Ethereum's next hard fork, Glamsterdam, entered its first generalized devnet in the final week of April 2026, combining all proposed components into a single test environment for the first time. The upgrade targets a gas limit increase from 60 million to 200 million per block, parallel transactio...
"Replicating Ethereum, scaling it up, and becoming more centralized is meaningless." — Vitalik Buterin, Keynote at Hong Kong Web3 Carnival, April 20, 2026
Ethereum's next hard fork, Glamsterdam, entered its first generalized devnet in the final week of April 2026, combining all proposed components into a single test environment for the first time. The upgrade targets a gas limit increase from 60 million to 200 million per block, parallel transaction execution via Block-Level Access Lists (EIP-7928), and the enshrining of Proposer-Builder Separation directly into the consensus layer (EIP-7732). If delivered on schedule, the fork would increase Layer 1 throughput toward 10,000 transactions per second and cut smart contract gas costs by approximately 78%.
The timeline, however, is under pressure. Originally targeted for H1 2026, the upgrade faces delays driven primarily by the complexity of ePBS implementation. Core developers have acknowledged that Q2 delivery is unlikely, with Q3 now the more realistic window. ETH trades at approximately $2,404 as of April 22, 2026, with a market capitalization of roughly $233 billion.
Glamsterdam is a combined hard fork updating both Ethereum's Execution Layer (Amsterdam) and Consensus Layer (Gloas). It follows the Pectra and Fusaka upgrades shipped in 2025 and represents the most structurally ambitious fork since the September 2022 Merge.
The Ethereum Foundation adopted a "headliner" framework for Glamsterdam, proposed by Tim Beiko and the core-dev process leads: select at most two high-impact EIPs — one per layer — as the upgrade's anchors, then attach lower-risk proposals around them. The two headliners are now locked: EIP-7732 (ePBS) on consensus and EIP-7928 (BALs) on execution.
A total of 22 EIPs are included in the current scope. Over 30 additional proposals were evaluated and declined, including six-second slot times (EIP-7782), multidimensional gas metering (EIP-8011), and post-quantum signature verification. These are expected to target the Hegota fork in H2 2026.
EIP-7732 — Enshrined Proposer-Builder Separation (ePBS). Today, approximately 90% of Ethereum blocks are constructed via MEV-Boost, an out-of-protocol relay system where validators outsource block building to specialized builders through third-party infrastructure. Five builders produce over 80% of blocks, according to Flashbots data. EIP-7732 moves this process on-chain. Under ePBS, the proposer commits to a block header; a separate builder constructs the execution payload. Builders become staked protocol participants with on-chain identity. No external relay is required.
The change eliminates the trust dependency on off-chain relay operators — Ultra Sound relay currently handles 32.3% of MEV-Boost payloads, Titan Relay 24.75%, bloXroute relays a combined ~26%, and Flashbots' own relay just 3.44%, according to relayscan.io data from late 2025. ePBS replaces this fragmented off-chain infrastructure with a single on-chain mechanism.
EIP-7928 — Block-Level Access Lists (BALs). Ethereum currently processes transactions sequentially, even when they touch entirely unrelated parts of state. State access — specifically, random disk reads from a growing database — is the primary execution bottleneck. BALs are structured declarations specifying exactly which storage slots and accounts each transaction will read or write. A hash of this access list (the "BAL root") is embedded directly in the block header.
Gabriel Trintinalia of Consensys described BALs as addressing "the biggest bottleneck we have," explaining that the mechanism enables clients to "preload all the necessary data from disk into memory first, rather than keep going back to read the disk sequentially." With state pre-loaded, non-conflicting transactions can execute in parallel.
EIP-7904 — Gas Repricing. The third significant component realigns gas costs with actual computational resource consumption. Many current gas prices were set years ago and no longer reflect execution on modern hardware. The repricing targets a 78.6% reduction in gas fees for smart contract calls, according to devnet benchmarks.
Ethereum's block gas limit was raised from 30 million to 60 million over the course of 2025, following validator signaling. The increase was automatically applied on November 25, 2025, after more than half of validators — over 513,000 — signaled approval. Ethereum Foundation researcher Toni Wahrstätter called it a "2x increase in a single year."
Glamsterdam targets a further increase to 200 million — a 3.3x jump from the current level. The increase depends on two enabling conditions: parallel execution (via BALs) and gas repricing (via EIP-7904). Without both, a raw gas limit increase would impose unsustainable load on validator nodes.
According to research by ethpandaops, after the gas limit was raised to 60 million, approximately 90% of blocks could be first discovered within 1,016 milliseconds. Block propagation delay increased slightly but remained within acceptable bounds. Based on propagation constraints, the theoretical upper limit estimated on testnets is approximately 150 million — below the 200 million target, suggesting that BAL-enabled parallel execution is essential to bridge the gap.
