L1 vs L2: Does Ethereum Still Need Its L2s after Glamsterdam in 2026?

TL;DR

• For years, Ethereum planned to stay lean at the base and push activity out to L2s, a plan set in motion by the Dencun upgrade (March 2024) that made L2 costs cheap. Glamsterdam rebalances that plan by making the base layer itself much more capable.
• Every Ethereum block has a “gas limit,” a cap on how much computation it can hold. That cap is around 60M today and is targeted toward a 200M floor after Glamsterdam, which roughly triples base-layer capacity.
• Ethereum captured only 4.9% of its application-layer fees in Q2 2026, so more L1 activity could improve how much value ETH captures.
• For builders, the old rule that “L1 is too expensive” no longer holds, and a small set of contracts should be tested for the gas repricing. 
• The question Glamsterdam raises is what an L2 should exist for once the base layer becomes cheap and capable on its own.

Ethereum Is Reconsidering What “Scaling” Means

For most of the last six years, “scaling Ethereum” meant one thing: move activity off the base layer onto L2s. Glamsterdam is set to scale the base layer directly.

The upgrade is the largest protocol change since The Merge, shipping ten Ethereum Improvement Proposals, the formal design documents known as EIPs, grouped under a master proposal called EIP-7773. The headline is a strategic rebalancing of where computation happens.

The old plan was simply to keep the base layer lean and secure, and let L2s handle the volume of everyday transactions.

The emerging plan is different. The base layer itself becomes much more capable, and L2s are pushed to specialize on features the base layer cannot offer.

To get better context on the upcoming upgrade, read our full overview on Glamsterdam.

How Ethereum Became a Rollup-Centric Chain

Ethereum‘s scaling plan since 2020 was “rollup-centric,” meaning rollups did most of the work. 

A rollup is the most common type of L2: it bundles a large batch of transactions together off-chain, then posts a compressed summary back to Ethereum for security.

The plan had three parts: keep L1 lean, push execution to rollups, and let rollups inherit Ethereum‘s security. L2s acted as secure extensions of Ethereum, often called “branded shards”, meaning offshoots that carry the Ethereum name and trust model.

Then came the economic accelerant. EIP-4844 (an upgrade named Dencun, March 2024) introduced “blobs,” a cheap temporary data lane that rollups use to post their summaries to L1.

Blobs reduced average L2 fees by 80 to 90%, bringing costs to a fraction of a cent on major networks. Activity and fees then migrated to rollups in large volume.

By 2026, L2Beat data shows the total value held across Ethereum Layer 2 networks has exceeded $40 billion, with Arbitrum and Base holding the largest shares. However, once rollups switched to the cheap blob lane, mainnet base fees compressed and Ethereum collected little from all that L2 activity.

Recent upgrades like Dencun and Fusaka were tuned to help L2s, while Glamsterdam is the first in a while aimed at the base-layer engine itself.

Glamsterdam: Scaling the Ethereum Engine

Glamsterdam does its work through three coordinated changes. They are easier to understand together than as three isolated EIPs, so each one below comes with a plain-English purpose.

ePBS (EIP-7732) separates the job of building and proposing the block, making block production more reliable and less dependent on outside middlemen.

Today that handoff runs through off-chain helpers called relays, such as MEV-Boost, which ePBS makes optional once the protocol settles builder payments on its own. ePBS also extends the time window nodes have to share block data.

Ethereum‘s roadmap says this could expand the main data-sharing window from about two seconds to roughly nine seconds. That extra time is what makes larger blocks safe to send around the network without slowing it down.

BALs (EIP-7928), short for Block-Level Access Lists, let the network process many transactions at once instead of one after another. Each block has a map of every account and storage slot the block will touch, so nodes can read data in parallel.

It means a bigger block would not take proportionally longer to validate, letting the capacity rise without punishing the people who run nodes.

Gas repricing (EIP-8037 / EIP-8038) updates the cost of certain operations so prices match the real hardware burden they create, keeping growth sustainable.

EIP-8037 was the final change that gave client teams a safe ceiling under which the 200M gas limit could rise without bloating the database past 120 GiB of new data per year. Runaway database growth would otherwise make running a node too expensive for ordinary participants.

Here is how the three fit together:

Change	Purpose
ePBS	Makes bigger blocks easier to build and pass around
BALs	Lets transactions be processed in parallel, so bigger blocks stay fast
Repricing	Keeps higher capacity affordable to run long-term

What the 200M Gas Limit Unblocks

Gas here is the unit that measures how much computational work a transaction takes, and the “gas limit” caps how much work fits in each block, so a higher limit means more transactions per block.

200M is not written into the fork, which is the point most articles blur. Glamsterdam does not include any EIP that raises the gas limit itself, because that number is chosen by validators through a voting mechanism.

