Ethereum's Glamsterdam hard fork — the network's most extensive protocol overhaul since The Merge in September 2022 — has entered its final devnet phase as of mid-June 2026, with Devnet-5 achieving approximately 95% participation across client teams. The upgrade ships ten Ethereum Improvement Pro...
"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 most extensive protocol overhaul since The Merge in September 2022 — has entered its final devnet phase as of mid-June 2026, with Devnet-5 achieving approximately 95% participation across client teams. The upgrade ships ten Ethereum Improvement Proposals tracked under Meta EIP-7773, headlined by EIP-7732 (Enshrined Proposer-Builder Separation) and EIP-7928 (Block-Level Access Lists), and clears the path for a 200 million gas-limit floor that would roughly triple current Layer 1 capacity.
The fork targets mainnet activation in Q3 2026, with an optimistic internal window around late August and a firmer base case of September to December, pending two to four months of public-testnet seasoning on Holesky and Hoodi. At its core, Glamsterdam addresses a structural dependency that has quietly defined Ethereum's post-Merge era: 80–90% of blocks currently route through off-chain MEV-Boost relays operated by a handful of entities, with a single builder — Titan — assembling 47.76% of all blocks as of June 16, 2026.
The economic stakes are substantial. Ethereum's average gas price has fallen to 0.052 Gwei in April 2026 from 3.146 Gwei a year prior, compressing fee revenue while L1 transaction costs sit at $0.16–$0.22. Glamsterdam's gas repricing (EIP-7904) is projected to cut complex smart-contract interaction fees by up to 78%, while BALs enable parallel transaction execution that could push throughput toward 10,000 TPS-equivalent under realistic workloads. Whether this generates meaningful new economic activity — or merely accelerates L1 fee compression — remains the central question.
Since The Merge, Ethereum has operated with an informal but deeply entrenched division of labor: validators propose blocks, but the actual assembly of those blocks — the ordering of transactions that determines who captures Maximal Extractable Value (MEV) — has been outsourced to an off-chain system called MEV-Boost, developed by Flashbots.
The numbers paint a concentrated picture. As of June 16, 2026, relay market share breaks down as follows:
| Relay | Payload Share | |---|---| | relay.ultrasound.money | 35.49% | | titanrelay.xyz | 22.94% | | bloxroute (max profit) | 13.85% | | bloxroute (regulated) | 12.24% | | aestus | 8.39% |
On the builder side, concentration is sharper. Titan Builder controls 47.76% of blocks, followed by Quasar at 19.99% and beaverbuild.org at 16.36%. Five builders control 96.7% of all MEV blocks. The Herfindahl-Hirschman Index (HHI) sits at approximately 2,140, placing the builder market in "moderately concentrated" territory — comparable to the U.S. airline industry before its merger wave.
This concentration creates three distinct risks: censorship (regulated relays can filter transactions), single-point-of-failure fragility (loss of a dominant builder disrupts block production), and value extraction asymmetry (a small number of builders capture disproportionate MEV revenue). An academic paper published in January 2026 (arXiv:2601.12989) found that under ePBS simulations, the Gini coefficient for builder profits rises from 0.1749 under standard PoS to 0.8358, indicating that while ePBS solves the validator-layer centralization problem, builder-layer concentration may intensify.
EIP-7732 moves the proposer-builder handoff from MEV-Boost's trust-based middleware directly into Ethereum's consensus layer. The mechanical change is straightforward but consequential:
Before (MEV-Boost): Builders assemble blocks → submit bids to relays → relays verify bids and forward headers to proposers → proposers select highest-paying header blind → relays reveal full block after commitment.
After (ePBS): Builders assemble blocks and cryptographically seal contents → submit bids on-chain → proposers select highest bid without seeing contents → protocol enforces payload delivery and payment atomically.
The key addition is the Payload Timeliness Committee (PTC): a randomly selected committee of exactly 512 validators per slot whose sole job is to attest that the builder's payload arrived on time and matches the commitment. PTC members do not execute the block. They verify delivery timing only, creating a dual-deadline logic that separates consensus-block attestation from execution-payload attestation.
This eliminates the trusted intermediary role that relays currently play. Validators no longer need to trust a relay to honestly report a builder's bid or faithfully deliver the block. The protocol handles both enforcement and settlement.
The practical effect: MEV-Boost becomes optional rather than essential. Validators can still use external relays if they prefer, but the protocol now provides a native, trustless alternative for builder-proposer coordination.
The second headline EIP addresses execution throughput. EIP-7928 introduces Block-Level Access Lists (BALs), which allow blocks to declare in advance which accounts and smart-contract storage slots they will access. This pre-declaration enables something Ethereum has never had: parallel transaction execution within a single block.
Current Ethereum execution is strictly sequential. Each transaction must complete before the next begins, because the EVM cannot know in advance whether two transactions touch the same state. BALs solve this by making state access explicit at the block level. If two transactions access disjoint state, the client can execute them simultaneously.
The throughput implications are significant. Combined with the 200M gas-limit target, BALs could enable throughput approaching 10,000 TPS-equivalent under realistic workloads — a figure that would put Ethereum L1 in the same ballpark as Solana's practical throughput, though achieved through a fundamentally different architectural approach (parallel EVM execution versus Solana's Sealevel runtime).
BALs also enable "executionless sync," where nodes can preload required data more efficiently, reducing sync times and making it more feasible for consumer hardware to run full nodes — a persistent concern as state growth accelerates.
