On February 3, 2026, Vitalik Buterin said something on X.

The shockwaves this statement sent through the Ethereum community were no less than those caused by his vigorous promotion of the "Rollup-centric" roadmap in 2020. In that post, Vitalik admitted: "The initial vision of Layer2 as 'Branded Sharding' to solve Ethereum's scalability is no longer valid."

One sentence, almost declaring the end of Ethereum's mainstream narrative of the past five years. The Layer2 camp, once seen as the lifeline and savior of Ethereum, is facing its biggest legitimacy crisis since its birth. More direct criticism followed. Vitalik wrote mercilessly in the post: "If you create an EVM that handles 10,000 transactions per second, but its connection to L1 is through a multi-signature bridge, then you are not scaling Ethereum."

Why has the former lifeline become the burden to be discarded today? This is not just a shift in technical路线, but a brutal game of power, profit, and ideals. The story begins five years ago.

How Did Layer2 Become Ethereum's Lifeline?

The answer is simple: it wasn't a technical choice, but a survival strategy. Let's go back to 2021. Ethereum was then mired in the quagmire of being a "noble chain".

The data doesn't lie: On May 10, 2021, Ethereum's average transaction fee reached a historical peak of $53.16. During the craziest period of the NFT boom, Gas prices once soared above 500 gwei. What does this mean? A simple ERC-20 token transfer could cost tens of dollars, and a token swap on Uniswap could cost $150 or even more.

The DeFi Summer of 2020 brought unprecedented prosperity to Ethereum, with the Total Value Locked (TVL) skyrocketing from $700 million at the beginning of the year to $15 billion by the end of the year, an increase of over 2100%. But the price of this prosperity was extreme network congestion. By 2021, when the NFT wave hit, the minting and trading of blue-chip projects like Bored Ape Yacht Club made the network worse. The Gas fee for a single NFT transaction often amounted to hundreds of dollars. A collector was once offered over 1000 ETH for a Bored Ape in 2021 but ultimately gave up due to the high Gas fees and complex transaction process.

At the same time, a challenger named Solana emerged. Its data was staggering: tens of thousands of transactions per second throughput, transaction fees as low as $0.00025. The Solana community not only mocked Ethereum's performance but also directly attacked its bloated and inefficient architecture. The rhetoric that "Ethereum is dead" was rampant, and the community was filled with anxiety.

It was against this backdrop that in October 2020, Vitalik formally proposed a concept in "A Rollup-Centric Ethereum Roadmap": positioning Layer2 as Ethereum's "Branded Sharding". The core of this idea was that Layer2 would process massive transactions off-chain, then package and send the compressed results back to the mainnet, theoretically achieving infinite scaling while inheriting the security and censorship resistance of the Ethereum mainnet.

At that point in time, the entire future of the Ethereum ecosystem was almost entirely bet on the success of Layer2. From the Dencun upgrade in March 2024 introducing EIP-4844 (Proto-Danksharding), specifically providing cheaper data availability space for Layer2, to various core developer meetings, everything was paving the way for Layer2. After the Dencun upgrade, Layer2 data publishing costs dropped by at least 90%, and Arbitrum's transaction fees plummeted from about $0.37 to $0.012. Ethereum tried to gradually push L1 into the background, making it a quiet "settlement layer".

But why didn't this bet pay off?

Those "Centralized Databases" with $1.2 Billion Valuations

If Layer2 had truly realized its initial vision, it wouldn't be falling out of favor today. But the problem is, what exactly did they do wrong?

Vitalik pointed out the fatal flaw in his article: the progress towards decentralization is too slow. The vast majority of Layer2s have not yet reached Stage 2—having a fully decentralized fraud or validity proof system and allowing users to withdraw assets permissionlessly in emergencies. They are still controlled by centralized sequencers that manage the packaging and ordering of transactions. In essence, they are closer to centralized databases disguised as blockchains.

