Ethereum's next hard fork, Glamsterdam, entered its first generalized devnet in the final week of April 2026 after months of component-level testing. The upgrade packages two structural changes — Enshrined Proposer-Builder Separation (ePBS, EIP-7732) and Block-Level Access Lists (BALs, EIP-7928) ...
"Every compromise of values that Ethereum has made up to this point — every moment where you might have been thinking, is it really worth diluting ourselves so much in the name of mainstream adoption — we are making that compromise no longer." — Vitalik Buterin, Ethereum Co-Founder
Ethereum's next hard fork, Glamsterdam, entered its first generalized devnet in the final week of April 2026 after months of component-level testing. The upgrade packages two structural changes — Enshrined Proposer-Builder Separation (ePBS, EIP-7732) and Block-Level Access Lists (BALs, EIP-7928) — alongside benchmarked gas repricing (EIP-7904). Together, they target a tenfold throughput increase to approximately 10,000 TPS on the base layer, a 78.6% reduction in gas costs for standard operations, and the elimination of Ethereum's current 90%+ dependence on third-party MEV-Boost relays for block construction.
The upgrade was initially slated for Q2 2026, with June as a tentative mainnet date. According to the Ethereum Foundation's Checkpoint #9 report published April 10, 2026, a Q2 activation is now "unlikely." The ePBS component — which fundamentally restructures how blocks are produced inside consensus — is proving more complex than anticipated. Public testnet activations on Holesky and Sepolia have not yet been scheduled. A slip into Q3 or later remains the base case.
The economic stakes are material. Ethereum processed a record 200.4 million base-layer transactions in Q1 2026. Three block builders — Titan (57%), BuilderNet (23%), and Quasar (13%) — control approximately 93% of block production. Glamsterdam's ePBS would move this process on-chain, reducing the centralization risk that Buterin himself identified as "serious backsliding" in a March 2026 address.
Glamsterdam simultaneously updates Ethereum's Execution Layer (Amsterdam) and Consensus Layer (Gloas). It is the first upgrade since The Merge in September 2022 to restructure how blocks are physically built and validated at the protocol level, rather than adding features on top of the existing architecture.
The upgrade contains three headliner EIPs:
| EIP | Function | Layer | |-----|----------|-------| | EIP-7732 | Enshrined Proposer-Builder Separation | Consensus | | EIP-7928 | Block-Level Access Lists | Execution | | EIP-7904 | Benchmarked Gas Repricing | Execution |
Additional proposals bundled under the gas repricing umbrella include EIP-8007 (repricing bundle), EIP-7954 (maximum contract size increase), EIP-8011 (multidimensional gas metering), and EIP-8032 (size-based storage gas pricing). The current gas limit baseline target stands at 60 million, with testing ongoing at higher limits to determine the implications.
Today, Ethereum validators outsource block construction to specialized builders via an off-protocol system called MEV-Boost. According to relayscan.io data from April 2026, the relay market is dominated by a handful of operators: ultrasound.money (33.9%), Titan relay (24.2%), and BloXroute (23.7% combined across two relays). Approximately 90% of Ethereum blocks are produced through these third-party relays.
EIP-7732 moves the proposer-builder separation mechanism directly into the consensus protocol. Under the new design:
This eliminates the trust assumption currently placed in relay operators and reduces the attack surface for MEV extraction via front-running and sandwich attacks. The proposer no longer has the ability to inspect or reorder transactions, closing a structural vulnerability that has persisted since The Merge.
The implementation complexity is substantial. According to the Ethereum Foundation's Checkpoint #9, ePBS "splits block production into two parties acting in sequence inside consensus instead of the way it currently happens outside of the protocol, so the protocol now has to handle disagreement or failure between them." The upgrade touches "practically everything" in the consensus layer.
The second headliner addresses Ethereum's single-threaded execution model. Currently, Ethereum processes transactions sequentially because it cannot predict which accounts and storage slots a given transaction will access until it executes.
EIP-7928 introduces a new field in the block header — block_access_list_hash — containing the Keccak-256 hash of all accounts and storage locations accessed during block execution, along with their post-execution values. This enables:
Historical analysis of Ethereum mainnet data shows that 60-80% of transactions access disjoint storage slots, enabling effective parallelization without conflict. The remaining 20-40% can still be parallelized using post-transaction state diffs included in the BAL.
Average BAL size based on historical data is approximately 70 KiB per block. This overhead is modest relative to the throughput gains: combined with ePBS, the target is approximately 10,000 TPS at the base layer, up from an effective rate of roughly 1,000 TPS today.
The Ethereum Foundation describes BALs as "a fundamental rethinking of how gas and state access work." The BAL-specific devnets have been making "predictable progress through these expectedly difficult problems," per Checkpoint #9.
Many of Ethereum's current gas prices for EVM opcodes were set years ago and no longer reflect actual computational costs on modern hardware. EIP-7904 recalibrates gas costs using empirical benchmarks.
