with Ethereum's core value proposition of trustless execution.
The transition will not be instantaneous. Native rollups remain at the proof-of-concept stage. Existing L2s with established user bases (Base's $4.3B TVL, Arbitrum's $16.6B TVS) will persist as execution environments, likely specializing by application domain rather than competing on generic EVM throughput. The "zombie chains" — the long tail of undifferentiated rollups — face extinction.
What has changed permanently: the default assumption that Ethereum scales through external execution layers. The protocol is pivoting toward a model where "large quantities of block space backed by the full faith and credit of Ethereum" are provided natively. For the economic value distribution of the ecosystem, this concentrates fee revenue at the validator layer and structurally devalues standalone L2 governance tokens.
The 2026 L2 shakeout is not a market cycle phenomenon. It is an architectural reclassification.