Ethereum's Glamsterdam hard fork — the network's most structurally significant upgrade since the Merge — entered generalized devnet testing in late April 2026 after months of fragmented component trials. The upgrade bundles Enshrined Proposer-Builder Separation (EIP-7732) and Block-Level Access L...
"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 structurally significant upgrade since the Merge — entered generalized devnet testing in late April 2026 after months of fragmented component trials. The upgrade bundles Enshrined Proposer-Builder Separation (EIP-7732) and Block-Level Access Lists (EIP-7928) into a single fork targeting a 78% reduction in gas fees, a gas limit increase from 60 million to 200 million per block, and a theoretical throughput path toward 10,000 transactions per second.
Current Ethereum L1 throughput sits at approximately 23 TPS. Average transaction costs have fallen to $0.23 as of May 5, 2026, with base fees around 0.05–0.12 gwei — already historically low due to Fusaka's blob expansion and Layer 2 migration. Glamsterdam aims to push execution capacity further by enabling parallel transaction processing, while simultaneously addressing the protocol's most persistent centralization risk: the block builder duopoly that currently produces over 80% of all Ethereum blocks via off-chain MEV-Boost relays.
The Ethereum Foundation's Checkpoint #9, published April 10, described progress as "slow but steady," with ePBS identified as the primary bottleneck. A June 2026 mainnet target remains aspirational; Q3 2026 is the realistic deployment window.
Glamsterdam is a portmanteau of Gloas (consensus layer) and Amsterdam (execution layer), following Ethereum's convention of pairing dual-layer upgrades under a single name. It succeeds Fusaka (December 2025), which introduced PeerDAS and expanded blob capacity to 14 target / 21 maximum.
Two EIPs are confirmed for inclusion:
| EIP | Name | Function | |-----|------|----------| | 7732 | Enshrined Proposer-Builder Separation (ePBS) | Moves proposer-builder coordination into the protocol | | 7928 | Block-Level Access Lists (BALs) | Declares block state access upfront for parallel execution |
Four additional proposals remain under consideration:
| EIP | Title | Purpose | |-----|-------|---------| | 7904 | General Repricing | Realigns EVM opcode costs with modern hardware benchmarks | | 8037 | State Creation Gas Cost Increase | Raises account/storage slot creation costs | | 8038 | State-Access Gas Cost Increase | Raises cold read gas costs | | 7954 | Increase Maximum Contract Size | Raises bytecode size limits |
More than 30 proposals were explicitly excluded from scope, including FOCIL inclusion lists (deferred to Hegotá), reduced slot times, multidimensional gas metering, and post-quantum signature verification.
The current Ethereum block production pipeline relies on an off-chain infrastructure layer. Approximately 90% of all Ethereum blocks are produced through MEV-Boost, a middleware system where validators outsource block construction to specialized builders who compete to assemble the most profitable block. Validators then select the highest-bidding builder's block without inspecting its contents.
This system functions but introduces trust assumptions. Validators must trust relay operators to faithfully deliver builder payloads. Builders must trust that their blocks will not be stolen or front-run by proposers. The entire exchange happens outside protocol consensus.
EIP-7732 moves this exchange on-chain. Under ePBS:
The stated goal is to eliminate the need for third-party relay infrastructure while preserving the competitive builder marketplace that has reduced MEV extraction costs for users.
According to Tomasz Stańczak, former Ethereum Foundation co-executive director, the gas limit will be phased — reaching 100 million per block initially and scaling to 200 million once ePBS is fully operational.
EIP-7928 addresses a fundamental architectural limitation: Ethereum currently processes transactions sequentially. Every transaction in a block executes one after another, regardless of whether they interact with the same state. A Uniswap trade touching pool X and an Aave deposit touching market Y are processed in series despite having zero state dependency.
Block-Level Access Lists (BALs) require that each block declare upfront which accounts and storage slots its transactions will touch. This declaration is committed in the block header and verified against the actual execution trace. Transactions touching non-overlapping state can then execute in parallel.