Vitalik Buterin indicated in late 2025 that future increases would be "more targeted / less uniform" rather than blanket doublings. The longer-term path to even higher gas limits depends on validator adoption of ZK-proof verification. Ethereum Foundation researcher Justin Drake estimates that approximately 10% of validators would need to switch to ZK clients before gas limits could rise substantially beyond 200 million.
Until April 2026, Glamsterdam testing was split across separate networks: an epbs-devnet for ePBS and a bals-devnet for BALs. In the final week of April, the core development team launched the first generalized devnet combining all components — a milestone that integrates EIP-7732, EIP-7928, EIP-7904, and the remaining 19 EIPs into one environment.
According to the Ethereum Foundation's Checkpoint #9 update published April 10, 2026, ePBS implementation is the primary bottleneck. The proposal splits block production into a two-party sequential process within consensus, requiring every part of the stack to handle "partial blocks" and two-party coordination. Client teams (Geth, Besu, Prysm, and others) must achieve interoperability across this new architecture.
The path to mainnet follows a fixed sequence: stable devnet, client releases, security reviews, public testnet deployments (Holesky, then Sepolia), then mainnet fork date announcement. Holesky and Sepolia testnet activation dates will be the clearest signal of an imminent mainnet deployment.
The original H1 2026 target is under pressure. Core developers have acknowledged that Q2 delivery is unlikely. June remains theoretically possible but Q3 2026 is the more realistic expectation. A slip to Q4 cannot be ruled out.
Coinbase's Base engineering team has publicly warned that adding FOCIL (forced inclusion lists) alongside ePBS could push the upgrade beyond 2026 entirely. FOCIL has since been moved to Hegota.
The transition from off-chain to on-chain PBS addresses a structural tension that has persisted since Ethereum's move to proof-of-stake in September 2022. In 2022, Flashbots' relay handled approximately 81% of MEV-Boost payloads. By late 2025, that share had dropped to 3.44% as competitors — Ultra Sound, Titan, bloXroute — absorbed volume. The relay market diversified, but the builder market did not: five builders still produce over 80% of blocks.
This concentration creates what the Flashbots team has described as a "negative-sum orderflow game" — large builders reinvest MEV profits to secure exclusive orderflow, creating a flywheel that locks out smaller participants. MEV-Boost currently lifts validator APY from roughly 4% to approximately 5.69%, according to Flashbots documentation.
Flashbots launched BuilderNet in November 2024 — a decentralized block-builder running in trusted execution environments, jointly operated with Beaverbuild and Nethermind — and ceased operating centralized block builders on Ethereum in December 2024. ePBS takes a different approach: rather than decentralizing the builder through off-chain coordination, it makes PBS a native protocol function with staked participants and on-chain accountability.
The risk, acknowledged by researchers, is that capital-intensive builders may dominate on-chain block construction just as they dominate the current off-chain market, structurally reducing competition within MEV markets regardless of where the mechanism sits.
On February 18, 2026, the Ethereum Foundation reorganized protocol development into three tracks:
Scale — led by Ansgar Dietrichs, Marius van der Wijden, and Raúl Kripalani. Combines the previously separate "Scale L1" and "Scale Blobs" initiatives. Responsible for gas limit increases, execution engine performance, BALs, ePBS, blob parameter expansion, and zkEVM attester client development.
Improve UX — led by Barnabé Monnot and Matt Garnett. Focused on account abstraction, user-facing protocol improvements, and cross-layer coordination.
Harden the L1 — led by Fredrik Svantes, Pari, and Thomas Thiery. Covers security audits, quantum resistance preparation, and the preservation of Ethereum's core properties.
The consolidation under a unified Scale track reflects what the Foundation described as a "practical reality": gas limit increases depend on execution engine performance, blob scaling depends on networking and consensus changes, and both touch the same client code. Separating them created coordination overhead.
The Foundation's February update noted that 2025 was "one of Ethereum's most productive years at the protocol level," having shipped Pectra and Fusaka, doubled blob throughput, raised validator balance caps, and increased the gas limit from 30 million to 60 million. Glamsterdam and Hegota are intended to sustain that pace.
Several significant proposals were excluded from Glamsterdam's scope:
Glamsterdam is a structural overhaul, not an incremental upgrade. It rewrites how Ethereum builds blocks (ePBS), how it executes transactions (BALs), and how it prices computation (EIP-7904). The ambition is proportional to the risk: ePBS introduces two-party coordination into the core consensus loop, and parallel execution via BALs requires ecosystem-wide tooling changes before its benefits fully materialize.
The market context adds urgency. ETH trades at $2,404, down from a peak near $5,000 in August 2025. Layer 2 transaction fees have dropped approximately 95% over two years, but L1 throughput has not kept pace with demand from DeFi, where $292 million was drained from a single bridge exploit on April 18 and $13 billion fled Aave in the aftermath. Ethereum's L1 must process more, cost less, and resist manipulation at the protocol level — Glamsterdam is the attempt to deliver all three simultaneously.
Whether it ships in Q3 or slides to Q4, the fork will define whether Ethereum's 2026 is remembered as a year of execution or a year of delays.