Ethereum Glamsterdam: ePBS, BALs, and Gas Repricing Deep Dive

Ethereum Is Rewriting How Blocks Get Built — and It Matters More Than You Think

Sometime in the first half of 2026, Ethereum will activate Glamsterdam, a hard fork that fundamentally changes how blocks are constructed, priced, and validated on the network. This is not an incremental patch. It is the most ambitious execution-layer overhaul since The Merge transitioned Ethereum to proof of stake in September 2022.

Three pillars define the upgrade: EIP-7732 enshrines proposer-builder separation directly into the consensus layer, EIP-7928 introduces block-level access lists that unlock parallel transaction execution, and a package of repricing EIPs — led by EIP-7904 — recalibrates gas costs to reflect what modern hardware actually spends on each operation. Together, they represent Ethereum’s attempt to solve three problems simultaneously: trust assumptions in block production, sequential execution bottlenecks, and gas prices set in a different era.

The Block-Building Problem Glamsterdam Actually Solves

To appreciate what Glamsterdam changes, you have to understand what Ethereum’s block production looks like today — and it is not pretty.

Since The Merge, Ethereum has relied on an off-chain system called MEV-Boost to separate the roles of block proposers (validators) and block builders. Validators outsource the actual assembly of transactions to specialized builders who compete to construct the most valuable block. The winning bid gets forwarded through trusted relays, and the validator signs it sight-unseen.

This system works, but it depends on trusted intermediaries. QuickNode’s analysis estimates that the vast majority of Ethereum block production — between 80% and 90% — currently depends on off-chain builders. The concentration is striking: according to Bitfinex, the top three builders control more than 80% of all PBS blocks. Meanwhile, private transaction routing has grown to where over half of high-value Ethereum transactions bypassed the public mempool by mid-2025, per the same Bitfinex analysis.

The relay layer is a single point of failure. If relays go down, block production degrades. If relays collude, censorship becomes trivial. Ethereum’s credible neutrality rests on infrastructure that is neither enshrined in the protocol nor subject to its incentive mechanisms.

Glamsterdam’s answer is to bring the entire market on-chain.

EIP-7732: Enshrined Proposer-Builder Separation

EIP-7732 moves the proposer-builder handoff into the consensus layer itself. The mechanism uses a commit-reveal pipeline that eliminates relays entirely.

Here is how it works. A builder assembles a block and publishes a bid containing a payload commitment — a hash of the execution payload, not the transactions themselves. The proposer selects the highest bid and includes the commitment in their beacon block. Only after the beacon block is accepted does the builder reveal the full execution payload.

A new protocol actor called the Payload Timeliness Committee (PTC) validates that the reveal arrives on time and matches the commitment. The PTC does not execute transactions; it only checks timing and hash consistency. If a builder fails to reveal, the slot goes empty — but critically, the proposer still keeps the bid payment while the builder forfeits the full bid value. This asymmetry is deliberate: it penalizes builders for non-delivery while protecting validators from lost income.

This design solves several problems at once. Validators no longer need to trust external relays. MEV handoff rules are standardized at the protocol level. And the commit-reveal structure means validators cannot front-run builders, preserving the economic incentives that make the builder market competitive.

The Free Option Problem

Nothing this complex ships without trade-offs. Academic research has identified what is called the free option problem in ePBS: a timing asymmetry where builders can exploit the window between commitment and reveal to gain optionality on volatile blocks. Estimates cited by both Bitfinex and CryptoSlate put this at roughly 0.82% of blocks on average, rising to around 6% during volatile periods.

In practice, this means a small but meaningful percentage of blocks could see builders strategically withholding reveals when market conditions shift unfavorably between commitment and deadline. The forfeiture penalty mitigates this, but it does not eliminate it — builders with deep capital reserves can absorb occasional forfeitures if the option value justifies it.

There is also the centralization question. Capital-intensive builders may dominate the on-chain market just as they dominate the current relay-based one. Enshrining PBS removes the relay trust assumption, but it does not inherently solve builder concentration. The Ethereum Foundation’s 2026 priorities update acknowledges this by placing FOCIL (Fork-choice enforced Inclusion Lists, EIP-7805) under the Harden the L1 track — a censorship-resistance mechanism that would let a validator committee force-include transactions regardless of builder preferences. However, FOCIL did not make the Glamsterdam cut. Base’s engineering team has publicly warned that bundling FOCIL with ePBS could push the upgrade past 2026.

EIP-7928: Block-Level Access Lists

If ePBS restructures who builds blocks, EIP-7928 restructures how blocks execute.

Today, Ethereum processes transactions sequentially. Each transaction executes, and only then does the next one know what state it is working with. This is safe but slow. Even when two transactions touch completely different parts of the state — say, one interacts with a Uniswap pool and another updates an ENS record — the EVM processes them one after another.