The conflict between commercial reality and technological ideals is exposed here. Take Arbitrum as an example. Its development company, Offchain Labs, received a $120 million investment in its Series B round in 2021, valuing the company at $1.2 billion, with investors including top-tier firms like Lightspeed Venture Partners. But until today, this behemoth with over $15 billion in locked funds, holding about 41% of the Layer2 market share, remains at Stage 1.

Optimism's story is equally intriguing. This project, led by Paradigm and Andreessen Horowitz (a16z), completed a $150 million Series B round in March 2022, with total funding reaching $268.5 million. In April 2024, a16z privately purchased $90 million worth of OP tokens. But even with such strong capital support, Optimism also only reached Stage 1.

The rise of Base reveals another dimension of the problem. As a Layer2 launched by Coinbase, Base quickly became a market darling after its mainnet launch in August 2023. By the end of 2025, Base's TVL reached $4.63 billion, capturing 46% of the entire Layer2 market share, surpassing Arbitrum to become the Layer2 with the highest DeFi TVL. But Base is even less decentralized, as it is entirely controlled by Coinbase, making its technical architecture closer to a centralized sidechain.

Starknet's story is even more ironic. This Layer2 using ZK-Rollup technology, developed by Matter Labs, has raised a total of $458 million, including a $200 million Series C round led by Blockchain Capital and Dragonfly in November 2022. But its token STRK price has shrunk by 98% compared to its historical high, with a market cap of about $283 million. According to on-chain data, the protocol revenue it generates daily is not even enough to cover the operating costs of a few servers, and its core nodes remain highly centralized, only reaching Stage 1 by mid-2025.

Some project teams have even privately admitted that they may never fully decentralize. Vitalik cited a case in his post: a certain project argued that they would never further decentralize because "the regulatory needs of their clients require them to have ultimate control." This彻底激怒了Vitalik, and he responded毫不客气地:

"This might be the right thing for your clients. But it's clear that if you do this, then you are not 'scaling Ethereum.'"

This comment almost sentences to death all projects that fly the Ethereum L2 flag but refuse to decentralize. Ethereum wants a分身 that can extend decentralization and security to a broader space, not a group of附庸 that wear the skin of Ethereum but practice centralization.

The deeper problem is that there is an irreconcilable conflict between decentralization and commercial interests. Centralized sequencers mean project parties can control MEV (Maximal Extractable Value) income, respond more flexibly to regulatory requirements, and iterate products faster. Full decentralization means giving up this control and handing power over to the community and the validator network. For those with venture capital funding and growth pressure, this is a difficult choice.

If Layer2 had truly achieved full decentralization, would they still fall out of favor? The answer might still be yes. Because, Ethereum itself has changed.

When the Mainnet is Faster and Cheaper than Sidechains

Why does Ethereum no longer need Layer2 for scaling as much?

As early as February 14, 2025, Vitalik released a key signal. He published an article titled "Even in an L2-heavy Ethereum, there are reasons to have a higher L1 Gas limit," clearly stating that "L1 is scaling." This sounded more like comfort for mainnet fundamentalists at the time, but looking back now, it was actually the charge for the Ethereum mainnet to start competing with Layer2 again.

Over the past year, the scaling speed of Ethereum L1 has far exceeded everyone's expectations. Technological breakthroughs came from multiple dimensions: EIP-4444 reduced the storage需求 for historical data, stateless client technology made node operation lighter, and most critically, the continuous increase in the Gas Limit. At the beginning of 2025, Ethereum's Gas Limit was still 30 million; by mid-year, it had increased to 36 million, a 20% growth. This was the first significant increase in Ethereum's Gas Limit since 2021.

But this was just the beginning. According to the plans of Ethereum core developers, there will be two major hard fork upgrades in 2026. The Glamsterdam upgrade will introduce perfect parallel processing capabilities, and the Gas Limit will soar from 60 million to 200 million, an increase of over 3 times. The Heze-Bogota fork will add the FOCIL (Fork-Choice Enforced Inclusion Lists) mechanism, further improving block construction efficiency and censorship resistance.

