Ethereum's most consequential protocol upgrade since The Merge — codenamed Glamsterdam — has slipped from its original June 2026 target to Q3 2026. The fork enshrines Proposer-Builder Separation (ePBS) directly into the consensus protocol via EIP-7732, eliminating dependence on off-chain relay in...
"Ethereum has effectively been running in single-lane mode: one long queue of transactions, executed strictly in order. Even with multi-core CPUs, validation stays sequential. BALs change that." — Toni Wahrstätter, Lead Author EIP-7928, Ethereum Foundation Researcher
Ethereum's most consequential protocol upgrade since The Merge — codenamed Glamsterdam — has slipped from its original June 2026 target to Q3 2026. The fork enshrines Proposer-Builder Separation (ePBS) directly into the consensus protocol via EIP-7732, eliminating dependence on off-chain relay infrastructure that currently routes over 90% of Ethereum blocks through two dominant builders.
Simultaneously, Block-Level Access Lists (EIP-7928) introduce parallel transaction execution to a network that has processed transactions sequentially since launch. Combined with EIP-7904's empirical gas repricing — benchmarked against modern hardware rather than 2016-era assumptions — the upgrade targets a gas limit expansion from 60 million to 200 million per block. The aggregate effect: approximately 10,000 transactions per second on the base layer, a 10x increase over current throughput.
The delay stems from ePBS implementation complexity, leadership changes at the Ethereum Foundation, and a deliberate scope reduction that moved censorship-resistance features (FOCIL) to the subsequent Hegotá fork. Devnets are running. Public testnets remain weeks away.
Glamsterdam — combining "Gloas" (consensus layer) and "Amsterdam" (execution layer) — was initially announced for H1 2026 activation. The Ethereum Foundation confirmed in May 2026 that devnets (Devnet-0 through Devnet-5) achieved stability in ePBS after multi-client testing at the Soldøgn interop conference.
The slip to Q3 2026 results from three converging factors:
Technical complexity. EIP-7732's ePBS implementation requires fundamental changes to the slot structure, expanding the data propagation window from 2 seconds to approximately 9 seconds. This is not a parameter tweak — it restructures how blocks are proposed, built, and validated.
Scope management. Over 25 non-headliner EIPs were under consideration for inclusion. Base's engineering team publicly warned that adding FOCIL (Fork-Choice Inclusion Lists) alongside ePBS could push the upgrade beyond 2026 entirely. Core developers responded by deferring FOCIL to the Hegotá fork (H2 2026), keeping Glamsterdam's scope manageable.
Leadership transition. The Ethereum Foundation underwent significant restructuring, with departures including joint executive director Tomasz Stańczak and key contributors Josh Stark and Trent Van Epps. Vitalik Buterin stated the Foundation would become a "smaller ship," but some developers have expressed concern that personnel changes may impact upgrade cadence.
Current status: if the ePBS testnet remains stable, the first general Glamsterdam public testnet launches within weeks, followed by client releases, security audits, and staged testnet deployments before mainnet activation.
Today, 90%+ of Ethereum blocks flow through off-chain relay infrastructure — primarily MEV-Boost, operated by Flashbots. As of April 2026, the builder market is heavily concentrated: Titan captures 52.16% of blocks and BuilderNet handles 24.63%. Two entities construct over three-quarters of all Ethereum blocks.
This architecture works, but it introduces a trust dependency. Validators must trust relays to honestly report block contents and builder payments. The relay layer was never intended to be permanent infrastructure — it was a stopgap while protocol-level solutions were developed.
Vitalik Buterin stated directly: "ePBS only prevents builder centralization from spilling over into the staking layer, but the issue of builder centralization itself remains."
EIP-7732 formally separates the proposer role (selecting the consensus block) from the builder role (assembling the execution payload) at the protocol level. The key mechanics:
This removes MEV-Boost as a systemic dependency. Validators no longer need to trust third-party middleware for block construction. The exchange of block payload for payment becomes a protocol-level guarantee.
Top relays by payload count (April 2026): ultrasound.money (33.92%), titanrelay.xyz (24.19%), bloxroute max-profit (14.67%), aestus.live (10.03%), bloxroute regulated (9.07%). Post-Glamsterdam, these services become optional rather than essential infrastructure.
Since launch, Ethereum has executed transactions sequentially — one after another, in order. Even on modern multi-core hardware, block validation cannot be parallelized because the protocol provides no advance information about which state slots a transaction will access.
EIP-7928 changes this. Each block carries an enforced access list documenting every account and storage slot touched during execution, along with post-execution values. This enables:
According to protocol specifications, the overhead averages ~35 KiB per block at 36M gas, with worst-case sizes remaining below calldata worst-case sizes.
Toni Wahrstätter, lead author of EIP-7928, explained the mechanism: "With explicit mappings, disk I/O + EVM execution + post state root calculations become fully parallelizable. This cuts worst-case block validation latency. Faster block validation = higher throughput for all Ethereum users."
The significance extends beyond raw speed. Lower validation latency means the network can safely increase the gas limit without compromising decentralization — validators with consumer hardware can still keep up.
Many EVM opcode gas costs were set in 2016 and have not been updated to reflect modern hardware capabilities. Storage operations that were expensive on spinning disks now execute on NVMe SSDs. Arithmetic that was costly on 2016 CPUs completes in nanoseconds on current hardware.
