Ethereum's Glamsterdam hard fork — the network's largest protocol change since the September 2022 Merge — has entered its final testing phase. The Sepolia testnet fork is scheduled for August 3, 2026, with mainnet activation targeted around September 16. Three headline EIPs define the upgrade: EI...
"The storage redesign is probably the single most disruptive part of the plan." — Vitalik Buterin, Ethereum Co-Founder
Ethereum's Glamsterdam hard fork — the network's largest protocol change since the September 2022 Merge — has entered its final testing phase. The Sepolia testnet fork is scheduled for August 3, 2026, with mainnet activation targeted around September 16. Three headline EIPs define the upgrade: EIP-7732 (Enshrined Proposer-Builder Separation), EIP-7928 (Block-Level Access Lists), and EIP-7904 (gas repricing). Together, they eliminate trusted MEV relays, enable parallel transaction execution, and reduce gas costs by an estimated 78.6%.
The upgrade arrives at a critical juncture. Over 88% of Ethereum blocks are currently built off-chain through MEV-Boost, a relay-dependent system that concentrates block production among three operators. Ethereum's base staking yield has compressed to 2.78-2.82% annually, and the network processes approximately 25 TPS at its current 60 million gas limit. Glamsterdam addresses all three constraints simultaneously, clearing a path toward a 200 million gas limit and an estimated 10,000 TPS throughput ceiling in subsequent upgrades.
Glamsterdam (formally defined under EIP-7773) is the consensus-layer and execution-layer hard fork following Fusaka, which shipped in December 2025. The Soldøgn interop devnet concluded on May 2, 2026. Devnet-4, running multi-client tests with the full slate of planned EIPs, has been operational since mid-June 2026.
The current schedule, according to Ethereum core developer communications:
| Milestone | Target Date | |-----------|-------------| | Soldøgn interop devnet concluded | May 2, 2026 | | Final devnet (devnet-4) multi-client | June 2026 | | Sepolia testnet activation | August 3, 2026 | | Hoodi testnet activation | TBD (August-September) | | Mainnet activation | ~September 16, 2026 |
These dates are aspirational. Given that recent forks required two to four months of public-testnet seasoning, multiple staking infrastructure providers cite a September-to-December 2026 window as the firmer base case.
Vitalik Buterin described the broader effort, termed "Lean Ethereum," as the protocol's "third major iteration" — a multi-year rebuild touching censorship resistance, privacy, and quantum resistance. Glamsterdam is the first hard fork in that sequence, with Hegotá expected to follow in Q4 2026 or Q1 2027.
The headline proposal. Currently, Ethereum validators outsource block construction to specialized builders via MEV-Boost, a trust-based relay infrastructure maintained by Flashbots and a small number of operators. According to Relayscan data from April 2026, three relays dominate payload delivery:
| Relay | Market Share | |-------|-------------| | relay.ultrasound.money | 33.92% | | titanrelay.xyz | 24.19% | | bloxroute.max-profit | 14.67% |
These three operators control approximately 73% of relay traffic. Builders operate without on-chain identity, relays can censor transactions, and validators have no protocol-enforced recourse if a builder fails to deliver a payload.
EIP-7732 enshrines the proposer-builder handoff directly into the consensus protocol. The key mechanical change: the data propagation window expands from 2 seconds to approximately 9 seconds, separating the consensus attestation from execution payload delivery. A new Payload Timeliness Committee (PTC) — a validator committee — attests that the builder's payload arrived on time.
The projected impact: up to 70% reduction in MEV extraction losses, according to Ethereum Foundation estimates. The existing MEV-Boost infrastructure continues to function; the new ePBS market is opt-in, not a forced migration.
For validators, MEV-Boost currently adds 10-30% on top of the base 2.78% APR, according to Figment's Q2 2026 Ethereum Validator Report. A more transparent, protocol-native auction mechanism may improve builder competition, potentially increasing the net amount flowing to stakers versus intermediaries. However, the PTC introduces a new validator duty with tighter timing windows.
Block-Level Access Lists (BALs) introduce a pre-computed map included in every block that identifies every account and storage slot the block will touch, along with post-execution state values. Nodes receive this map before processing, enabling two operational improvements:
At the current gas limit of 60 million, Ethereum processes approximately 25.59 TPS (March 2026 average, per on-chain data). BALs do not directly increase the gas limit — that remains a community governance decision — but they remove the technical constraint blocking safe increases. According to analysis cited by Bitget and Lookonchain, the combination of BALs and ePBS clears a path to a 200 million gas limit floor, roughly tripling current L1 capacity.