The fork makes a higher limit safe, and validators then raise it gradually by signaling their preference over time. The trajectory runs through four staged steps:

1. Roughly 60M gas today.
2. Higher reference limits during testing. Developers are currently testing at a 150M reference block gas limit to set accurate prices.
3. A 200M floor after Glamsterdam. In May 2026, the Ethereum Foundation aligned on a credible post-Glamsterdam target of 200 million, which applies once the upgrade ships and validators raise the limit.
4. Potentially higher later.

Some outlets cite a 10,000 TPS target, roughly ten times what Ethereum handles today, though that figure depends heavily on the type of activity. Real throughput depends on six inputs:

• transaction type,
• calldata, the data a transaction carries,
• execution complexity,
• state access, how much stored data it reads or writes,
• block composition,
• whether you count base-layer transactions or the compressed rollup transactions posted on top.

200M gas roughly triples available L1 capacity, with lower and more predictable fees, and no single TPS number does it justice.

State growth is the reason repricing exists. “State” is the running record of every account and balance, and it must be stored by every node forever, so the real test is whether Ethereum can process 200M gas while keeping a node affordable to run.

The Bigger Question: Does Ethereum Still Need Its L2s?

Why the original thesis is being reconsidered

In February 2026, Vitalik Buterin, Ethereum‘s co-founder, argued that the original vision of L2s no longer makes sense, citing two developments that undermined the plan.

The first development is decentralization stalling. L2s are measured on a maturity scale from Stage 0 (a central operator can still intervene) up to Stage 2 (fully trustless).

He noted that L2 progress toward that top stage has been far slower than expected. By 2025, only three major L2s, Arbitrum, OP Mainnet, and Base, reached Stage 1, while most rollups stayed at Stage 0 with central control.

The second development is a cheaper base layer. Following recent upgrades, fees stayed low and the assumption that L1 would be too expensive for most users became largely untrue.

On mainnet, basic transactions have run around $0.02, rising to roughly $0.48 for a token swap. Those prices are low enough that many everyday users no longer rely on L2 to afford Ethereum.

Buterin was blunt about what does not count as scaling Ethereum. He wrote that a fast EVM chain connected to L1 only by a “multisig bridge,” a link controlled by a small group of signers, does not contribute to Ethereum in a meaningful way.

Why L2s are not going away

In short, L1 cannot hold the volume that L2s handle today.

One Q2 2026 dataset shows rollups processing around 1,270 user operations per second, versus just 20.4 on the Ethereum mainnet. A user operation is a single user-facing action, such as a transfer or a swap.

A cheaper base layer can reclaim some of that activity. It cannot absorb roughly 60 times the volume at near-term capacity.

Rollups are also settling into a workable structure. A May 2026 analysis describes the L2 market organizing into a hub-and-spoke model: a few high-throughput general chains, surrounded by smaller application-specific rollups that share underlying infrastructure.

Layer	Primary advantage
Ethereum L1	Security, neutrality, composability, and a growing supply of blockspace
General-purpose L2	Cheap, high-volume execution
Specialized L2 / app-chain	Custom execution, privacy, tailored user experience, app-specific design

The Economic Question: Who Captures the Value?

In Q2 2026, applications on Ethereum produced $1.79 billion in fees, but the base layer itself kept only $88.4 million, or 4.9% of the total. The near-term picture is even starker at the settlement level, meaning what L2s pay Ethereum in practice.

Per growthepie data, L2 payments to Ethereum have run near zero for stretches. Even Base has paid about $6.41K on a busy day and Arbitrum under $400 per day for all of its activity.

Why did this happen? Dencun made L2 settlement extremely cheap by design, and the cheap blob fees that resulted reduced the fee pressure that used to power ETH burn.

Daily L1 fees fell from tens of millions in 2021 and 2022 to single-digit millions in 2026, and the argument that heavy burn makes ETH scarcer over time, lost momentum.

If Ethereum offers abundant cheap blockspace, fees can stay low even as usage climbs. Higher capacity can lift activity while holding down the fees per transaction.

That is why three separate ideas should not be blurred together:

1. Network usage. The count of transactions, apps, and active users.
2. L1 fee rewards. How much of that activity Ethereum itself collects.
3. ETH value accrual. Burn, staking economics, and demand for ETH as the network’s native asset.

Glamsterdam may improve Ethereum‘s value-capture position, though the link between capacity, fees, burn, and ETH demand is not a straight line.

Rollup fee migration is a deep pattern that a single upgrade will not reverse. The optimistic case is that if real-world asset settlement and stablecoin activity grow enough to strain blob supply, fee pressure comes back to L1 and burn picks back up.

Who Wins and Who Loses?

Glamsterdam‘s impact lands differently on each group. Here is a stakeholder view.

Stakeholder	Glamsterdam impact
L1 builders	More capacity and better parallel execution via BALs
L2s generally	Greater pressure to differentiate
Generic EVM L2s	Potentially weaker value proposition
Specialized L2s	Potentially stronger case
Mainnet applications	Fewer capacity constraints
Users	Potentially lower and more predictable fees
Validators	More demanding block-production infrastructure
Stakers	Greater potential L1 activity and economic relevance
ETH	Better potential value capture, with no assured price effect

What Glamsterdam Means for Builders

For some applications, staying on L1 wins because of:

• liquidity, the depth of assets already on L1,
• composability, the ability of contracts to plug into each other,
• security,
• existing infrastructure,
• cheap transactions.

