Ethereum's next hard fork, Glamsterdam, cleared three engineering milestones during a week-long interop event in Svalbard, Norway that concluded May 2. Over 100 core contributors confirmed: a 200 million gas-limit floor (up from 60 million today), stable multi-client ePBS devnets with external bu...
"Under their coordination, Protocol launched tracks and helped to ship Fusaka to mainnet in December 2025, introducing PeerDAS and raising the mainnet gas limit on the path to 200M and beyond." — Ethereum Foundation, Protocol Cluster Update, May 2026
Ethereum's next hard fork, Glamsterdam, cleared three engineering milestones during a week-long interop event in Svalbard, Norway that concluded May 2. Over 100 core contributors confirmed: a 200 million gas-limit floor (up from 60 million today), stable multi-client ePBS devnets with external builders running end-to-end, and finalized EIP-8037 state-creation repricing. On May 11, the Ethereum Foundation published its Protocol Cluster update confirming these results and announcing a leadership transition — Tim Beiko and Barnabé Monnot will depart, replaced by three new co-leads. Mainnet activation, originally targeted for June, is now expected Q3 2026 by most observers.
The upgrade packages three structural changes into a single fork: enshrined proposer-builder separation (ePBS) via EIP-7732, block-level access lists (BALs) via EIP-7928, and gas repricing via EIP-7904 and EIP-8037. Together they target a 3.3x increase in execution capacity, approximately 10,000 TPS, and a 78.6% reduction in gas costs for both simple transfers and complex contract interactions. The economic implication is significant: if gas fees fall as modeled, Ethereum L1 becomes cost-competitive with several Layer-2 rollups for the first time since Fusaka.
From late April through May 2, over 100 Ethereum core contributors convened in Longyearbyen, Svalbard — above the Arctic Circle — for the Soldøgn Interop. The event, named after the Norwegian term for the midnight sun, produced six successive devnets: bal-devnet-4 through bal-devnet-6 and glamsterdam-devnet-0 through glamsterdam-devnet-2.
According to the Ethereum Foundation's Soldøgn Interop Recap published May 2, every spec change that landed in a devnet was running across client implementations within hours. By Friday of that week, nearly all execution and consensus clients were operating on glamsterdam-devnet-2 with the external builder pipeline tested end-to-end.
Three core deliverables were confirmed complete:
The current MEV pipeline relies on off-protocol relays — third-party intermediaries that sit between block builders and validators. As of early 2026, MEV-Boost handled over 92.5% of Ethereum blocks, according to data from Flashbots. BuilderNet alone produced 25.5% of blocks by January 2026. This concentration creates systemic risk: a relay outage or compromise could disrupt block production across a supermajority of the network.
EIP-7732 moves proposer-builder separation into Ethereum's consensus layer. Under the new design, block builders assemble execution payloads and cryptographically seal their contents. Proposers select the highest-paying sealed block header without seeing or manipulating the transaction order. Payload contents are revealed only after the block is committed, eliminating one vector for sandwich attacks and other MEV extraction strategies that rely on transaction reordering by proposers.
The practical effect: relays become optional rather than load-bearing infrastructure. This does not eliminate MEV — searchers and builders still compete for value — but it removes the trusted intermediary from the critical path. According to research published on arXiv in 2026, the current PBS framework violates three prerequisites for competitive markets: diminishing returns to scale, information symmetry, and low entry barriers. ePBS addresses the intermediary problem but does not directly solve builder concentration.
EIP-7928 introduces Block-Level Access Lists (BALs), which require each block to declare its read/write footprint — the accounts and storage slots it will access — before execution begins. This metadata allows clients to pre-fetch state data and execute non-overlapping transactions in parallel rather than sequentially.
The throughput gain is structural, not incremental. Sequential execution has been Ethereum's default since genesis. BALs enable three concurrent operations:
Combined with the 200 million gas limit, BALs target approximately 10,000 TPS on L1 — up from the current effective throughput of roughly 15-30 TPS depending on transaction complexity. The EIP was discussed extensively at the Fellowship of Ethereum Magicians forum, with developers noting it required minimal consensus-layer changes.
EIP-7904 recalibrates gas costs for EVM opcodes based on benchmarks run against current hardware. Many of Ethereum's gas schedules date to 2016-2020 and no longer reflect actual computation costs. Operations that were expensive on 2016 hardware are now trivial on modern processors, yet their gas prices remained unchanged.
The recalibration produces a 78.6% reduction in gas costs for both simple ETH transfers and complex smart contract calls, according to the upgrade's specification documents. In practical terms: a Uniswap trade that currently costs $3–8 in gas could fall below $1. Multi-step DeFi transactions involving multiple contract calls would see proportionally larger savings.
