L2 narrative ending? From user abandonment to Vitalik backtracking, Ethereum scaling roadmap reaches a turning point

In February 2026, Ethereum co-founder Vitalik Buterin publicly stated in a lengthy article that the roadmap established five years ago, which viewed Layer 2 as the primary means of Ethereum’s scalability, “has failed.” This statement is not an isolated technical discussion but is corroborated by a structural reversal in on-chain data.

According to statistics from Token Terminal, the monthly active addresses on Ethereum Layer 2 networks plummeted from about 58.4 million in mid-2025 to about 30 million in February 2026, a drop of nearly 50%. Meanwhile, the active addresses on the Ethereum mainnet surged from about 7 million to 15 million, achieving a doubling in growth. Users are flowing back from L2 to L1—this trend is completely contrary to the mainstream expectation over the past three years that “L2 would carry the vast majority of Ethereum’s transactions.”

The reversal in user ratios occurred against the backdrop of mainnet Gas fees dropping to historic lows. The PeerDAS (peer data availability sampling) technology introduced by the Fusaka upgrade, along with the adjustment of Blob target capacity from 6 to 14 (with a maximum of 21), significantly increased the transaction processing capacity of the Ethereum mainnet compared to the early stages post-merge. As the Gas limit on L1 has been raised to 60 million units, with plans to further increase to 100 million or even 200 million units, the necessity of L2 as a “cheap and fast” scaling tool has begun to be fundamentally questioned.

What are the driving mechanisms for users flowing back to the mainnet?

The superficial reason for the change in user behavior is the convergence of Gas fees, but the deeper driving mechanisms involve three dimensions: technology, economics, and security.

On the technical level, the Ethereum mainnet has significantly reduced L1 transaction costs through incremental scalability upgrades. After the Dencun upgrade, the data availability costs paid by L2 to Ethereum dropped by over 90%. However, this upgrade, which was originally intended to benefit L2, has instead weakened L2’s competitive advantage—when the Gas fees on L1 are low enough to be nearly indistinguishable from those of L2, users no longer need to migrate to L2 to save money.

From an economic perspective, the value capture mechanism of L2 tokens has exposed fundamental flaws. In 2025, the total revenue of the entire L2 sector plummeted by 53% year-on-year, dropping to about $129 million, with most revenue belonging to centralized sequencer operators, leaving token holders with almost nothing. The core use of mainstream L2 tokens like ARB and OP is limited to governance voting, lacking both staking rewards and burn mechanisms, earning them the label of “worthless governance assets” in the market. When L2 tokens cannot capture the consensus premium of network operations, the motivation for users to hold these tokens diminishes.

The security aspect is even more critical. Vitalik pointed to the core issue: an EVM chain with a processing capacity of 10,000 TPS, if its connection to L1 is only regulated by a multisig bridge, has not truly scaled Ethereum but merely established a trust-based independent platform. According to L2beat statistics, of the top 20 Rollup projects, only one has reached Stage 2 (fully trustless), while 12 projects are in Stage 0, heavily relying on multisignatures and auxiliary functions. When users realize that the safety of their funds ultimately depends on a few private key holders rather than the mathematical guarantees of Ethereum, flowing back to L1 becomes a rational choice.

What costs does this structural transformation bring?

The downgrade of L2 from “official Ethereum sharding” to “specialized plugins” has resulted in asymmetric cost distributions across the ecosystem.

For L2 project teams, the cost is most direct. Mainstream L2 tokens have plummeted over 90% from historical highs, with the total market cap of the sector shrinking to about $7.95 billion. Leading projects like Arbitrum and Optimism saw their tokens drop by 15% to 30% in January 2026. More importantly, the funding window is closing—previously, the valuations in the L2 space were built on the narrative of “inheriting Ethereum’s security,” and when this narrative was denied by Vitalik himself, the valuation logic in the primary market faces reconstruction.

For the Ethereum ecosystem, the cost is reflected in fragmented liquidity and dispersed developer attention. Over the past five years, L2 has carved the Ethereum ecosystem into dozens of isolated islands, with users incurring bridging risks and tolls when crossing between different L2s. L2 project teams tend to establish their own token economies and ecosystems rather than feeding back to L1. This situation of “feudal fragmentation” has gradually reduced the Ethereum mainnet to a purely settlement layer, diluting ecological synergy.