Glamsterdam does not enforce a specific gas limit. The 200 million figure is the design target that the upgrade's technical changes — particularly BALs and ePBS — are engineered to support safely. The actual limit is set by validator consensus through standard gas-vote signaling, currently coordinated around 60 million.
The path from 60M to 200M is incremental. Validators would step the limit up progressively as node operators demonstrate that larger blocks propagate without degradation. This is the same process that raised the gas limit from 30M to 60M in 2025, but the jump to 200M is substantially larger and introduces new propagation challenges.
At 200 million gas, L1 capacity roughly triples. Fee reduction projections are substantial: EIP-7904's gas repricing is expected to slash costs for complex smart-contract interactions by up to 78%. Combined with Ethereum's already low gas prices — 0.052 Gwei in April 2026, down from 3.146 Gwei a year earlier — this raises questions about the economic sustainability of validator fee revenue in a post-Glamsterdam environment.
The ten EIPs in Glamsterdam span consensus, execution, and economic layers:
Consensus Layer:
Execution Layer:
Additional EIPs:
The gas repricing EIPs work in tandem. While EIP-7904 reduces execution costs for legitimate use cases, EIPs 8037, 7976, and 7981 increase costs for specific operations that have historically been underpriced — state creation, calldata storage, and access-list usage. The net effect is a recalibration of Ethereum's internal cost model rather than a blanket fee reduction.
As of June 2026, Glamsterdam has progressed through multiple devnet iterations:
Ethereum Foundation developer Parithosh Jayanthi stated that teams are working on devnets with "all the EIPs in them right now" and that developers have made "massive progress," though "no fixed timeline" is set.
Projected timeline:
| Phase | Estimated Date | |---|---| | Final devnet completion | July 2026 | | Holesky testnet fork | July–August 2026 | | Hoodi testnet fork | August 2026 | | Mainnet activation (optimistic) | Late August 2026 | | Mainnet activation (base case) | September–December 2026 |
Ethereum Foundation contributors have noted that Glamsterdam is "proving trickier and slower" than the preceding Fusaka fork. The stated priority: getting ePBS right outranks any fixed date.
The subsequent upgrade — Hegota — is expected in late 2026 or early 2027, addressing state growth, node sustainability, and censorship resistance. This two-fork cadence reflects Ethereum's strategic shift toward smaller, more frequent upgrades rather than bundled, high-risk releases.
Glamsterdam's economic impact operates across multiple vectors:
Validator Economics: ePBS changes how validators earn MEV revenue. Under MEV-Boost, validators passively select the highest relay bid. Under ePBS, the same auction occurs on-chain, but the PTC attestation duty adds a new obligation. MEV-Boost currently adds 10–30% to base validator rewards, contributing roughly 0.5–1% additional staking yield. Whether ePBS preserves, increases, or reduces this premium depends on builder competition dynamics in the on-chain auction.
Fee Revenue: L1 transaction fees averaged $0.16–$0.22 in Q1 2026, already near historic lows. A further 78% reduction in complex-interaction costs could push per-transaction revenue below $0.05 for many use cases. The bet is that dramatically cheaper L1 execution attracts sufficient new activity — DeFi protocols, on-chain computation, applications currently priced out of L1 — to offset per-unit fee compression.
Builder Market: The academic evidence is ambiguous. On-chain builder auctions remove relay trust requirements but may intensify builder concentration. The January 2026 arXiv paper found that ePBS auctions create winner-take-most dynamics for builders with superior MEV extraction capabilities. Vitalik Buterin has acknowledged this, stating that ePBS addresses validator centralization but not builder centralization.
L2 Economics: Layer 2 rollups currently pay for L1 data availability via blob fees (post-Dencun). A tripled L1 gas limit does not directly reduce L2 data costs but may shift the competitive landscape. If L1 becomes cheap enough for certain applications, some activity that currently routes through L2s for cost reasons may migrate back to L1 — a dynamic that would compress L2 revenue while boosting L1 fee income.
Execution risk: ePBS is the most complex consensus-layer change since The Merge. Multi-client interoperability across all EVM clients must hold under adversarial conditions, not just cooperative devnet environments.
Block propagation: At 200M gas, blocks are significantly larger. If propagation times increase, this advantages geographically proximate validators and large staking pools, potentially undermining the decentralization objectives that ePBS aims to serve.
Builder centralization persistence: ePBS solves the relay trust problem but may not address the economic forces driving builder concentration. The top three builders currently control 84% of blocks. If this concentration persists or worsens post-ePBS, the censorship and fragility risks merely shift from the relay layer to the builder layer.
Timeline slippage: The upgrade has already slipped from its original H1 2026 target. Further delays are possible given the complexity of ePBS implementation and the Ethereum Foundation's stated preference for correctness over speed. The Foundation's recent 40% budget cut and 20% workforce reduction add resource constraints to the development timeline.
Glamsterdam represents Ethereum's attempt to resolve a three-year-old structural compromise: the reliance on off-chain middleware for block construction that emerged as a pragmatic workaround after The Merge. The upgrade's ambitions are considerable — trustless proposer-builder separation, parallel execution, and a path to tripled capacity — but the execution risk matches the scope. The fork has already slipped from H1 to H2 2026, the Ethereum Foundation is operating with a reduced workforce and budget, and the academic evidence suggests that the builder-centralization problem ePBS targets may merely shift rather than resolve. Whether Glamsterdam delivers on its capacity promises while maintaining Ethereum's decentralization guarantees will become clearer in the months of public-testnet operation ahead. For now, the data shows a network engineering its way toward structural improvements while confronting the limits of what protocol-layer changes can achieve against entrenched economic incentives.