Ethereum was cheaper than expected in 2020, and rollup decentralization was slower than promised in 2021. These two realities are forcing the ecosystem to rewrite the purpose of “Layer 2.”
Vitalik Buterin’s recent post on Ethereum research bluntly states this change. Ethereum’s original vision of a Layer 2 (L2) blockchain as a “branded shard” is no longer viable, and the ecosystem needs a new path forward.
However, this is not a waiver. Instead, this re-layers expectations and more clearly defines what different types of rollups are actually building.
What’s at stake now is the new job description, as the assumptions underlying the rollup-centric roadmap have weakened.
Stage 2 is rare
L2BEAT provides the clearest framework for understanding rollup decentralization through the Stages system.
Stage 0 indicates that the training wheels remain in place and meaningful trust assumptions persist.
Stage 1 represents partial decentralization with stronger escape hatches and proof guarantees, but still retains meaningful upgrades or governance trust.
Stage 2 is a “no training wheels” milestone, where critical safety characteristics are enforced by code rather than by arbitrary actors.
This is illustrated by the current distribution of value secured across the L2 ecosystem. According to the L2BEAT rollup scaling overview, approximately 91.5% of the listed values are included in the Stage 1 rollup, 8.5% are included in Stage 0, and approximately 0.01% are included in Stage 2.
The top three rollups by value account for about 71% of the total, indicating that “stage 2 progress” is largely dependent on the decisions of a few of the largest projects, rather than what smaller experimental chains try.
A central hurdle is whether the proof system can be overridden and whether upgrades will face significant delays and constraints.
Upgrade discretion remains common in the largest rollups, and moving beyond that is proving to be slower and more difficult than the optimism of 2020-2021 anticipated.
Some projects have made it clear that they may not wish to proceed beyond Stage 1 due to regulatory requirements that require absolute control, as well as technical constraints related to zkEVM safety.
While this is a legitimate product decision for certain customer bases, it makes clear that these chains are not “scaling Ethereum” in the sense that their rollup-centric roadmaps originally meant.
| project | stage | TV ($) | proof type | Upgrade key/existence of security council? | Precautions |
|---|---|---|---|---|---|
| Arbitrum One | 1 | 16.16B | optimistic | yes | Emergency passes can skip delays |
| base chain | 1 | 10.99B | optimistic | yes | Upgrades approved by multiple stakeholders. No delay |
| OP mainnet | 1 | 1.88B | optimistic | yes | Security Council Instant Upgrade Authority |
| writer | 0 (app chain) | 1.27B | valid | yes | 21 days delay, emergency may be 0 |
| stark net | 1 | 676.17M | valid | yes | Security Council can be upgraded without delay |
| ink | 1 | 523.71M | optimistic | yes | Security Council + Foundation Approval. No regular delays |
| Linea | 0 | 492.93M | valid | yes | Multisig can be upgraded without delay |
| ZKsync era | 0 | 417.07M | valid | yes | Emergency board can avoid upgrade delays |
| Katana | 0 | 297.94M | valid | yes | Security Council can eliminate upgrade delays |
| unichain | 1 | 168.81 million | optimistic | yes | There is no exit window for regular upgrades. instant power |
Why the constraint was changed
On October 2, 2020, the Fellowship of Ethereum Magicians’ post “Rollup-centric Ethereum Roadmap” described the original paper.
Gasoline prices rose, some applications were forced to shut down, and the ecosystem came to the conclusion that it would be an “all-in roll-up” in the short to medium term.
Base layer scaling must prioritize rollup data capacity, and users increasingly live on L2.
Since then, two hard facts have changed. First, L1 is currently significantly cheaper. Etherscan shows that the 7-day average trading fee is around $0.35 and the gas snapshot is a fraction of 1 gwei.
On January 16th, Ethereum hit an all-time high of 2,885,524 transactions in a single day. The narrative is “busier, cheaper”, the exact opposite of the 2020 crisis that motivated the rollup-centric roadmap.
Second, L1 executive ability has improved. Ethereum’s block gas limit was raised to approximately 60 million from the long-standing limit of 30 million after extensive validator signaling in late 2025.
In blocks of about 12 seconds, 60 million gases equals about 5 million gases per second.
Ambitious community discussions have mentioned a lofty goal of 180 million gas, equivalent to a three-fold increase, but it remains a direction rather than a commitment.
Clean Interpretation: The 2020 assumption that “L1 doesn’t scale to most users” is weakened by today’s pricing structure. This creates room for L2 to be a range of trade-offs between security and sovereignty, rather than competing solely on price as “shards” that are all nearly identical.
