Core thesis:

Vitalik Buterin is walking back a foundational assumption of Ethereum's rollup-centric roadmap: that Layer 2s primarily exist to provide generic, cheaper execution as extensions of L1. With Ethereum mainnet fees structurally lower, gas limits already rising, and deeper protocol-level scaling work underway, "L2s as branded shards" is no longer a sufficient justification.

This is not an anti-L2 position. It is a reframing of what counts as legitimate scaling in an ecosystem where L1 itself is evolving faster than expected.

The rollup-centric vision meets a scaling L1

The original rollup roadmap relied on three assumptions moving in parallel:

L2s would decentralize quickly (progressing to Stage 2).
Cross-rollup interoperability would mature toward near-synchronous composability.
L1 blockspace would remain scarce enough to push most activity outward.

The third assumption is increasingly false.

Ethereum's gas limit has already risen to roughly 60M gas, up from the long-standing ~30M ceiling, and core developers have openly discussed further increases toward 80M gas and beyond. Longer-term proposals such as EIP-9698 outline a gradual, scheduled path toward much higher limits starting in 2026, though these remain proposals rather than protocol guarantees.

Crucially, L1 usage has continued to grow without fee pressure. Ethereum processed around 2.2 million daily transactions in late 2025 while average fees remained far below prior cycle highs. This undermines the idea that generic L2s are the only viable path to scale Ethereum execution.

L2 decentralization: the slowest-moving constraint

Decentralization progress on L2s has lagged far behind early expectations.

Most Ethereum rollups remain at Stage 0 or Stage 1, relying on centralized sequencers, upgrade keys, or governance controls that can override user safety. Stage 2 rollups, those that are effectively trustless, are almost nonexistent.

Vitalik has been unusually direct about the implications. In February 2026, he wrote:

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.

"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."

Vitalik Buterin, X thread, Feb 3, 2026

https://x.com/VitalikButerin/status/2018711006394843585?s=20

Elsewhere, he has summarized the problem more bluntly: L2 decentralization has progressed "far slower" than expected, while the base layer has quietly improved.

This matters because the cost advantage of L2s is narrowing, while their trust assumptions are not.

Activity vs. capital: what the data actually shows

On-chain metrics reveal a widening gap between usage and economic commitment across L2s.

L2Beat's Activity dashboard shows that aggregate rollup user operations per second (UOPS) exceed Ethereum L1's throughput, driven primarily by a small number of rollups such as Arbitrum and Base.
Source: https://l2beat.com/scaling/activity

However, L2Beat's Total Value Secured (TVS) data shows where capital actually concentrates:

Arbitrum One: around $15B.
Base: around $9.8B.
OP Mainnet: around $1.6B.
Starknet: around $0.6B.

Source: https://l2beat.com/scaling/tvs

More tellingly, reporting based on L2Beat data indicates that total value secured across rollups declined roughly 13% year-over-year, even as transaction activity increased. This suggests users are willing to transact on L2s, but less willing to park capital under their security guarantees.

The long tail is even starker: over 80% of Ethereum L2s record fewer than 1 user operation per second, effectively placing them below the threshold of sustained economic relevance.

What Vitalik is actually proposing

Vitalik's pivot is not about abandoning L2s, but about narrowing their mandate.

He has repeatedly pointed toward L2s that do at least one of the following:

Offer execution environments L1 cannot or should not support, such as privacy-preserving VMs.
Pursue aggressive specialization, including ultra-low-latency trading, high-throughput social or gaming, or app-specific rollups.
Embed protocol-level features inappropriate for L1, such as native oracles, identity primitives, or explicit compliance hooks.

He has also been explicit that L2s handling ETH or systemically important assets should meet minimum decentralization standards (at least Stage 1). Below that bar, they increasingly resemble independent L1s with Ethereum settlement as a branding layer.

Vitalik Tracks Ecosystem Reaction and Doubles Down on Innovation

In the days following his initial critique of generic Layer-2 execution environments, Vitalik Buterin explicitly acknowledged the community responses and sharpened his focus on the underlying issue: the prevalence of “copy-paste” EVM chains that add little beyond low-cost blockspace.