Block-Level Access Lists (BALs) change this by making the state access pattern of an entire block explicit before execution begins. A BAL records every account and storage slot that transactions in the block will touch. The BAL root — a hash of this list — is included in the block header, while the full list travels with the block payload.

This unlocks several performance strategies. Clients can prefetch the entire working state into memory before executing a single transaction. Non-interfering transactions can be identified and executed in parallel. And — perhaps most significantly for the long-term roadmap — BALs enable executionless state reconstruction, where a node can verify a block’s state transitions without re-executing every transaction, a prerequisite for ZK-based light clients.

From Theory to Throughput

The throughput implications are substantial but depend heavily on implementation. CryptoSlate’s analysis cautions that real-world gains hinge on whether clients adopt concurrency across actual bottlenecks and whether the extra data and verification steps avoid becoming their own latency tax.

There is a tension here. BALs add data to every block — the full access list must propagate across the network alongside the execution payload. The Ethereum Foundation’s roadmap targets gas limits toward and beyond 100 million with BALs in place. Bitfinex goes further, citing an initial target of 100 million gas per block, with a longer-term target of 200 million once ePBS is fully operational. For context, Ethereum’s gas limit currently sits at approximately 60 million, per CryptoSlate — a level reached after the first gas limit increase since 2021, as noted in the Ethereum Foundation’s 2026 update.

The infrastructure changes are non-trivial. Per QuickNode, the Engine API must be extended for BAL payload delivery, a new eth/71 wire protocol handles peer-to-peer BAL serving, and execution-layer clients must retain BALs for 3,533 epochs. This is not a drop-in optimization — it is a new data layer woven into Ethereum’s networking stack.

A correctness risk also exists. If two clients compute different access lists for the same block, they will disagree on the BAL root, causing a consensus split. Datawallet’s risk assessment highlights deterministic access list mismatches as a potential source of chain forks during early activation.

Gas Repricing: Aligning Costs With Reality

The third pillar of Glamsterdam is less dramatic but arguably more impactful for everyday users: a sweeping recalibration of gas costs.

Many of Ethereum’s current gas prices were set years ago, when the EVM ran on different hardware and the state was a fraction of its current size. Operations that are cheap on modern processors are still priced as if they strain the network, while others that have become genuinely expensive as state grows remain artificially cheap.

EIP-7904 leads the repricing effort with a benchmark-driven approach: every EVM opcode, precompile, and memory operation is measured against actual CPU, storage, and bandwidth costs on current client implementations. The adjustments are conservative by design — the official EIP specification prioritizes increases for operations that empirical data show as underpriced, while applying decreases conservatively to avoid compromising security.

The repricing adjustments vary by category. Some compute-heavy operations see reductions where benchmarks confirm they were overpriced relative to actual resource consumption. Others — particularly those that were historically underpriced — see increases. For example, the KECCAK256 hashing operation has its base cost raised from 30 to 45 gas, reflecting updated measurements of its actual computational burden. The overall direction of EIP-7904 is to bring each operation’s gas cost into alignment with its real-world resource footprint, whether that means raising or lowering the price.

The Full Repricing Stack

EIP-7904 does not operate alone. EtherWorld’s analysis documents ten repricing EIPs spanning three categories:

Compute and memory: EIP-7904 (general repricing), EIP-7667 (hashing cost updates for ZK-VM compatibility), and EIP-7923 (replacing quadratic memory growth with a paged memory model).

State and storage: EIP-8032 (size-based SSTORE pricing that scales with contract storage depth), EIP-8037 (state creation cost increases), and EIP-8038 (SLOAD/SSTORE constant adjustments).

Data and accounting: EIP-7981 (access list floor costs), EIP-7976 (increased calldata costs to reduce spam), EIP-2780 (intrinsic transaction gas reduction), and EIP-7778 (refund accounting changes).

The intrinsic gas change in EIP-2780 deserves special attention. Base’s engineering team reports that it lowers the base transaction cost from 21,000 to 4,500 gas, bringing simple ETH transfers down to 6,000 gas total. To prevent spam exploiting the lower floor, it adds a GAS_NEW_ACCOUNT penalty of 25,000 gas for transfers that create new empty accounts.

The net effect of the full repricing package is substantial. Bitfinex projects a roughly 78% reduction in fees. But this number comes with caveats. Actual fee savings depend on network demand, block space utilization, and the specific operations in each transaction. Compute-heavy operations that were previously overpriced will see the largest drops; state-heavy operations may actually become more expensive as storage repricing better reflects the real cost of state growth.