The result: a 78.6% reduction in gas fees for both simple ETH transfers and complex smart contract interactions. The repricing also includes companion proposals — EIP-8011 for multidimensional gas metering (separating compute, storage, and bandwidth costs) and EIP-8032 for size-based storage gas pricing.
For context, Ethereum mainnet fees have already dropped approximately 95% from their peak to roughly $0.01 per transaction in early 2026, driven primarily by L2 migration. Glamsterdam's repricing would compress L1 costs further, potentially narrowing the cost differential between L1 and L2 execution for certain transaction types.
As of the third week of April 2026, the Glamsterdam development pipeline stands as follows:
| Milestone | Status | |-----------|--------| | Component devnets (ePBS, BALs separate) | Completed | | First generalized devnet (all components combined) | Launched ~April 20, 2026 | | Public testnet (Holesky) | Not yet scheduled | | Public testnet (Sepolia) | Not yet scheduled | | Mainnet activation | Original target: June 2026 | | Revised estimate | Q3 2026 or later |
The Ethereum Foundation's April 10 Checkpoint #9 stated that ePBS implementation "is proving to be trickier than anticipated." The report's author assessed that a Q2 mainnet launch is "unlikely." The path to mainnet requires: devnet stabilization, client releases, security audits, and public testnet rehearsals on Holesky and Sepolia — a sequence that typically consumes 8-12 weeks minimum.
Meanwhile, planning for the following upgrade — Hegotá — has already begun. FOCIL (EIP-7805), a fork-choice enforced inclusion list mechanism, has been selected as the consensus layer headliner. Non-headlining feature proposals opened on April 9, 2026, with a deadline to be announced at least two weeks in advance.
The economic rationale for ePBS is rooted in measurable centralization. As of April 2026, Ethereum's block building market is controlled by three entities:
| Builder | Market Share | |---------|-------------| | Titan Builder | ~57% | | BuilderNet (Flashbots) | ~23% | | Quasar Builder | ~13% | | All others | ~7% |
This represents a de facto oligopoly. In March 2025, the market was a duopoly: Beaverbuild and Titan Builder collectively produced approximately 86% of mainnet blocks. BuilderNet's introduction partially distributed this concentration, but the structural dependency on off-protocol infrastructure remains.
Under ePBS, block construction moves on-chain. Builders compete in a protocol-native auction. Proposers cannot inspect sealed block contents. The relay layer — currently the single point of failure between builders and proposers — becomes unnecessary. This is not a marginal improvement; it is a structural delayering of Ethereum's block production supply chain.
Glamsterdam does not displace Layer 2 networks. Even at the projected 10,000 TPS, L2s remain cheaper for cost-sensitive activity and offer sub-second finality that L1 cannot match. The more likely outcome is a cleaner division of labor.
Current L2 economics illustrate the dynamic. L2 sequencers collectively earned approximately $161 million annually as of early 2026 (Base: $93M, Arbitrum: $42M, Optimism: $26M). Of this, only a fraction returns to L1 as data availability fees — Base paid roughly $4.9 million to L1 in blob fees in 2024, a 5% capture rate.
Glamsterdam affects rollups in two ways:
L2 networks currently account for approximately 95% of Ethereum's total transaction throughput, processing roughly 2 million daily transactions. Glamsterdam reinforces rather than challenges this architecture.
Viewed through an economic value distribution lens, Glamsterdam reshapes how fees flow between network participants.
Validators lose MEV-Boost relay dependence but retain proposer revenue through ePBS auctions. The mechanism changes from off-protocol relay selection to on-chain sealed-bid competition. Whether this increases or decreases proposer revenue depends on builder auction dynamics post-launch — data is not yet available.
Builders face a structurally different market. Sealed-bid auctions reduce information advantages. The current oligopoly of three builders controlling 93% of blocks may fragment if protocol-native auctions lower barriers to entry.
L2 sequencers benefit from cheaper, faster settlement on L1. However, further L1 fee compression narrows the cost advantage that justifies L2 usage for high-value transactions, potentially shifting some activity back to L1.
End users are the primary beneficiaries: lower gas costs (78.6% reduction from repricing alone), faster execution (parallel processing), and reduced MEV extraction (sealed block contents).
Glamsterdam is Ethereum's most structurally ambitious upgrade since The Merge. It addresses two of the network's most persistent vulnerabilities — single-threaded execution and off-protocol block construction — with protocol-level solutions rather than middleware patches. The scope of the change explains the development delays: ePBS touches "practically everything" in the consensus layer, per the Ethereum Foundation.
The economic implications are significant but uncertain. Moving block construction on-chain should reduce centralization risk, but whether it reduces builder profits or redistributes them depends on auction dynamics that cannot be modeled before mainnet activation. Gas repricing benefits users directly. Parallel execution benefits infrastructure operators by reducing node resource requirements.
The upgrade's timeline risk is real. ePBS complexity has already pushed the likely activation from Q2 to Q3 2026 or later. But the development trajectory — component devnets completed, generalized devnet launched, public testnets next — suggests consistent forward progress rather than stalled development. The question is when, not whether.