Ethereum's Glamsterdam hard fork — combining the Amsterdam execution-layer upgrade with the Gloas consensus-layer upgrade — has slipped from its original June 2026 target to Q3 2026, according to the Ethereum Foundation's May 11 protocol cluster update. The delay follows completion of the Soldøgn...
"ePBS only prevents builder centralization from spilling over into the staking layer, but the issue of builder centralization itself remains." — Vitalik Buterin, Ethereum Co-Founder
Ethereum's Glamsterdam hard fork — combining the Amsterdam execution-layer upgrade with the Gloas consensus-layer upgrade — has slipped from its original June 2026 target to Q3 2026, according to the Ethereum Foundation's May 11 protocol cluster update. The delay follows completion of the Soldøgn Interop in Svalbard, Norway, where 100+ core contributors locked in a 200-million post-fork gas limit floor (up from ~60 million today), stabilized multi-client ePBS devnets with external builders, and finalized EIP-8037 repricing parameters.
Glamsterdam is the largest Ethereum protocol change since the September 2022 Merge. It enshrines Proposer-Builder Separation (ePBS) directly into consensus, pre-declares block state access via Block-Level Access Lists (BALs), and reprices EVM opcodes — changes projected to reduce gas costs by up to 78% for complex smart contract calls and raise theoretical throughput toward 10,000 TPS. The upgrade also triggers a generational leadership transition: longtime protocol coordinators Tim Beiko and Barnabé Monnot have departed the Ethereum Foundation, replaced by Will Corcoran, Kev Wedderburn, and Fredrik as the new protocol cluster leads.
The Glamsterdam upgrade was originally scoped during the Ethereum Foundation's Checkpoint #8 in January 2026, positioning it as the successor to Fusaka, which activated on mainnet December 3, 2025. A headliner freeze was established in late 2025, with early ePBS and BAL devnets running by Q1 2026.
The critical milestone came during the Soldøgn Interop, held the week of May 2, 2026, in Longyearbyen, Svalbard — 78°N, above the Arctic Circle. Just over 100 core contributors converged for a week of intensive, single-track work sessions that frequently extended past midnight under the Arctic's perpetual daylight.
By the interop's Friday close, three core deliverables were locked:
cost_per_state_byte parameters delivered on bal-devnet-6The Ethereum Foundation's April 2026 Checkpoint #9 acknowledged that "ePBS implementation is proving to be trickier than anticipated." By the May 11 protocol update, the Foundation confirmed Glamsterdam has officially slipped to Q3 2026. Public testnets on Holesky and Sepolia and dual audit phases remain ahead before any mainnet fork date is announced.
The headline technical change is EIP-7732, which moves Proposer-Builder Separation from the current off-chain relay infrastructure into the Ethereum protocol itself.
Under the current system, 96% of all Ethereum blocks are outsourced to independent block builders via the MEV-Boost relay network, according to relay data. The average MEV-Boost block is 5.57x more valuable than a locally built block, creating overwhelming economic pressure for validators to delegate construction. This dependency chain runs through seven unique relay operators controlling 99% of payloads — with relay.ultrasound.money at 33.92% and titanrelay.xyz at 24.19% of relay market share.
ePBS restructures the slot mechanism. The protocol introduces explicit deadlines for block construction, payload reveal, and attestations. A new Payload Timeliness Committee monitors builder compliance. Proposers commit to a builder's bid on-chain; the builder then reveals the execution payload within protocol-enforced timing windows. If a builder fails to deliver, the protocol handles it natively rather than relying on relay fallback logic.
Key infrastructure changes include an extension to the Engine API and introduction of eth/71, a new wire protocol version. Execution clients must store BALs for at least 3,533 epochs.
Glamsterdam addresses one specific vector of centralization but does not resolve builder concentration itself — a point Buterin has publicly acknowledged.
As of April 2026, Titan Builder produces approximately 50% of all Ethereum mainnet blocks (3,528 blocks, 49.97% market share), followed by BuilderNet at 23.36% and Quasar at 16.60%. This concentration traces partly to exclusive order-flow agreements: in April 2023, Titan and Banana Gun entered a private deal where Banana Gun routes nearly all user transaction orders to Titan for block bundling. Titan's market share subsequently increased from under 1% to over 40%.
Titan and Beaverbuild collectively controlled approximately 86% of block production over a two-week period in March 2025, forming what researchers describe as a de facto duopoly. A May 2026 paper published on arXiv (arXiv:2605.04471) examines order-flow exclusivity and value extraction mechanisms driving this consolidation.
ePBS ensures this builder centralization does not infect the validator layer — proposers no longer need to trust relays or maintain relationships with specific builders. But as Buterin noted, toxic MEV (sandwich attacks, front-running) may simply persist elsewhere. The Ethereum Foundation has flagged FOCIL (Fork-Choice Inclusion Lists) and encrypted mempools as subsequent mitigations, with FOCIL scheduled as the consensus-layer headliner for the follow-up Hegotá fork in late 2026.
The second confirmed headliner is EIP-7928, which requires each block to carry a deterministic map of every account and storage slot it will touch before execution begins.
This turns Ethereum execution into a dependency graph. Nodes can verify that a block's declared access list matches its actual state touches, and — in future iterations — use this information to parallelize execution across multiple threads. In the near term, BALs improve state reconstruction speed and reduce node workload by making execution patterns predictable.
BAL implementation progressed through dedicated devnets, separate from ePBS chains, using individual feature flags for isolation testing. A benchmark dashboard and leaderboard were established during the Soldøgn Interop to track client performance. The spec stabilized on bal-devnet-6.