The Fusaka upgrade completed on December 3, 2025, already allowed the market to witness the power of L1 scaling. After the upgrade, Ethereum's daily transaction volume increased by about 50%, the number of active addresses rose by about 60%, and the 7-day moving average of daily transaction volume reached a historical high of 1.87 million, surpassing the records of the 2021 DeFi peak period.

The results are astounding: Ethereum mainnet transaction fees have dropped to extremely low levels. In January 2026, the average transaction fee on Ethereum dropped to $0.44, a decrease of over 99% compared to the peak of $53.16 in May 2021. During off-peak hours, the cost of a transaction is often below $0.1, sometimes even as low as $0.01, with Gas prices as low as 0.119 gwei. This number is already close to Solana's level, and Layer2's biggest cost advantage is being quickly erased.

Vitalik did a detailed calculation in that February article. He assumed an ETH price of $2500, a Gas price of 15 gwei (long-term average), and a demand elasticity close to 1 (i.e., doubling the Gas Limit would halve the price). Under this assumption:

Censorship resistance demand: Currently, enforcing a transaction censored by L2 through L1 requires about 120,000 gas, costing $4.5. To reduce the cost to below $1, L1 needs to scale 4.5 times.

Cross-L2 asset transfer: Currently, withdrawing from one L2 to L1 requires about 250,000 gas, and depositing into another L2 requires 120,000 gas, totaling $13.87. With an ideal optimized design, it would only require 7,500 gas, costing $0.28. To reach the target of $0.05, it needs to scale 5.5 times.

Mass exit scenario: Taking Sony's Soneium as an example, PlayStation has about 116 million monthly active users. Using an efficient exit protocol (7,500 gas per user), the current Ethereum can just support the emergency exit of 121 million users within a week. But to support multiple applications of this scale, L1 needs to scale about 9 times.

And these scaling goals are being realized step by step in 2026. Technological progress has彻底改变了the rules of the game. When L1 itself can become fast and cheap, why should users still endure the cumbersome cross-chain bridging, complex interaction experience, and potential security risks of Layer2?

The security issues of cross-chain bridges are not unfounded fears. In 2022, cross-chain bridges became a major target for hacker attacks. In February, the Wormhole bridge was hacked for $325 million; in March, the Ronin bridge suffered the largest DeFi attack in history, losing $540 million; and bridge protocols like Meter and Qubit were successively breached. According to Chainalysis statistics, in 2022 alone, the total amount of cryptocurrency stolen from cross-chain bridges reached $2 billion, accounting for the majority of all DeFi attack losses that year.

Liquidity fragmentation is another pain point. With the surge in the number of Layer2s, the liquidity of DeFi protocols is分散到dozens of different chains, leading to increased transaction slippage, reduced capital efficiency, and worsened user experience. A user wanting to move assets between different Layer2s needs to go through a complex bridging process, wait for long confirmation times, and bear additional costs and risks.

This leads to the next, and most brutal, question: what should those Layer2 projects that raised huge funds and issued tokens do now?

Valuation Bubbles and Ghost Towns

Where did all the Layer2 money go?

In the past few years, the Layer2 track has been more like a huge financial game than a technological revolution. Venture capital firms waved checks, pushing the valuations of L2 projects to staggering heights. zkSync raised a total of $458 million, Arbitrum's背后的Offchain Labs was valued at $1.2 billion, Optimism raised $268.5 million, Starknet raised $458 million. Behind these numbers are top VCs like Paradigm, a16z, Lightspeed, Blockchain Capital.

Developers were keen on "套娃" (nesting) across different L2s, building complex DeFi legos to attract more liquidity and airdrop hunters. But real users were worn out by the tedious cross-chain operations and high hidden costs.

A残酷的现实is that the market is highly concentrated towards the top. According to data from crypto research firm 21Shares, the three major L2s—Base, Arbitrum, and Optimism—already control nearly 90% of the transaction volume. Base, leveraging Coinbase's traffic advantage and user base, achieved explosive growth in 2025, its TVL soaring from $1 billion at the beginning of the year to $4.63 billion by the end of the year, with quarterly transaction volume reaching $59 billion, a 37% increase quarter-on-quarter. Arbitrum followed closely with about $19 billion TVL, and Optimism after that.

