Ethereum's Glamsterdam hard fork — the network's largest protocol overhaul since the Merge — entered its final devnet phase in June 2026 and is targeting mainnet activation in Q3-Q4 2026. The upgrade's headline proposal, EIP-7732, enshrines Proposer-Builder Separation (ePBS) directly into the con...
"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 — the network's largest protocol overhaul since the Merge — entered its final devnet phase in June 2026 and is targeting mainnet activation in Q3-Q4 2026. The upgrade's headline proposal, EIP-7732, enshrines Proposer-Builder Separation (ePBS) directly into the consensus layer, eliminating the trusted third-party relays that currently mediate over 90% of Ethereum block production. A second flagship proposal, EIP-7928, introduces Block-Level Access Lists (BALs) that enable parallel transaction execution across CPU cores.
Together with EIP-8037's state-creation gas repricing, these changes unlock a path to a 200 million gas limit — roughly triple the current ~60 million — with a theoretical throughput ceiling near 10,000 TPS. Go-ethereum client v1.17.5, released July 27, 2026, laid the final client-side groundwork. Devnet-6 has been live since June 25, with public testnet forks on Holesky and Hoodi still required before mainnet activation.
The economic implications are significant. Over 95% of Ethereum validators currently depend on MEV-Boost, a sidecar system that routes block construction through a concentrated set of relays and builders. Two builders alone control 95.7% of the MEV-Boost market. Glamsterdam replaces this trust-dependent pipeline with an on-chain auction mechanism, redistributing structural power within the $77.7 billion staked ETH ecosystem.
Ethereum's current block production pipeline relies on an off-protocol system called MEV-Boost, developed by Flashbots. The mechanism separates the roles of block proposers (validators) and block builders (entities that order transactions to capture Maximal Extractable Value). Relays sit between the two, acting as trusted intermediaries.
As of July 27, 2026, relay market share data from relayscan.io shows the following concentration:
| Relay | Market Share | |-------|-------------| | Ultra Sound | 31.75% | | bloXroute Max Profit | 22.04% | | bloXroute Regulated | 18.48% | | Titan | 16.18% | | Aestus | 7.04% |
The top five relays control approximately 95.5% of all block payloads. At the builder layer, concentration is more extreme: two builders control 95.7% of the MEV-Boost market. This structure creates censorship vectors — relays classified as "OFAC-filtering" (including Flashbots, bloXroute Max Profit, and bloXroute Regulated) can exclude transactions at government direction, affecting roughly 60% of block production.
MEV extraction totaled between $851 million and $1.3 billion in 2024-2025, according to Flashbots data. Sandwich attacks alone generated approximately $60 million in trader losses on Ethereum between November 2024 and October 2025, affecting 1.2% of all DEX trades with an average loss of 0.41% per sandwiched trade. Monthly MEV profit fell from $10 million in late 2024 to $2.5 million by October 2025, indicating declining extraction margins even as the structural dependency persists.
EIP-7732 is a consensus-layer-only upgrade that decouples execution validation from consensus validation and moves the proposer-builder auction on-chain. The mechanism works as follows:
This eliminates the relay layer entirely. The roughly 90% of validators currently running MEV-Boost will no longer need to trust independent relay operators or risk builder-relay collusion.
The security properties were formally analyzed in a paper published on arXiv in June 2026 (arXiv:2506.18189) by Luca Zanolini and collaborators, which states and proves the security guarantees of the ePBS mechanism.
A critical limitation, acknowledged by Buterin himself: ePBS prevents builder centralization from bleeding into the validator set, but does not solve builder centralization at its source. The two-builder oligopoly at the construction layer remains an unsolved problem that Glamsterdam does not address.
EIP-7928 introduces Block-Level Access Lists (BALs) — structured records attached to every block that catalog every account and storage slot the block touches, along with post-transaction state diffs.
Historical analysis of Ethereum mainnet data shows that 60-80% of transactions access disjoint storage slots. BALs allow nodes to identify these non-conflicting transactions and distribute them across multiple CPU cores for parallel execution. The remaining 20-40% with overlapping state access can still be parallelized using the post-transaction state diffs included in the BAL.
