Ethereum's Glamsterdam hard fork — its largest protocol upgrade since the Merge — passed Devnet-11 testing the week of September 14, 2026, with 84,000 validators across a multi-client configuration. The upgrade bundles ten Ethereum Improvement Proposals under meta-EIP-7773, headlined by enshrined...
"I can just spin up a thousand builders. Any teenager can do this." — Potuz, Ethereum Consensus Developer, on testnet builder-abuse vectors in Glamsterdam
Ethereum's Glamsterdam hard fork — its largest protocol upgrade since the Merge — passed Devnet-11 testing the week of September 14, 2026, with 84,000 validators across a multi-client configuration. The upgrade bundles ten Ethereum Improvement Proposals under meta-EIP-7773, headlined by enshrined Proposer-Builder Separation (EIP-7732) and Block-Level Access Lists (EIP-7928). Together, these clear the technical path for raising the block gas limit from 60 million to 200 million, tripling Layer 1 execution capacity.
The Sepolia public testnet fork is confirmed for October 6, 2026, per the September 17 All Core Developers Consensus call. Mainnet activation remains unscheduled but targets Q4 2026. However, the compressed seven-day window between the client software release deadline (September 29) and the Sepolia fork has drawn scrutiny, and a builder-abuse vulnerability specific to the ePBS mechanism presents a known attack surface on public testnets where test ETH carries no economic cost.
Glamsterdam Devnet-11 launched on September 14, 2026, as a controlled "happy-path" rehearsal with 84,000 validators. Genesis occurred on September 14, the Gloas transition on September 16, and the block gas limit was raised from 60 million to 200 million shortly afterward. Nodes successfully built and validated blocks under the new rule set.
The Nethermind execution client passed all 2,302 performance tests during the exercise, processing 570.7 billion gas in 3 minutes and 15 seconds. These results confirmed that at least one major execution client can handle the proposed throughput without degradation.
Not all clients performed equally. Earlier adversarial testing on the Platåberget testnet — a longer-running public test environment launched by the Ethereum Foundation on August 17, 2026 — exposed fragility: a stress scenario knocked 12 of 13 Besu nodes offline and slowed both Erigon and Ethrex nodes. The five execution clients participating are Geth, Nethermind, Besu, Erigon, and Reth; the six consensus clients are Lighthouse, Prysm, Teku, Nimbus, Lodestar, and Grandine.
Earlier Devnet-9, launched September 1 with 1,000 validators, failed to finalize. Developers shifted directly to Devnet-11 with a substantially larger validator set.
EIP-7732: Enshrined Proposer-Builder Separation (ePBS). This EIP moves the relationship between validators (proposers) and specialized block builders into the protocol itself, replacing the current dependence on external relay infrastructure such as MEV-Boost. Under ePBS, the proposer selects a consensus block without seeing or rearranging the execution payload inside it. Builders compete by committing bids to the protocol; the winning builder then delivers the execution payload.
The mechanism introduces a Payload Timeliness Committee (PTC) and dual-deadline logic, allowing validators to attest to the consensus block and the execution payload timeliness separately. This extends the effective execution-payload propagation window from approximately 2 seconds to approximately 9 seconds per slot, accommodating larger blocks without forcing faster hardware requirements on validators. Ethereum's slot duration remains 12 seconds.
EIP-7928: Block-Level Access Lists (BALs). This EIP attaches a structured record to every block — a log of every account and storage slot touched, along with what changed after each transaction. BALs enable four capabilities: parallel execution of unrelated transactions, prefetching of state data from disk before a block begins executing, computation of post-state roots during block processing rather than after, and chain synchronization without the slow peer-negotiation healing phase that snap sync currently requires.
BALs are the technical prerequisite for operating safely at 200 million gas. Without parallel execution, blocks of that size would create unacceptable propagation delays and centralization pressure on validators.
Glamsterdam ships ten EIPs tracked under meta-EIP-7773:
| EIP | Description | |---|---| | EIP-7732 | Enshrined Proposer-Builder Separation (ePBS) | | EIP-7928 | Block-Level Access Lists (BALs) | | EIP-8159 | eth/71 P2P protocol upgrade | | EIP-8037 | State creation gas cost increase | | EIP-7954 | Max contract code size increase (24 KiB → 32 KiB) | | EIP-8038 | State-access gas cost updates | | EIP-7997 | Deterministic factory predeploy | | EIP-7708 | ETH transfers and burns emit logs | | EIP-2780 | Cheaper ETH transfers (standard: 21,000 gas; zero-value: 15,000 gas; self-transfer: 12,000 gas) | | EIP-7843 | SLOTNUM opcode |
The state-cost EIPs (8037 and 8038) address a growing sustainability concern. According to Chainstack analysis, Ethereum's state growth rate sits at approximately 380 GiB per year — a pace developers consider unsustainable. The targeted sustainable rate is 120 GiB per year. By increasing gas costs for state-creation operations, Glamsterdam aims to reduce state bloat without penalizing read-heavy workloads.
EIP-2780 reduces the cost of basic ETH transfers by up to 71%, directly lowering the cost of the most common on-chain operation.
Ethereum's block gas limit has followed a gradual expansion path: 30 million through most of 2024, rising to 60 million by December 2025 via validator gas-vote signaling. The 200 million target represents a 3.3x increase from the current floor.
