Ethereum's next hard fork, Glamsterdam, entered generalized devnet testing in the final week of April 2026 after months of component-level development. The upgrade combines two headliner EIPs — Enshrined Proposer-Builder Separation (EIP-7732) and Block-Level Access Lists (EIP-7928) — with at leas...
"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 next hard fork, Glamsterdam, entered generalized devnet testing in the final week of April 2026 after months of component-level development. The upgrade combines two headliner EIPs — Enshrined Proposer-Builder Separation (EIP-7732) and Block-Level Access Lists (EIP-7928) — with at least ten additional proposals targeting gas repricing, state sustainability, and validator operations. Developers are targeting a June 2026 mainnet activation, though multiple core contributors have characterized Q3 as more realistic given the complexity of two-party consensus coordination.
The stakes are structural. Today, two builders — Titan (52.2% of blocks) and BuilderNet (24.6%) — produce over 76% of Ethereum's blocks through off-chain relay infrastructure with no protocol-level accountability. Glamsterdam would move this market on-chain, replacing trust assumptions with cryptographic commitments. Simultaneously, Block-Level Access Lists would enable parallel transaction execution, targeting a 10x throughput increase from roughly 1,000 TPS to 10,000 TPS, while gas repricing proposals aim to cut execution costs by up to 78%. For a network with $280.7 billion in market capitalization and 35.8 million ETH staked, these are not incremental changes.
The Ethereum Foundation's Checkpoint #9 blog post, published April 10, 2026, confirmed the first generalized Glamsterdam devnet was being prepared for launch the following week. According to the Foundation, the generalized devnet merges previously separate testing environments — the epbs-devnet and bals-devnet — into a single integrated environment where all Glamsterdam components coexist and interact for the first time.
The devnet launched on schedule in the final week of April 2026. Developers had previously operated isolated devnets for each headliner feature but needed to verify cross-feature compatibility before proceeding. The Foundation acknowledged that "ePBS implementation is proving to be trickier than anticipated" and that non-headliner features like gas repricing "have their own complexities to work through."
The path from generalized devnet to mainnet follows a defined sequence: devnet stabilization, client releases, security audits, public testnet deployments (Holesky and Sepolia), and then a mainnet fork date announcement. Ethereum Foundation co-executive director Tomasz Stańczak has stated the gas limit would increase to 100 million per block initially, doubling to 200 million once ePBS is fully operational, with the possibility of reaching 300 million before year-end.
June 2026 remains the aspirational target. Most developers consider Q3 2026 more realistic.
EIP-7732 addresses what has become Ethereum's most concentrated supply chain. According to relay data from relayscan.io as of April 2026, 80-90% of Ethereum block production relies on off-chain builders via external relays with no cryptographic guarantees. The top builder, Titan, controls 52.16% of block production. BuilderNet, Flashbots' decentralized builder network launched in response to centralization concerns, handles 24.63%. The Flashbots relay itself has declined to 4.22% of payloads, down from its earlier dominance.
This off-chain market operates on trust. Builders construct blocks and route them through relays to proposers (validators), who select the highest-value block. But relays are unregulated intermediaries. There is no protocol-level enforcement if a relay censors transactions or a builder reneges.
EIP-7732 replaces this with a commit-reveal pipeline embedded in the consensus layer:
The practical effect is to make the builder "a first-class protocol participant: an actor with a registry identity, signed bids, and protocol-enforced accountability," according to QuickNode's technical analysis. The propagation window expands from roughly 2 seconds to approximately 9 seconds, giving builders more time to construct optimized blocks and enabling larger data payloads including additional blobs for Layer 2 settlement.
A January 2026 academic paper modeling ePBS in the presence of MEV found that while the mechanism reduces validator-side concentration, it "significantly amplifies profit and content centralisation" among builders, because access to private order flow confers a structural bidding edge that compounds over time. This finding aligns with Buterin's own assessment that ePBS contains centralization at the validator layer but does not resolve it at the builder layer — a problem deferred to the subsequent Hegotá upgrade, which has selected FOCIL (Fork-choice enforced Inclusion Lists, EIP-7805) as its consensus-layer headliner.
Ethereum currently executes transactions sequentially. A node discovers which accounts and storage slots a transaction touches only during execution, creating disk I/O bottlenecks that cap throughput regardless of available compute.