Additional performance gains from BALs include:
Average BAL size runs approximately 70 KiB per block, according to historical modeling. The overhead is modest relative to the parallelization gains.
The gas limit increase from ~60 million to 200 million — a 3.3x expansion — is made feasible by EIP-8037, which introduces two mechanisms:
Without EIP-8037, tripling the gas limit would proportionally increase state bloat, eventually making full node operation infeasible for consumer hardware. The separate state-growth reservoir decouples throughput from storage, allowing the network to process more transactions without proportional database expansion.
The throughput implications are substantial. At 200 million gas per block, theoretical throughput approaches 10,000 TPS — up from roughly 30 TPS at the current ~60 million gas limit. According to KuCoin research, ETH transfer fees could decline by approximately 71%.
Glamsterdam includes ten EIPs across both the consensus layer ("Gloas") and execution layer ("Amsterdam"):
| EIP | Description | |-----|-------------| | EIP-7708 | ETH transfers and burns emit a log | | EIP-7732 | Enshrined Proposer-Builder Separation (ePBS) | | EIP-7778 | Block gas accounting without refunds | | EIP-7843 | SLOTNUM opcode | | EIP-7928 | Block-Level Access Lists | | EIP-7954 | Max contract size increase (24 KiB → 32 KiB) | | EIP-7975 | eth/70 partial block receipt lists | | EIP-8024 | Backward-compatible SWAPN, DUPN, EXCHANGE opcodes | | EIP-8037 | State-creation gas-cost increase | | EIP-8159 | eth/71 Block Access List Exchange |
Development has progressed through multiple devnet iterations:
Remaining milestones before mainnet: Public testnet forks on Holesky and Hoodi, security reviews, and final client releases. Recent Ethereum hard forks have required two to four months of public-testnet testing.
Timeline estimate: Internal teams target Q3 2026 (as early as late August). Given the public-testnet runway required, September-December 2026 is the more realistic window. Cross-client implementation parity and ePBS complexity are the primary risk factors for delay.
As of July 2026, approximately 41 million ETH ($77.7 billion notional) is staked across roughly 900,000 active validators, representing 34% of circulating supply. Annual validation rewards stand at 2.62%.
Glamsterdam restructures the economic relationship between validators and builders:
Builder centralization persists. ePBS firewalls builder power from the consensus layer but does not reduce the two-builder oligopoly. Post-Glamsterdam, the same builders likely dominate on-chain auctions as they do relay-mediated ones.
Gas repricing side effects. EIP-8037's higher state-creation costs will increase deployment gas for smart contracts that write significant state. DeFi protocols and rollups posting state to L1 may face higher costs, partially offsetting throughput gains for certain use cases.
Cross-client risk. Glamsterdam touches both consensus and execution layers simultaneously across ten EIPs. Historical precedent (the Merge in 2022, Fusaka in 2025) shows that multi-layer upgrades carry elevated coordination risk.
State growth monitoring. The 120 GiB annual ceiling under EIP-8037 is a modeled estimate. Actual state growth under real-world load at 200M gas remains untested at scale.
Glamsterdam addresses a structural dependency that has persisted since the Merge: Ethereum's reliance on off-protocol infrastructure for its most critical function — block production. By enshrining PBS and enabling parallel execution, the upgrade reduces trust assumptions while expanding throughput capacity by approximately 3x.
The economic stakes are measurable. The $77.7 billion staked ETH ecosystem, the $851 million to $1.3 billion annual MEV market, and the relay infrastructure serving 900,000 validators all face structural rearrangement. Whether this rearrangement reduces concentration or merely relocates it from relays to on-chain auctions remains an open empirical question.
The upgrade's technical scope — ten EIPs across both protocol layers — makes it the most complex Ethereum hard fork since the Merge. The development team's track record on Soldøgn devnets and the July 27 Geth release suggest engineering readiness. The question is no longer whether ePBS works, but whether the remaining testnet and security review timeline permits a 2026 activation or pushes into early 2027.