This repricing is bundled with approximately 10 individual gas-cost EIPs, collectively tracked under the Glamsterdam repricing package. The Soldøgn interop locked final numbers for these EIPs, closing an extended calibration process that began in late 2025.
A higher gas limit without guardrails would accelerate Ethereum's state growth — the total size of data that full nodes must store. At a 60 million gas limit, daily state growth runs approximately 349 MiB, or 124 GiB per year, according to analysis published on Ethereum Research. At 100 million gas, that rises to 553 MiB daily and 197 GiB per year. At the new 200 million target, unmitigated state growth would exceed node storage thresholds within roughly two years.
EIP-8037 addresses this by raising the gas cost of state-creation operations — deploying new contracts and opening new storage slots. The proposal uses a fixed cost_per_state_byte structure designed to target state growth of approximately 60 GiB per year even at a 300 million gas block limit. Contract deployment costs rise roughly 10x and new account creation approximately 8.5x under the new pricing.
The design explicitly decouples state creation from execution gas. Until Glamsterdam, both operations shared the same gas resource, meaning any increase in compute capacity automatically increased the rate of new state. Post-Glamsterdam, scaling execution does not proportionally scale state bloat.
The spec initially carried dynamic per-state-byte pricing tied to the block gas limit, but teams agreed to drop dynamic pricing in favor of fixed costs during the Soldøgn interop, with future repricing handled at fork boundaries.
The May 11 Protocol Cluster Update confirmed that two of Ethereum's most prominent protocol coordinators will leave the Foundation. Tim Beiko, who led developers through EIP-1559, the Merge, Pectra, and Fusaka, is departing. Barnabé Monnot, who spent over six years at the Foundation working on mechanism design and economic analysis, is also leaving. Researcher Alex Stokes is taking a sabbatical.
Three new co-leads have been named:
The leadership transition coincides with a period of accelerating protocol ambition. The Foundation is simultaneously advancing Glamsterdam toward mainnet, scoping Hegotá features, and maintaining the long-term Strawmap roadmap.
Glamsterdam's scope was deliberately narrowed to ensure deliverability. Three significant features were deferred to Hegotá, Ethereum's second planned 2026 hard fork (named by blending the Bogotá Devcon venue with the star Heze), now targeted for late 2026:
The decision to move FOCIL was notable. While ePBS restructures how blocks are built, FOCIL addresses whether transactions can be censored. Together they form a two-part response to centralization concerns in Ethereum's block production pipeline, but the Foundation chose to ship them sequentially rather than risk delaying both.
The economic logic of Glamsterdam follows a pattern identified in earlier Ethereum upgrades: reduce unit costs to increase volume. If gas fees drop 78.6% as modeled, Ethereum L1 becomes viable for transaction types that have migrated to Layer-2 networks over the past two years.
This creates a direct competitive dynamic. L2 rollups — including Optimism, Arbitrum, Base, and others — justified their existence partly through lower fees than L1. A 3.3x capacity increase with parallel execution narrows that gap. L2s retain advantages in specialized execution environments and application-specific optimizations, but the fee differential that drove initial adoption compresses.
For validators, ePBS restructures revenue flows. Relay operators lose a privileged position in the MEV supply chain. Builders still compete, but do so within a protocol-defined auction rather than a trust-dependent off-chain marketplace. The net effect on validator revenue is uncertain — ePBS may increase total MEV capture efficiency while redistributing value away from relay operators.
State-creation repricing (EIP-8037) introduces a new cost consideration for developers. Contract deployments become significantly more expensive, which may accelerate the trend toward proxy patterns and minimal-deployment architectures already prevalent in DeFi.
Glamsterdam represents the largest single increase in Ethereum L1 execution capacity since genesis. The engineering work is substantially complete — the remaining timeline risk is testing and client coordination, not design uncertainty. The leadership transition at the Foundation adds organizational risk, though the incoming co-leads have deep technical backgrounds.
The more consequential question is economic: whether cheaper L1 execution draws activity back from L2s or simply expands the total addressable market. Historical precedent from Fusaka (December 2025) suggests both effects occur simultaneously — L2 fees fell but L2 usage also increased as lower costs attracted new users. Glamsterdam's 3.3x capacity increase is larger in magnitude than Fusaka's blob expansion, and the introduction of parallel execution via BALs is architecturally distinct from prior scaling approaches.
The data from Soldøgn devnets is encouraging but limited. Multi-client stability on a six-devnet sequence does not guarantee mainnet readiness. The Q3 timeline allows for additional testing rounds, and the Foundation's decision to defer FOCIL and Verkle Trees to Hegotá suggests a preference for conservative scoping over ambitious timelines.