EIP-7928 requires builders to record all accounts and storage slots accessed during block execution into a Block-Level Access List (BAL). The BAL root hash is included in the block header, and the full access list travels with the block to all network nodes via a new wire protocol, eth/71 (EIP-8159).
This enables three capabilities:
The throughput target is approximately 10,000 TPS, up from an effective base-layer rate of roughly 1,000 TPS. This estimate assumes combined effects from BALs, gas limit increases, and repricing — it is not attributable to any single EIP.
Multiple EIPs target Ethereum's gas economics, which have not been comprehensively updated since their initial design:
EIP-7904 (General Repricing) recalibrates EVM opcode costs using empirical benchmarks from modern hardware. According to Bitfinex's technical analysis, the combined repricing results in a 78.6% reduction in gas costs for both simple ETH transfers and complex smart contract interactions. A DeFi swap currently costing $4-8 would drop proportionally.
EIP-2780 (Reduce Intrinsic Transaction Gas) targets standard ETH transfers specifically, making them up to 71% cheaper by reducing the base fee floor for simple transactions.
EIP-8037 (State Creation Gas Cost Increase) addresses the opposite problem: state bloat. The network currently grows by nearly 200 GiB per year in permanent state data. This EIP ties state creation fees to actual data size, introduces a reservoir model for predictable cost management, and separates gas accounting for computation from long-term storage obligations.
EIP-8038 (State-Access Gas Cost Update) increases gas costs for state-access opcodes to align pricing with actual hardware performance, closing denial-of-service vectors that exploit underpriced cold reads.
The net effect is a significant redistribution of costs: computation becomes cheaper, storage becomes more expensive. This is economically rational — compute is abundant and stateless, storage is scarce and permanent — but it will change the cost profile of every deployed contract.
Beyond the headliners, Glamsterdam includes several operational improvements:
More than 30 proposals were declined for this fork, including reduced slot times, multidimensional gas metering, and post-quantum signature verification. These deferrals reflect a deliberate scope constraint: ship ePBS and BALs cleanly before adding further complexity.
MEV concentration shift. ePBS solves the validator-centralization problem but may intensify builder concentration. Builders with access to private order flow retain a structural bidding advantage. The free-option problem — where builders can delay payload delivery to exploit price movements — affects roughly 0.82% of blocks on average, rising to 6% during periods of high volatility.
Implementation complexity. The Ethereum Foundation's Checkpoint #9 explicitly noted that ePBS introduces fundamentally new consensus dynamics: the protocol must now handle disagreement between two parties (proposer and builder) acting sequentially within a single slot. This is architecturally distinct from prior upgrades and increases the surface area for consensus-layer bugs.
Smart contract cost changes. Gas repricing will alter the cost profile of every deployed contract. Protocols optimized for current gas schedules may become more or less expensive to operate. The magnitude of the shift — up to 78% in some cases — could affect protocol economics in ways that are difficult to model in advance.
Timeline risk. Ethereum has historically shipped upgrades later than initial targets. The Merge was delayed multiple times. Pectra shipped in May 2025 after repeated postponements. Glamsterdam's June 2026 target, described as "aspirational" by the development team itself, carries similar risk. The devnet launched only in late April, leaving limited runway for the full testnet-to-mainnet pipeline within two months.
BAL correctness. If a builder produces an incorrect Block-Level Access List, the block becomes invalid. This creates a new failure mode at the block production layer. Mismatches between declared and actual state access could cause consensus splits if client implementations diverge in how they validate BALs.
Glamsterdam represents Ethereum's most structurally significant upgrade since the Merge. Where the Merge changed how blocks are validated (proof-of-stake), Glamsterdam changes how blocks are built (ePBS) and how they are executed (parallel processing via BALs). The economic scope is equally broad: gas repricing will redistribute costs across the entire protocol stack, making computation cheaper and storage more expensive.
The upgrade targets real, measurable problems. Builder concentration is empirically documented. Sequential execution is a measurable throughput constraint. Gas mispricing is a quantifiable economic distortion. Whether the solutions work as designed is the open question — one that devnets, testnets, and audits are meant to answer over the coming weeks.
For a network securing $280.7 billion in market capitalization with 35.8 million ETH staked, the margin for error is narrow. The development team's own characterization of June 2026 as "aspirational" suggests they are aware of the stakes. The question is not whether Glamsterdam ships, but whether it ships correctly.