For users, the cost is an increase in cognitive burden. Ordinary users find it difficult to distinguish between the security stages of different L2s and cannot assess which L2 truly inherits Ethereum’s security. Vitalik’s concept of the “trust spectrum”—ranging from Stage 0’s centralized multisignatures to Stage 2’s fully trustless—although helpful in clarifying differences, requires users to possess a high level of technical literacy to make safe choices.

What does this mean for the landscape of the cryptocurrency industry?

This structural adjustment is reshaping the power dynamics of the Ethereum ecosystem and spilling over into the broader cryptocurrency industry.

First, the Ethereum mainnet has re-established its central position for value capture. Over the past three years, L2 tokens have siphoned off capital and attention that could have flowed into ETH. Now, with the enhanced scalability of L1 and the failure of L2 token value capture mechanisms, funds are beginning to reassess the safety and scarcity of ETH as a foundational asset. Vitalik’s proposal for “native Rollup precompiles”—allowing Ethereum to directly verify ZK-EVM proofs—further strengthens the position of L1 as the final verification layer.

Second, the L2 sector is undergoing a brutal survival of the fittest. A report released by 21Shares at the end of 2025 indicated that among over 50 L2s, Base, Arbitrum, and Optimism accounted for nearly 90% of the transaction volume, with Base alone exceeding 60% market share. The activity of small Rollups has decreased by 61%, with some projects like Kinto ceasing operations and Blast’s TVL plunging by 97%. The industry is transitioning from a “hundred flowers blooming” to a “winner takes all” consolidation phase.

Third, the competitive landscape is polarizing towards monolithic chains and L1 native scalability. Some funds are flowing towards high-performance monolithic chains like Solana, which has demonstrated the potential for millions of transactions per second through the Firedancer client in testing. Another portion of funds is returning to the Ethereum mainnet, seeking certainty in security premiums. This polarization means that L2 projects must make strategic choices between “deep integration with Ethereum” and “fully independent development.”

How might the future evolve?

Vitalik has not completely denied the value of L2’s existence but has instead indicated a new direction for L2’s evolution—from “scaling tools” to “specialized plugins.”

The first evolutionary path is to move towards the higher end of the “trust spectrum.” Vitalik explicitly requires that L2s managing Ethereum assets reach at least Stage 1 security standards, meaning that smart contracts begin to have limited governance rights rather than relying entirely on multisig. For L2 projects pursuing long-term development, evolving to Stage 2 (fully trustless) becomes key to establishing competitive barriers. Native Rollup precompiles are seen as critical infrastructure that allows Ethereum to directly verify proofs and keep pace with protocol upgrades.

The second evolutionary path is to deeply specialize in vertical fields. Vitalik suggests that L2s explore new value directions “beyond scalability,” including privacy virtual machines, application-specific optimizations, and dedicated architectures for non-financial scenarios (social, identity, artificial intelligence). For example, on-chain identity and payment infrastructure designed specifically for AI agents (such as the x402 protocol and ERC-8004) are forming a technological closed loop. Such applications do not need to compete with L1 for general computing power but rather provide unique features that L1 finds difficult to achieve.

The third evolutionary path is to form a complementary rather than substitutive relationship with L1. Some industry observers propose that the future of L2 lies not in competing with L1 for transaction execution rights but in providing liquidity entry points and user reach channels for L1. When L1 handles core asset settlement and high-value transactions, L2 can focus on high-frequency, low-value, delay-sensitive application scenarios, forming a clearly defined symbiotic relationship.

What potential risks and limitations exist?

This transformation process is accompanied by multiple risks that could affect the stable evolution of the Ethereum ecosystem.

Technical risks are concentrated in the large-scale verification of ZK proofs. Although native Rollup precompiles can theoretically solve the security alignment issues between L2 and L1, the upgrade cycle is complex and has not been verified on a large scale. The transition of Ethereum’s underlying verification mechanisms to zero-knowledge proofs will take years, at least until 2027, during which time execution risks and market uncertainties will be present. There remains disagreement within the developer community over the optimal architectural path, and technical friction is inevitable.