The Smart Contract Risk

Gas repricing carries a unique risk that neither ePBS nor BALs share: it can break existing smart contracts. Any contract that hardcodes gas values — for example, forwarding exactly 2,300 gas in a transfer call, a pattern common before EIP-1559 — may behave differently when the underlying costs change. Datawallet warns that repricing may alter execution costs in ways that expose latent bugs in deployed contracts.

This is not theoretical. Previous repricing events — notably EIP-2929 in the Berlin hard fork — required EIP-2930’s access lists specifically to prevent contracts from becoming unusable after gas cost changes. The Glamsterdam repricing is far more comprehensive, touching compute, memory, storage, and data costs simultaneously.

The Bigger Picture: Ethereum’s Scaling Architecture

Glamsterdam does not exist in isolation. It is the third major upgrade in a twelve-month span, following Pectra (May 2025) and Fusaka (December 2025). Each delivered specific components of Ethereum’s scaling roadmap:

• Pectra introduced EIP-7702 (smart contract execution for EOAs) and doubled blob throughput
• Fusaka activated PeerDAS (EIP-7594), enabling what the Foundation describes as an 8x theoretical blob capacity increase
• Glamsterdam now targets the execution layer itself — how blocks are built, how transactions are processed, and what operations cost

The Ethereum Foundation’s three-track organizational restructuring — Scale, Improve UX, and Harden the L1 — reflects the scope of ambition. Glamsterdam is primarily a Scale-track deliverable, but its ripple effects touch all three. ePBS hardens the L1 by removing relay trust assumptions. BALs advance the zkEVM attester client initiative. Gas repricing improves UX by making transactions cheaper and more predictable.

The follow-up upgrade, Hegota, is already being planned for later in 2026. The features that did not make Glamsterdam — FOCIL inclusion lists, reduced slot times, multidimensional gas metering, post-quantum signatures — form its candidate list.

What Could Go Wrong

Glamsterdam bundles more protocol-level complexity than any Ethereum upgrade since The Merge. The risk surface is proportional.

Consensus fragility: BALs introduce a new consensus-critical data structure. If clients disagree on access list computation, the chain splits. Cross-client testing must be exhaustive — and Datawallet notes that public testnets and dual audits are planned for early-to-mid 2026, a compressed timeline for this scope of change.

Builder market dynamics: Enshrining PBS on-chain removes one centralization vector (relays) but may entrench another (capital-advantaged builders). Without FOCIL, the censorship-resistance improvements are incomplete.

Validator role transformation: CryptoSlate highlights a longer-term concern — validators are transitioning from re-executing blocks to verifying ZK execution proofs. This changes the hardware profile, economics, and operational requirements for running a validator. The changes in Glamsterdam accelerate this transition.

Timeline risk: The aspirational June 2026 target is aggressive. Base has explicitly warned that scope expansion — particularly FOCIL — could push the upgrade past 2026 entirely. Even without FOCIL, the combination of ePBS, BALs, and ten repricing EIPs represents significant cross-client implementation and testing work.

Implications for the Ecosystem

If Glamsterdam ships successfully, it redefines Ethereum’s competitive position. Parallel execution narrows the throughput gap with chains like Solana and Sui that were designed for concurrency from day one. Cheaper gas makes L1 transactions viable for use cases that had migrated entirely to L2s. And enshrined PBS gives Ethereum a credible-neutrality story that no other major chain can match.

For L2 rollups, the implications are mixed. Cheaper L1 data costs (via repricing and expanded gas limits) reduce rollup operating expenses. But a more capable L1 also reduces the urgency of moving activity to L2s in the first place. The Ethereum Foundation’s simultaneous investment in blob scaling suggests they see L1 and L2 as complementary rather than competitive — but the market will decide.

For DeFi protocols and dApp developers, gas repricing demands attention. Contracts that assume specific gas costs need auditing. Protocols that rely on gas-based access control or calldata pricing assumptions need to verify their logic against the new cost schedule. The testing window between devnet activation and mainnet deployment is where this work happens.

Key Takeaways

• EIP-7732 (ePBS) moves block building on-chain via commit-reveal, eliminating relay dependency but introducing new dynamics around builder concentration and the free option problem
• EIP-7928 (BALs) makes per-block state access explicit, enabling parallel execution and future ZK verification — but adds a new consensus-critical data structure with correctness risks
• EIP-7904 and nine companion repricing EIPs recalibrate gas costs to match modern hardware benchmarks, with notable reductions for compute-heavy operations and increases for state-heavy ones
• The upgrade scope is massive: two headliner EIPs plus ten repricing proposals, requiring new wire protocols, API extensions, and extensive cross-client testing
• Glamsterdam is part of a rapid three-fork sequence (Pectra → Fusaka → Glamsterdam) that collectively reshapes Ethereum’s execution, data availability, and block production layers in under twelve months