But outside the top tier, most L2 projects, after losing the drive of airdrop expectations, saw their real user numbers quickly drop to freezing point, becoming名副其实的"ghost towns". Starknet is the most typical example. Although its token price has fallen 98% from its high, its price-to-earnings ratio is still in a very high bubble range relative to its extremely low daily active users and fee income. This means there is a huge gap between the market's expectations for its future and its current ability to create real value.

More ironically, when Layer2 fees dropped significantly due to EIP-4844, the data availability fees they paid to L1 also plummeted, which in turn reduced the fee income of Ethereum L1. In January 2026, an analysis pointed out that the Dencun upgrade caused a large number of transactions to shift from L1 to cheaper L2s, which was one of the main reasons for Ethereum network fees dropping to their lowest level since 2017. While Layer2 was reducing its own costs, it was also "emptying" the economic value of L1.

21Shares predicted in its 2026 Layer2 Outlook report that most Ethereum Layer2s may not survive 2026, the market will experience a残酷的整合, and ultimately only those projects with high performance, true decentralization, and unique value propositions will prevail.

This is the real intention behind Vitalik's recent "challenge". He wants to burst this "infrastructure circle-jerk" bubble and pour a bucket of cold water on this sick market. If a Layer2 cannot provide functions more interesting and valuable than L1, then it will ultimately become just an expensive transitional product in Ethereum's development history.

Ethereum is Taking Back Its Sovereignty

Vitalik's latest suggestion points out a new way out for Layer2: abandon "scaling" as the only selling point, and instead explore functional additional value that L1 cannot or is unwilling to provide in the short term. He specifically listed several directions: privacy protection (achieving on-chain private transactions through zero-knowledge proof technology), application-specific efficiency optimization (such as games, social networks, AI computing), ultra-fast transaction confirmation (millisecond level instead of second level), and exploration of non-financial use cases.

In other words, the role of Layer2 will change from Ethereum's "分身" (avatar) to functional "plugins". They are no longer the only savior for scaling, but a functional extension layer in the Ethereum ecosystem. This is a fundamental shift in positioning and a return of power—Ethereum's core value and sovereignty will be re-anchored on L1.

Vitalik also proposed a new framework: view Layer2 as a spectrum, not a binary classification. Different L2s can have different trade-offs in terms of decentralization degree, security guarantees, and functional characteristics. The key is to clearly explain to users what guarantees they provide, rather than all claiming to be "scaling Ethereum".

This清算has already begun. Those Layer2s that rely on expensive valuations but have no real daily activity are facing their final judgment. Those that can find their unique value positioning and truly achieve decentralization may survive in the new landscape. Base may continue to lead依靠Coinbase's traffic advantage and Web2 user onboarding ability, but it needs to face questions about its lack of decentralization. Arbitrum and Optimism need to accelerate the process towards Stage 2 to prove they are more than just centralized databases. ZK-Rollup projects like zkSync and Starknet need to prove the unique value of their zero-knowledge proof technology while significantly improving user experience and ecological prosperity.

Layer2 has not disappeared, but their era as Ethereum's only hope is彻底结束了. Five years ago, when cornered by competitors like Solana, Ethereum handed the hope of scaling to Layer2 and重构ed the entire technical roadmap for it. Five years later, it found that the best scaling solution is to make itself stronger.

This is not betrayal, but growth. And those Layer2s that cannot adapt to this evolution will become the cost. When the Gas Limit冲向200 million at the end of 2026, when Ethereum L1 transaction fees stabilize at a few cents or even lower, when users find they no longer need to endure the complexity and risks of cross-chain bridges, the market will vote with its feet. Those projects that once had sky-high valuations but did not create real value for users will be forgotten by history in this great浪淘沙.