An academic paper published on arXiv (June 2026) modeled the relationship between gas limits and throughput: doubling the gas limit to 120 million adds approximately 21 TPS. Reaching 100 TPS on mainnet would require a gas limit of roughly 278 million. The 10,000 TPS figure referenced by multiple sources represents a theoretical upper bound under idealized workloads, not a near-term operational target.
Many of Ethereum's current gas prices were set years ago and no longer reflect execution costs on modern hardware. EIP-7904 introduces systematic, benchmark-driven gas repricing that recalibrates costs against actual resource consumption.
The projected result: a 78.6% reduction in gas costs for both simple ETH transfers and complex smart contract interactions. Unlike previous ad hoc gas adjustments (such as EIP-2929 in the Berlin fork), EIP-7904 applies a methodical approach, measuring real computational overhead on contemporary validator hardware.
This has direct economic implications. Lower gas costs at the same gas limit mean more transactions per block. Combined with BALs enabling parallelism, the effective throughput improvement compounds.
Beyond the three headline proposals, Glamsterdam includes at least eight additional EIPs scheduled for inclusion:
The secondary EIPs reflect a dual strategy: expand capacity (larger contracts, more blobs) while raising costs on state-creation operations that contribute to long-term chain growth. This mirrors Ethereum's ongoing tension between throughput scaling and state sustainability.
According to Figment's Q2 2026 Ethereum Validator Report, covering April-June 2026:
Post-Pectra validator consolidation (which raised the maximum effective balance to 2,048 ETH per validator) has reduced active validator count while improving capital efficiency. The entry queue shows a backlog exceeding 3.5 million ETH with a 62-day wait; the exit queue sits at zero.
Glamsterdam introduces three changes relevant to stakers:
Institutional staking operators (Figment, Coinbase Cloud, Lido node operators) will need to update client software and potentially adjust timing configurations for PTC duties.
The ePBS propagation window expansion from 2 to 9 seconds has a secondary effect: Ethereum can carry more blobs per block. The upgrade expands blob capacity toward 72+ blobs per block — a 24x increase from Dencun's initial parameters (which launched with 3 blobs per block in March 2024).
For rollups that rely on Ethereum for data availability — including Arbitrum, Optimism, Base, and others — more blob space means lower data-posting costs and higher throughput ceilings. Dencun already reduced L2 costs by approximately 90% via blob transactions (EIP-4844). Glamsterdam extends this trajectory.
However, as noted in a June 2026 Crypto Briefing analysis: "Glamsterdam does not make higher throughput automatic. The gains will depend on client performance, validator hardware, gas repricing, and how much parallelism real blocks can actually support."
The upgrade also signals a strategic shift. By tripling potential L1 capacity, Ethereum is partially reversing the "L2-only scaling" narrative that dominated 2023-2025. L1 and L2 scaling are now positioned as complementary, not sequential.
Three factors could delay mainnet activation past September 2026:
ePBS implementation complexity. Enshrining proposer-builder separation into the consensus layer is architecturally significant. Cross-client parity between execution clients (Geth, Nethermind, Besu, Erigon) and consensus clients (Prysm, Lighthouse, Teku, Nimbus, Lodestar) must be validated.
Gas repricing edge cases. A 78.6% reduction in gas costs will change the economics of every deployed smart contract. Contracts with gas-dependent logic (e.g., stipend-based forwarding, gas estimation in DeFi protocols) may behave unexpectedly.
Validator readiness. The PTC introduces new timing constraints. Solo stakers running commodity hardware may face tighter operational requirements. The network's 99.76% participation rate could dip temporarily during the transition.
The 200 million gas limit target is explicitly not part of Glamsterdam itself. It is a subsequent community governance decision that Glamsterdam enables but does not enact.
Glamsterdam is the first concrete delivery in Ethereum's "Lean Ethereum" multi-year rebuild. Its three headline EIPs — ePBS, BALs, and gas repricing — collectively address MEV centralization, throughput constraints, and fee misalignment. The upgrade does not instantly deliver 10,000 TPS or a 200 million gas limit. It removes the technical prerequisites blocking those targets.
The near-term question is execution risk. Eight cross-client implementations must converge on testnet before mainnet activation. The gas repricing alone will require every deployed protocol to verify contract behavior under new cost parameters. Ethereum's track record on fork delivery — the Merge shipped months late, Dencun slipped by weeks — suggests the September 16 target carries meaningful schedule risk.
What the data shows: Ethereum is attempting to rebuild its L1 into a high-performance settlement layer while maintaining backward compatibility. Glamsterdam is the stress test for whether that ambition survives contact with multi-client engineering reality.