For other applications, moving to an L2 still wins because of:

• high throughput,
• custom execution,
• privacy,
• specialized sequencing, control over transaction ordering,
• custom fee structures.

The guidance is to decide with real gas math on today’s low mainnet costs. The old habit of assuming L1 is unaffordable no longer holds.

There is also a timely warning for existing contracts. On August 24, 2026, the Ethereum Foundation flagged that the Glamsterdam repricing changes the cost of creating and accessing state, and a small set of contracts rely on assumptions the new prices change.

Most flagged issues are fixed by raising a transaction’s gas limit, and the large majority of contracts are unaffected. 

Reports note that contracts using hardcoded gas amounts, such as 2,300-gas stipends, fixed call limits, or gasleft logic, face the greatest risk. Wallets and fee estimators are affected too, since sending ETH to a brand-new account will add a small extra charge on top of the usual 21,000-gas cost.

However, most contracts are fine, and developers should still test gas-sensitive applications, because EIP-8037 and EIP-8038 change what state creation and state access cost.

Developers can test on the Platåberget test network ahead of the public testnets.

Where Everstake Fits, and the Institutional Angle

Glamsterdam is, for validators, mainly a change in how blocks get built, built around ePBS, BALs, and repricing. Under ePBS, validators take on new duties around builder commitments, payload checks, and timing.

Staking smart contracts may need system changes to monitor builders and select them in a trustless way. The move toward proof-based verification is a longer-term roadmap goal, separate from anything Glamsterdam requires right now.

Glamsterdam raises the infrastructure bar, and a demanding upgrade is where the difference between operators becomes visible.

It affects anyone running or delegating stake. Everstake has operated Ethereum validators since the network’s move to Proof-of-Stake, maintaining a clean track record with no slashing events on major networks and serving over 1.6 million delegators to date.

As a non-custodial provider, Everstake lets delegators keep full control of their own assets while the operator handles node reliability, client upgrades, and readiness for demanding changes like ePBS where operational discipline separates validators.

For institutions weighing L1 versus L2 exposure, the base layer is becoming a more serious place to run applications, mainnet composability and neutrality remain advantages, and validator infrastructure is getting more demanding.

What Comes After Glamsterdam

Glamsterdam is one step in a longer, multi-upgrade plan for scaling L1. It is part of the Ethereum Foundation’s broader roadmap, which Everstake covers in the Strawmap five north stars explainer.

The next major upgrade is Hegotá, currently listed for 2027, headlined by FOCIL (EIP-7805), a feature aimed at making transaction inclusion harder to censor. Further out sits the long-term goal of verifying Ethereum blocks with zero-knowledge proofs, a cryptographic method that lets a node confirm a block is valid without re-running all of it, known as an L1 ZK-EVM, which has no confirmed date.

Native account abstraction, which would make wallets more flexible, is also on the improvement track. For readers new to how these transitions work, Everstake‘s soft fork vs hard fork guide covers the mechanics.

Ethereum‘s roadmap currently lists Glamsterdam for Q4 2026 and Hegotá for 2027.

The Sepolia test network fork is provisionally set on Oct. 6, 2026, at 13:53 UTC for Glamsterdam, a date that remains conditional because the fork has yet to run stably on any private devnet, though the developers deferred the final decision and no mainnet date exists yet. Treat the mainnet target as dependent on test network readiness.

FAQ

Is Ethereum abandoning rollups?

No. Ethereum is rebalancing the division of labor, so the base layer reclaims some workloads while L2s are pushed to specialize in what they do best.

When does Glamsterdam go live?

Mainnet is targeted for Q4 2026, subject to testing readiness. The Sepolia test network fork is provisionally Oct. 6, 2026, at 13:53 UTC, and no mainnet date is confirmed for Glamsterdam.

Will the gas limit definitely hit 200M?

It remains a target that depends on validator signaling, the ongoing vote in which operators choose the limit. The Glamsterdam fork makes 200M safe to reach, and validators raise the limit over time.

Does Glamsterdam lower gas fees?

Directionally, more capacity supports lower and more predictable fees. Actual fees on Ethereum still depend on how much demand competes for each block.

Is Glamsterdam positive for ETH?

It may improve the value-capture setup, though the translation to price is contested. Network usage, L1 fee rewards, and ETH value accrual are three separate things.

What is the difference between Glamsterdam and Fusaka?

Fusaka delivered foundational changes that mostly helped L2s. Glamsterdam targets the base-layer engine itself through ePBS, BALs, and repricing.

Disclaimer:
This article is informational and does not constitute financial, legal, or tax advice. Upgrade dates, gas targets, and EIP scope are moving targets and may change before mainnet.