Technical requirements include:
The combination of BALs with ePBS-enabled higher gas limits is projected to deliver a 78.6% reduction in gas costs across both simple transfers and complex smart contract interactions. According to developer estimates, Uniswap swaps could fall from the current $3–8 range to below $1.
The data on builder market concentration is unambiguous. As of April 2026, according to relayscan.io:
The top three builders control more than 91% of all PBS blocks. This represents a structural duopoly-turned-triopoly, with Titan alone commanding a majority.
Relay distribution tells a similar story. The ultrasound.money relay carries 33.92% of payloads, titanrelay.xyz handles 24.19%, and bloxroute.max-profit.blxrbdn.com processes 14.67%.
ePBS addresses one specific vector: it prevents builder centralization from contaminating the staking layer. By enshrining the builder-proposer separation in protocol, validators no longer need to trust relay operators, and the risk of relay-level censorship or manipulation is eliminated.
However, academic modeling cited in Bitfinex's technical analysis indicates ePBS may "significantly amplify profit and content centralisation" among builders themselves. The reasoning: private order flow access creates structural bidding advantages favoring large-scale operators. Over 50% of high-value Ethereum transactions are already routed through private channels (as of mid-2025) specifically to avoid MEV extraction — and this private flow disproportionately benefits the largest builders.
Buterin has acknowledged this limitation explicitly, noting that the block building pipeline requires additional interventions beyond ePBS — specifically "big FOCIL" (forced inclusion lists) and encrypted mempools — to address censorship resistance and extraction fairness.
Devnet Progression:
Until April 2026, testing was fragmented across separate networks — one for ePBS (epbs-devnet) and one for BALs (bals-devnet). The generalized devnet, launched in the final week of April, marked the first moment all Glamsterdam components coexist in a single environment.
The Ethereum Foundation's DevOps team has tested three of the proposed EIPs on Devnet-4, with Devnet-5 in progress. Public testnet activations (Holesky and Sepolia) are expected in the months leading up to mainnet activation.
Timeline Assessment:
| Milestone | Date | Status | |-----------|------|--------| | Component devnets (separate) | Q1 2026 | Complete | | Generalized devnet (merged) | Late April 2026 | Active | | Public testnets (Holesky, Sepolia) | Q2 2026 | Pending | | Mainnet activation (aspirational) | June 2026 | At risk | | Mainnet activation (realistic) | Q3 2026 | Likely |
The primary bottleneck is ePBS coordination complexity. As the Ethereum Foundation's Checkpoint #9 noted: "The protocol now has to handle disagreement or failure between [proposer and builder] — every part of the stack has to reason about 'partial blocks' and two-party coordination, a change that touches practically everything."
The "free option" problem — where builders can withhold a payload after winning a bid, affecting approximately 0.82% of blocks on average and rising to ~6% during volatility — requires careful protocol-level mitigation.
Current state (May 2026):
Post-Glamsterdam projections:
ETH market context:
The fee reduction is a double-edged sword for ETH's monetary policy. Lower fees mean less ETH burned under EIP-1559, potentially undermining the "ultrasound money" narrative that briefly held during high-activity periods. However, Fusaka's fee floor mechanism (introduced in January 2026) provides a baseline burn rate regardless of demand.
Glamsterdam represents a structural reorganization of how Ethereum builds blocks — not merely a parameter tweak. By enshrining the proposer-builder relationship in protocol consensus and enabling parallel execution through declared state access, the upgrade attempts to simultaneously scale throughput and reduce trust assumptions.
The trade-off is complexity. Two-party coordination inside consensus introduces failure modes that do not exist in the current relay-based system. The devnet phase will determine whether these failure modes are manageable or whether the June timeline slips further.
The builder centralization question remains open. ePBS isolates the problem — preventing staking centralization from tracking builder centralization — but does not solve it. With Titan building a majority of blocks and private order flow structurally advantaging large operators, the economic dynamics of block construction may prove more resistant to protocol intervention than the Ethereum research community anticipates. Hegotá's FOCIL mechanism is positioned as the next intervention point, but that is at minimum six months away.
For now, the data is clear: Ethereum's base layer is being rebuilt in production, one devnet at a time.