The gas limit increase is not part of Glamsterdam itself — validators will coordinate the increase via signaling after the fork activates. Glamsterdam creates the technical headroom (via BALs and ePBS) that makes 200 million operable.
The throughput implications are measurable. Ethereum L1 processed a daily average of 25.78 transactions per second in Q1 2026, up from 13.67 TPS in Q1 2024, according to academic research published in June 2026 (arxiv.org/html/2606.22206v1). At 200 million gas under realistic workloads, developers project the network can support up to 10,000 TPS-equivalent throughput — though this figure has not been validated under adversarial conditions.
During the Fusaka hard fork that preceded Glamsterdam (activated December 3, 2025), the network recorded 248 missed blocks across 42 epochs, network participation dropped to 75%, and validators collectively lost approximately 382 ETH in rewards. These numbers serve as a baseline for measuring Glamsterdam's activation risk.
The September 17 All Core Developers Consensus call surfaced a specific vulnerability in ePBS as deployed on public testnets. Consensus developer Potuz described the attack vector: because test ETH can be obtained at near-zero cost, an attacker can create numerous builder identities, submit bids priced far above legitimate participants, repeatedly win the right to supply execution payloads, and then intentionally withhold those payloads.
The result: blocks are stripped of their transaction content, disrupting the infrastructure testing that Glamsterdam requires before proceeding to mainnet.
According to Potuz, no unified protection mechanism currently exists across all clients that can directly block individual malicious builders. Existing safeguards only fall back to locally built blocks after multiple payloads have already gone missing — a delayed response that leaves a window for repeated abuse. The developer argued that clients need to actively identify and reject specific malicious builders rather than relying on reactive fallbacks.
The vulnerability is testnet-only and does not threaten mainnet, where economic incentives — builders must post real ETH as collateral — prevent the same attack from being cost-effective. However, if builder abuse disrupts Sepolia testing severely enough, it could delay the mainnet timeline.
Developer Fredrik Svantes noted that standard Ethereum process requires "at least 14 days between release-ready client software and the first public testnet activation." The Sepolia schedule provides only seven days — a September 29 software release deadline for an October 6 fork.
Core developer Alex Stokes urged teams to "release software earlier where possible so more reviewers could examine it." The compressed window raises the risk that edge-case bugs reach the testnet undetected, though it does not by itself indicate that the upgrade is unprepared — Devnet-11 results and Platåberget stress testing have already exercised the core logic.
Ethereum's public roadmap lists Glamsterdam for Q4 2026 but states that no mainnet date has been confirmed. A further delay into 2027 has been described as explicitly on the table by the development team.
Glamsterdam's capacity expansion has direct economic consequences for both Layer 1 and the 73 active Layer 2 rollups that collectively hold over $48 billion in TVL as of April 2026, according to L2Beat data.
Layer 1 fee compression. Tripling gas capacity at constant demand should reduce gas prices, lowering revenue per transaction for validators. The introduction of EIP-2780's cheaper ETH transfers (up to 71% cost reduction) amplifies this effect. For users, this makes L1 competitive with rollups for simple transfers — a dynamic that may shift the economic calculus for transaction routing.
Layer 2 data availability costs. Rollups post data to L1 via blobs (introduced in Dencun, March 2024). While Glamsterdam does not directly modify blob pricing, the broader capacity increase and continued danksharding roadmap progression (PeerDAS, eventually full danksharding) are projected to reduce blob fees by up to 90%, according to Symbiosis Finance analysis. Base currently charges a median fee of approximately $0.02 per USDC transfer; further L1 capacity expansion could push this lower.
L1-L2 competitive dynamics. Arbitrum One and Base account for approximately 77% of all Layer 2 DeFi TVL ($13.8 billion and $11.2 billion respectively). If L1 gas prices fall sufficiently post-Glamsterdam, some transaction volume may migrate back to mainnet — particularly for high-value DeFi operations where L1 security guarantees justify the remaining cost premium.
L2 total TPS rose from 78.60 in Q1 2024 to 226.92 in Q1 2026. Whether a more capable L1 complements or competes with this trajectory remains an open question that depends on post-fork gas price equilibria.
Glamsterdam represents the most significant structural change to Ethereum's execution and consensus layers since the Merge in September 2022. The combination of ePBS and BALs addresses two of the network's most persistent criticisms: reliance on off-protocol block-building infrastructure and sequential transaction processing.
The Devnet-11 results are positive but incomplete. Nethermind handled the load; Besu did not survive adversarial conditions on Platåberget. The compressed Sepolia timeline and the unresolved builder-abuse vulnerability introduce execution risk to the testing phase. Whether these are manageable engineering problems or indicators of insufficient preparation will become clear in October.
The economic stakes are substantial. A successful 200M gas limit deployment would triple L1 capacity at a time when 73 rollups and $48 billion in L2 TVL have been built on the assumption that L1 capacity is scarce. The resulting fee dynamics could reshape value distribution between Ethereum mainnet, its rollup ecosystem, and the validators and sequencers that operate them.
The data supports cautious optimism on the technical merits but flags timeline risk. Ethereum has not missed a hard fork yet, but the Fusaka activation cost — 248 missed blocks and 382 ETH in validator losses — demonstrates that even successful upgrades carry measurable economic friction.