The separate BAL and ePBS devnet tracks were designed to converge into the unified glamsterdam-devnet series, which reached glamsterdam-devnet-2 during the interop week.
Glamsterdam raises Ethereum's per-block gas limit from approximately 60 million to 200 million — a 3.33x increase. This is enabled by ePBS (which gives execution more headroom within restructured slot timing) and BAL optimizations (which reduce redundant state access overhead).
EIP-7904, under consideration for inclusion, realigns gas costs with actual computational resource consumption. During the Soldøgn Interop, developers shifted from dynamic to fixed cost_per_state_byte pricing and moved accounting from mid-execution to opcode level. Additional gas-related EIPs under consideration include:
| EIP | Description | |------|-------------| | EIP-7904 | General EVM opcode repricing using client benchmarks | | EIP-8037 | Raises state creation operation costs to curb excessive contract proliferation | | EIP-8038 | Aligns state-access gas costs with actual disk lookup overhead | | EIP-7954 | Increases maximum smart contract size |
The net result, according to developer estimates: a 78.6% reduction in gas costs for both simple ETH transfers and complex smart contract interactions. At current pricing, Uniswap trades costing $3–$8 could drop below $1.
The 200M gas limit does not displace Layer 2 networks. Even at 3.33x L1 capacity, rollups remain cheaper for cost-sensitive activity. The more likely outcome is a cleaner division of labor: L2s handle execution; L1 serves as the settlement and data-availability anchor.
The upgrade expands data blob capacity — the dedicated data space rollups use to post compressed transaction information back to Ethereum. More blob capacity reduces bidding wars for data availability, lowering L2 settlement costs for Arbitrum, Optimism, Base, and other rollup ecosystems.
Glamsterdam's throughput predictability — enabled by BALs and restructured slot timing — gives L2 sequencers more reliable cost estimates for L1 settlement batches. This is operationally significant for L2 operators managing fee markets.
Ethereum currently has approximately 35.86 million ETH staked (28.91% of total supply) across roughly 1.1 million active validators, earning an average 3.3% APY. The validator queue holds over 3 million ETH waiting to enter, reflecting sustained institutional demand.
Glamsterdam introduces new validator responsibilities. ePBS creates a Payload Timeliness Committee, adding monitoring duties for selected validators. Validators must run updated execution clients supporting the new Engine API extensions and eth/71 wire protocol.
According to Figment protocol analyst Benjamin Thalman, Glamsterdam will "materially improve exit liquidity" for validators — addressing a critical operational concern for institutional stakers managing large positions. Post-Pectra (May 2025), the maximum effective validator balance is 2,048 ETH, meaning large stakers operate fewer, more capital-efficient validators but need reliable exit mechanics.
Glamsterdam arrives alongside a generational change in Ethereum protocol coordination. Tim Beiko, who led All Core Developers Execution (ACDE) calls for years, has departed the Ethereum Foundation. Barnabé Monnot, a key researcher on MEV and mechanism design, has also left. Alex Stokes announced a three-month sabbatical starting after the Soldøgn Interop.
New protocol cluster leads:
| Name | Role | |------|------| | Will Corcoran | Research Coordinator; expertise in zkVM proving, post-quantum consensus | | Kev Wedderburn | Leads zkEVM team; bridges research and engineering | | Fredrik | Leads Protocol Security and the Trillion Dollar Security project |
ACDE co-leads are now Nixo and Ansgar. Pari serves as interim ACDC lead, with Barnabas as interim ACDT lead. This is the most significant coordination reshuffling since the pre-Merge era.
Builder non-delivery asymmetry. Under ePBS, a builder who wins a block slot can fail to deliver the payload. The protocol must handle this "free option" without creating exploitable timing games.
MEV market persistence. ePBS standardizes the proposer-builder handoff but does not eliminate MEV extraction. Capital-intensive builders with exclusive order-flow agreements retain structural advantages.
Cross-client complexity. Glamsterdam touches nearly every layer of the Ethereum stack — consensus timing, execution pricing, state management, and networking (eth/71). Every client team must implement and validate changes in parallel. The Soldøgn Interop achieved 4×3 (four execution clients, three consensus clients) stable configuration, but full multi-client coverage remains incomplete.
BAL correctness requirements. A block declaring an incorrect access list is invalid. Any mismatch between declared and actual state touches produces a consensus failure. This adds a new attack surface that client teams must rigorously test.
Smart contract cost disruption. Gas repricing changes the relative cost of EVM opcodes. Contracts optimized for current gas schedules may see altered execution economics. Developers must audit gas assumptions in production contracts.
Timeline uncertainty. The slip from June to Q3 2026 is the first delay. Given ePBS complexity, further slippage is possible if public testnet phases surface issues.
Glamsterdam represents Ethereum's most ambitious structural overhaul in four years. It rewires how blocks are built, how state is accessed, and how gas is priced. The 200M gas limit target — if achieved — positions Ethereum L1 as meaningfully more competitive on raw throughput while preserving the L2-centric scaling model.
The central tension is that ePBS solves a governance problem (relay dependency) while leaving the economic problem (builder concentration) intact. With one builder controlling half of all block production through private order-flow deals, the censorship resistance and neutrality properties that ePBS is designed to protect remain under pressure from market structure.
The Q3 slip is neither surprising nor alarming — ePBS touches nearly every subsystem in the protocol stack. The Svalbard interop demonstrated that multi-client devnets can run the full spec. The question now is whether public testnet phases and dual audits can be completed without further delays and whether the new generation of protocol coordinators can maintain the delivery cadence established by their predecessors.