L2 as a spectrum, not a clone
The reframing proposed by Buterin treats L2 as occupying the entire spectrum.
On one end, we have Ethereum’s full trust and trust-backed chain, which has unique characteristics such as not only EVM clones, but also privacy-focused systems, non-EVM execution environments, or ultra-low latency sequencers.
At the other end, there are options for different levels of Ethereum connectivity that users and applications can choose from based on their specific needs.
The new minimum standard is simple. If you are working with ETH or Ethereum issued assets, you must reach at least stage 1.
Otherwise, it’s just another L1 with a bridge, so you should call yourself that. The hurdle to differentiation is even more difficult: being the best in something other than “cheap EVM.”
Examples Buterin cited include privacy, application-specific efficiency, truly extreme scaling beyond even extended L1, fundamentally different designs for non-financial applications such as social and identity systems, ultra-low latency sequencing, or computationally unverifiable features such as built-in oracles and decentralized dispute resolution.
The mechanisms driving this are still under investigation. “Native rollup precompilation” allows Ethereum to verify standard zkEVM proofs within the protocol.
For the “EVM Plus Additional Features” rollup, this means that regular EVM validation is done trustlessly at the protocol level, and the rollup only needs to certify the custom extensions individually.
This allows for stronger interoperability and potentially paves the way for synchronous composability, where contracts between different rollups can interact within the same transaction. However, this is still a research trajectory and not an introduced feature.
The January 16th post “Combining base rollups and prechecks for synchronous composability” and the February 2nd post “Synchronous composability between rollups with real-time proofs” illustrate the design space, but do not represent changes to the shipped protocol.
Three buckets appear
Once this reorganization takes hold, we expect rollups to be split into more distinct categories.
The first bucket is a stage 2 tracking settlement rollup that maximizes Ethereum’s security inheritance.
These projects treat “scaling Ethereum” as their core mission and aim to achieve code-enforced guarantees with minimal discretionary governance.
The second bucket is a regulated or managed execution environment.
These are optimized for compliance, permitting, or specific organizational requirements. By design, it may never move beyond stage 1, and it needs to honestly sell that control as a feature, rather than pretending to offer complete decentralization.
The third bucket is specialized chains optimized for latency, privacy, app-specific execution, or non-financial use cases.
Privacy rollups that use zkProofs to hide transaction details, ultra-low latency sequencers for trading applications, or social or identity systems with fundamentally different state models all fall into this category.
They don’t have to be EVM compatible or even economical to justify their existence, they just have to provide value to users that they can’t get anywhere else.
Projects like Arbitrum One, Optimism, Base, zkSync Era, and Starknet each have to decide which category to pursue. Although the ecosystem is large enough to support all three, the assumption that all L2s perform the same function is fading.
Changes for users and builders
For users, the burden shifts to understanding warranties. Evacuation hatches, upgrade delays, attestation systems, and censorship resistance are product differentiators rather than assumed characteristics.
Wallets and interfaces need to more clearly label their trust assumptions, and the L2BEAT Stages framework aims to make these assumptions easier to read.
For builders, “cheap EVM” has become a commodity. Differentiation moves to privacy and custom virtual machines, ultra-low latency sequences, app-specific throughput optimization, non-financial applications in social, identity, or AI contexts, or compliance and permissions as an explicit product, without claiming to be “scaling Ethereum.”
As for the broader market story, we can expect to see a larger debate about whether L2 will actually “inherit the security of Ethereum” rather than as an aspiration.
This criticism was already a hot topic among rival L1 proponents, and the ecosystem’s recognition that many large rollups are stuck in Stage 1 due to discretionary governance has given it even more traction.
Is the L2 revolution about to begin?
It is unlikely that Ethereum will see an L2 revolution. Instead, retiering occurs.
The rollup-centric roadmap assumed that L2 would be nearly identical “branded shards” competing primarily on cost, while L1 would remain expensive and capacity-constrained.
That assumption no longer holds true. L1 is cheap and expanding, but L2, despite stage 2 decentralization, is divergent in security models and use cases faster than it converges.
The new path acknowledges that reality. L2s that store ETH or Ethereum-issued assets must meet at least Stage 1 minimum security standards. And beyond that, we need to compete on expertise and clear guarantees, rather than pretending to be compatible.
Research on native validation primitives and synchronous composability shows that Ethereum aims to facilitate that, but these are trajectories, not deployed features.
My job content has changed.
The minimum standard is to provide reliable security when dealing with Ethereum assets. The criteria for differentiation is being the best at something and being true to the trust model.
The rollup-centric roadmap has been upgraded to accommodate the reality that L1 is expanding and L2 is more diverse than the original vision.