EIP-7904 recalibrates gas costs using actual benchmarks from execution clients. The methodology: synthetic blocks that isolate individual EVM operations are generated and timed across multiple client implementations. Raw benchmark data was collected between January 5-22, 2026 using the EEST benchmark suite with Nethermind benchmarking tooling.
The repricing delivers a reported 78.6% reduction in effective gas costs for both simple ETH transfers and complex smart contract interactions — not because the network is subsidizing computation, but because previous prices overcharged relative to actual resource consumption.
Combined with BALs enabling parallel validation, the gas limit target rises from 60 million to 200 million per block. At current transaction sizes, this supports approximately 10,000 TPS on the base layer — roughly 10x current capacity.
For context: Ethereum's gas limit already doubled from 30M to 60M in early 2026. Current block utilization sits at 25-35%, well below the 50% target, with average transaction fees at $0.16-$0.22 (March 2026) versus $1.85 (mid-2025).
Under ePBS, validators (proposers) receive builder bids as guaranteed protocol-level payments rather than relay-mediated transfers. Revenue outcomes depend on builder market competitiveness — proposers rely on external bids rather than vertically integrated strategies.
The structural shift: proposer revenue becomes more predictable and decentralized. Validators no longer need sophisticated MEV extraction capabilities to maximize returns. The upgrade reduces incentives for vertically integrated staking-plus-building operations.
The uncomfortable reality: while ePBS decentralizes the proposer layer, it may intensify builder concentration. Builders with superior order flow, latency advantages, and MEV extraction algorithms maintain their competitive position. The protocol change removes the relay trust assumption but does not address the underlying economies of scale in block construction.
Academic research published in January 2026 (arXiv:2601.12989) confirms this tension: "Although ePBS achieves decentralization at the proposer level, it intensifies centralization among builders and worsens MEV-driven manipulation."
ePBS introduces a new game-theoretic challenge. After a proposer commits to a builder's bid, the builder has a brief window where they can choose not to reveal the block — essentially exercising a "free option" on the bid if market conditions change between commitment and reveal. Protocol designers are aware of this vector but consider it manageable within the current design parameters.
As of mid-2026, Layer 2 transaction costs are already at historical lows:
These levels reflect the cumulative impact of EIP-4844 (blobs, March 2024), the Fusaka upgrade (PeerDAS, December 2025), and the gas limit doubling.
The expanded 9-second propagation window under ePBS directly increases safe blob capacity. More blobs per block mean rollups face less competition for data availability space, further reducing L2 settlement costs.
The gas limit expansion from 60M to 200M provides headroom for increased calldata and blob usage without congestion-driven fee spikes. L2s operating at 95-99% margins (post-blob transactions) may see marginal cost reductions, though the primary beneficiary is throughput capacity rather than per-unit cost.
For rollup operators, the longer propagation window changes sequencer timing assumptions. Batch posting strategies may need recalibration. The shift from relay-dependent block production to protocol-level ePBS also affects L2 sequencers that currently optimize for specific builder relationships.
Builder centralization persists. ePBS solves relay trust, not market structure. Two builders controlling 76%+ of block construction remains an unresolved concern that subsequent upgrades (FOCIL in Hegotá) are designed to address.
Leadership instability. The Ethereum Foundation's restructuring introduces execution risk. Core developers maintain that the technical work proceeds regardless of organizational changes, but coordination overhead increases during transitions.
Gas limit safety. A 3.3x gas limit increase (60M to 200M) has not been tested at scale on mainnet. While benchmarks and testnets provide confidence, emergent behavior under adversarial conditions at production load remains unknown.
Hegotá dependency. FOCIL — the censorship-resistance mechanism removed from Glamsterdam's scope — is essential for the full ePBS security model. Without inclusion lists, builders retain unilateral transaction censorship capability. The gap between Glamsterdam (Q3 2026) and Hegotá (H2 2026) leaves this vulnerability temporarily unaddressed.
Parallel execution edge cases. BALs enable parallelism for non-conflicting transactions, but transactions touching the same state slots still execute sequentially. Under heavy DeFi activity (where many transactions interact with the same liquidity pools), effective parallelism may be lower than theoretical maximum.
Glamsterdam represents Ethereum's most ambitious base-layer performance upgrade since transitioning to proof of stake. The combination of enshrined PBS, parallel execution via block-level access lists, and empirically benchmarked gas costs addresses three long-standing criticisms: relay centralization, sequential bottlenecks, and mispriced computation.
The economic value redistribution is significant. Relay operators lose their privileged position in the block production pipeline. Validators gain protocol-guaranteed builder payments without trust assumptions. Builders retain their market position through skill rather than infrastructure monopoly.
The delay to Q3 2026 reflects engineering discipline rather than dysfunction — ePBS fundamentally restructures Ethereum's slot mechanics, and deploying it prematurely could compromise the network securing $400B+ in total value locked. The scope reduction (dropping FOCIL) is a pragmatic trade-off: ship the throughput gains now, address censorship resistance in six months.
What remains uncertain is whether enshrining PBS at the protocol level solves centralization or merely relocates it. Two builders constructing 76% of blocks is a structural market outcome, not a relay trust problem. Glamsterdam removes the intermediary; it does not change the game theory that produces concentration. That challenge falls to Hegotá and beyond.