In a February 5 post on X, Buterin wrote:

“Have been following reactions to what I said about L2s about 1.5 days ago. One important thing that I believe is: ‘make yet another EVM chain and add an optimistic bridge to Ethereum with a 1-week delay’ is to infra what forking Compound is to governance — something we’ve done…”
— vitalik.eth (X, Feb 5, 2026)
https://x.com/VitalikButerin/status/2019341766407725170?s=20

This phrasing deliberately parallels familiar crypto design patterns — comparing “yet another EVM chain + bridge” to the repetitive act of cloning governance models like Compound’s. The implication is clear: building more of the same offers diminishing returns in an ecosystem where the base layer is scaling and where trust assumptions vary widely across rollups.

Buterin’s follow-up suggests he is not discouraging experimentation per se, but discouraging inertia. Generic replication — whether of governance models or EVM chains with rudimentary bridges — no longer advances Ethereum’s architectural goals in a meaningful way. Instead, he is signalling that L2s should deliver genuinely new infrastructure, trade-offs, or primitives that the base layer or existing networks do not already provide.

Hyper-scaling state: the deeper context

Vitalik's broader thinking is captured in his Ethereum Research post, Hyper-scaling state by creating new forms of state:
https://ethresear.ch/t/hyper-scaling-state-by-creating-new-forms-of-state/24052

In it, he argues that Ethereum's long-term scalability depends not just on execution and data availability, but on introducing new, cheaper, and more restrictive forms of state. Developers will need to explicitly choose between flexibility and cost, and many applications will need to be re-architected to take advantage of these new state models.

This context matters. If both execution and state economics are changing at the L1 level, then L2s optimized purely for "more TPS" are addressing a shrinking part of the problem.

Case studies: L2s that are already specializing

Specialization is not theoretical. It has precedent.

StarkEx is an early example of an application-specific rollup optimized for exchange workloads. Rather than cloning the EVM, StarkEx tailored its execution and proving systems for high-throughput trading and settlement, supporting exchanges and trading venues with drastically lower per-transaction overhead.
https://starkware.co/starkex/

More broadly, Ethereum research has long discussed app-specific rollups as a distinct category: rollups built around a single application or narrow workload rather than general-purpose execution.
https://ethresear.ch/t/rollup-as-a-service-opportunities-and-challenges/13051

These systems sacrifice generality in exchange for performance, predictability, and simpler trust models, exactly the tradeoff Vitalik is now emphasizing.

DeFi implications: fewer chains, stronger differentiation

For DeFi, the implications are structural.

First, liquidity fragmentation becomes harder to justify. Dozens of near-identical EVM L2s dilute composability without offering unique guarantees or performance. As L1 becomes cheaper, some activity should rationally collapse back to mainnet.

Second, specialized environments become more attractive. Privacy-preserving lending, intent-based trading with sub-second latency, or protocol-specific rollups become easier to defend when general-purpose scaling is no longer the default.

Third, integration standards tighten. Supporting an L2 increasingly requires explicit reasoning about decentralization stage, bridge risk, upgrade authority, and interoperability, not just incentives or TVL.

Counterarguments: the monolithic and L1-maximalist view

Critics of the rollup-centric approach argue that multi-layer complexity imposes real costs: fragmented liquidity, weaker UX, withdrawal delays, and bridge risk. From this perspective, a highly capable monolithic L1 avoids many coordination failures inherent to multi-chain systems.

Community and industry debate often frames this as a tradeoff between architectural purity and operational simplicity. Some alternative ecosystems and modular chains implicitly critique Ethereum's approach by emphasizing simpler trust assumptions or fewer moving parts.

Vitalik's recent comments partially validate this critique, but only to a point. Rather than rejecting modularity, he is narrowing it to cases where it delivers clear, defensible advantages.

Ethereum's post-rethink role

Zooming out, Ethereum increasingly looks less like a congested execution layer and more like a high-capacity settlement and coordination layer.

Two second-order effects follow:

The argument that L2s systematically dilute L1 value capture weakens as more activity returns to mainnet, increasing fee revenue and ETH burn.
The L2 ecosystem faces a shakeout. Forks and clones become harder to defend; credible differentiation becomes mandatory.

If this transition succeeds, the outcome is fewer, more opinionated rollups and a DeFi ecosystem that is less fragmented, more interoperable, and more explicit about its tradeoffs.

Vitalik's L2 Rethink and What It Means for Ethereum Scaling