Economic risks manifest as challenges to the sustainability of L2 business models. After the Dencun and Fusaka upgrades, the gas price profits of L2 have been significantly compressed, and the total revenue of the industry plummeted by 53% year-on-year in 2025. If L2s cannot establish new revenue sources (such as MEV distribution after decentralizing sequencers or application layer service fees), many projects may exit the market due to an inability to cover operational costs. 21Shares predicts that most L2s will not survive through 2026, and this assessment is becoming a reality.

Governance risks involve the power struggle between the Ethereum Foundation and L2 interest groups. Some L2 project teams have explicitly stated that due to regulatory requirements, they may never wish to surpass Stage 1 security standards, as this requires relinquishing final control over the network. This creates tension with Ethereum’s core ideals of being permissionless and trustless. If L2 project teams choose to retain centralized control, the Ethereum ecosystem will face the long-term issue of “independent kingdoms masquerading under the Ethereum banner.”

Conclusion

The sharp decline in Ethereum L2 user ratios is not a temporary market fluctuation but a clear signal of structural transformation. When the mainnet Gas fees drop to historical lows, L2’s security progress lags behind expectations, and the value capture mechanisms of tokens fail, L2’s original mission as a “scaling tool” is coming to an end.

Vitalik’s roadmap revision is essentially a denial of the model that “a valuation of tens of billions can be sustained solely on an expansion narrative” over the past five years. Ethereum is transitioning from “Rollup-centric” to a new architecture centered on “L1 scalability, complemented by L2 specialization.” In this architecture, the survival of L2 no longer relies on the brand narrative of “inheriting Ethereum’s security” but must prove its existence through unique values such as privacy protection, application-specific optimization, and AI agent infrastructure.

For industry participants, this signifies a fundamental shift in evaluation standards—no longer asking “how high is this L2’s TPS,” but rather “what unique functionalities does this L2 provide that L1 cannot achieve?” The era of narratives is coming to an end, and the era of productivity is beginning.

FAQ

Q1: What is the core reason for the sharp decline in Ethereum L2 user ratios?

The core reasons for users flowing back from L2 to L1 include three aspects: first, the Ethereum mainnet significantly enhanced processing capabilities through the Fusaka upgrade, bringing Gas fees down to levels similar to L2; second, most L2s remain at Stage 0 or Stage 1 security phases, relying on centralized sequencers and multisig bridges, failing to truly inherit Ethereum’s security; third, L2 tokens lack effective value capture mechanisms and cannot provide staking rewards or burn mechanisms to benefit holders.

Q2: What is Vitalik’s latest positioning on L2?

In February 2026, Vitalik publicly denied the original roadmap centered on “Rollup,” believing that L2 should no longer be seen as Ethereum’s “brand sharding.” He proposed viewing L2 as a “trust spectrum,” ranging from Stage 0 (centralized multisig) to Stage 2 (fully trustless). L2 must prove its necessity by providing unique values that L1 cannot achieve (such as privacy protection, application-specific optimization, and non-financial scenarios).

Q3: Is there still survival space for L2 projects?

Yes, but the survival space will concentrate in two directions: first, evolving towards the higher end of the trust spectrum to reach Stage 1 or even Stage 2 security standards, deeply binding with Ethereum; second, transforming towards vertical specialization, focusing on privacy virtual machines, AI agent infrastructure, game-specific chains, and other scenarios that L1 cannot efficiently support. Generic L2s relying solely on a “cheap and fast” narrative will face the greatest survival pressure.

Q4: What are the follow-up plans for the Ethereum mainnet’s scalability?

The Ethereum mainnet plans to further increase Blob target capacity to 48 by June 2026, with Gas limits projected to rise to 100 million or even 200 million units. The Glamsterdam upgrade will focus on reducing MEV-related manipulation, stabilizing Gas fee rates, and laying the groundwork for future scalability. The long-term goal is for L1 to independently handle a large volume of transactions while achieving deep interoperability with L2 through native Rollup precompiles.

Q5: What does this mean for ordinary users?

Ordinary users can choose the more secure Ethereum mainnet for asset operations when transaction costs are similar. For scenarios requiring L2, users need to pay attention to the security stages (Stage 0/1/2) and trust assumptions of L2, prioritizing projects that reach Stage 1 and above and align with Ethereum’s security, avoiding storing assets on centralized L2s that rely on multisig bridges.