Ethereum's Glamsterdam hard fork — the network's largest protocol change since the 2022 Merge — enters public testing on Sepolia at 13:53:36 UTC on October 6, 2026. The upgrade combines the Amsterdam execution-layer release with the Gloas consensus-layer release, packaging two structural changes:...
"I can just spin up a thousand builders. Any teenager can do this." — Potuz, Ethereum Consensus Developer, on ePBS testnet risks
Ethereum's Glamsterdam hard fork — the network's largest protocol change since the 2022 Merge — enters public testing on Sepolia at 13:53:36 UTC on October 6, 2026. The upgrade combines the Amsterdam execution-layer release with the Gloas consensus-layer release, packaging two structural changes: EIP-7928 (Block-Level Access Lists) for parallel transaction processing, and EIP-7732 (Enshrined Proposer-Builder Separation) to move block construction into the protocol itself and eliminate dependence on third-party MEV relays.
Devnet-11 rehearsals completed on September 16 with 84,000 validators running a multi-client configuration. The gas limit rose from 60 million to 200 million without loss of finality. Nethermind's execution client passed all 2,302 performance tests, processing 570.7 billion gas in 3 minutes and 15 seconds — roughly 2.9 billion gas per second. If mainnet deployment proceeds on the Q4 2026 roadmap, Ethereum L1 throughput would target approximately 10,000 transactions per second, a 10x increase from the current effective rate of ~1,000 TPS.
Glamsterdam merges two coordinated releases. On the execution layer, Amsterdam introduces parallel processing, gas repricing, and expanded contract size limits. On the consensus layer, Gloas embeds proposer-builder separation directly into the protocol.
The full EIP roster scheduled for inclusion:
| EIP | Layer | Purpose | |------|-------|---------| | EIP-7732 | Consensus | Enshrined Proposer-Builder Separation (ePBS) | | EIP-7928 | Execution | Block-Level Access Lists (BALs) | | EIP-8037 | Execution | State creation gas cost increase | | EIP-8038 | Execution | State access cost realignment | | EIP-7976 | Execution | Calldata floor cost increase | | EIP-7981 | Execution | Access list entry cost increase | | EIP-7954 | Execution | Maximum contract size limit increase | | EIP-7778 | Execution | Gas refund removal from block accounting | | EIP-8024 | Execution | Backward-compatible stack-manipulation opcodes | | EIP-7843 | Execution | SLOTNUM opcode | | EIP-7708 | Execution | Native ETH transfers emit logs |
Both CL and EL must activate simultaneously. Client versions confirmed for Sepolia: Lodestar 1.49.0, Prysm 7.2.0, Teku 26.9.1 (consensus); Besu 26.9.0, Erigon 3.7.0, go-ethereum 1.17.6, Nethermind 2.0.0, Reth 2.7.0 (execution).
The current Ethereum Virtual Machine processes transactions sequentially. EIP-7928 introduces Block-Level Access Lists (BALs) — data structures that pre-declare which accounts and storage slots each block will touch. When two transactions access entirely different parts of the state, validators can process them simultaneously.
Nethermind's benchmark of this feature processed 570.7 billion gas across 2,302 tests in 195 seconds. The mechanism enables pre-fetching of state data from disk before execution begins, eliminating the I/O bottleneck that currently constrains throughput.
Combined with a gas limit increase from 60 million to 200 million per block, BALs target a throughput of approximately 10,000 TPS. Ethereum Foundation documentation projects a gas cost reduction of up to 78% across both simple transfers and complex smart contract interactions, according to the Ethereum Foundation's Glamsterdam roadmap page.
The economic implications are direct. Lower L1 gas costs compress the fee revenue that accrues to validators per transaction, but the 10x throughput increase is designed to more than offset this through volume. Whether that offset materializes depends on whether demand scales proportionally — a question the data cannot yet answer.
Today, Ethereum block construction relies on MEV-Boost, an out-of-protocol relay system maintained primarily by Flashbots. The concentration is measurable: Titan Builder produces approximately 46.5% of mainnet blocks, followed by Rsync Builder at 15.6%, Flashbots at 12.8%, and Beaverbuild at 9.4%. The top four builders collectively construct over 90% of blocks. The Herfindahl-Hirschman Index for Ethereum block building stands at 3,892 — well above the U.S. Department of Justice's 1,800 threshold for a "highly concentrated" market.
EIP-7732 moves this process on-chain. Under the new mechanism, builders submit signed execution-payload bids to the protocol. Proposers select bids without viewing full transaction contents. A new Payload Timeliness Committee verifies delivery. Payment from builder to proposer settles in-protocol, removing the trust dependency on relay middleware.
The execution-payload propagation window expands from approximately 2 seconds to 9 seconds, giving nodes additional time to distribute and validate larger payloads under the higher gas limit.
The MEV cost to ordinary users is substantial. Annual losses to sandwich attacks, frontrunning, and just-in-time extraction exceed $500 million, according to data compiled by Bex Research. Sandwich attacks alone represent 51% of MEV volume as of 2025. In a single 30-day period, researchers documented over 72,000 sandwich attacks targeting more than 35,000 victims. In one documented case, a user's $220,764 transaction was compressed to $5,271 — a 98% value extraction.
Research from the Flashbots Collective estimates ePBS could reduce MEV-related losses for users by up to 70% by creating a more competitive and transparent auction. However, the reduction is not guaranteed. Exclusive order flows — particularly those channeled through MEV-Share and similar private transaction systems — remain outside the protocol's reach. Unless these are dissolved or made public, some builders may retain disproportionate access to transaction flow, preserving a degree of centralization.
Two EIPs restructure gas pricing to reflect actual computational costs, addressing a divergence that has widened since the Berlin fork in 2021.
EIP-8037 increases and harmonizes state creation costs — new accounts, new storage slots, and deployed bytecode. It introduces a per-state-byte cost metric designed to bound average annual state growth to sustainable levels under the higher 200 million gas limit. Without this repricing, the 3.3x gas limit increase would accelerate state bloat proportionally.
EIP-8038 raises state access costs. Specific changes identified in the specification: cold account access increases from 2,600 gas to 3,000 gas. Account-write cost becomes 9,000 gas. Storage-write cost becomes 10,000 gas. CREATE access becomes 12,000 gas.
The Ethereum Foundation's August 2026 blog post flagged contracts using 2,300-gas stipends, fixed call limits, or gasleft() logic as facing the greatest compatibility risk. Developers have been urged to audit gas assumptions in deployed contracts. According to the Foundation, most flagged failures can be resolved by increasing transaction gas limits, but contracts with hardcoded gas values may require redeployment.
Ethereum's developer community has flagged a specific vulnerability in the Sepolia test. Under ePBS, builder identity is permissionless. On testnets where ETH is free, an attacker can register numerous builder identities, submit high bids, and then withhold payloads — stalling block production and degrading test quality.
Consensus developer Potuz described the attack surface succinctly: an attacker needs nothing more than free test ETH and disposable builder identities to repeatedly win block auctions and refuse delivery. The risk does not threaten mainnet funds, but it could undermine the reliability of the rehearsal itself.
The client software deadline for Sepolia was September 29 — only a 7-day review window, half the standard 14 days. The Hoodi testnet, which uses an open validator set, is tentatively scheduled for October 27. Mainnet activation remains unconfirmed within the Q4 2026 window.
Glamsterdam affects Layer 2 rollups through two mechanisms. First, lower L1 gas costs reduce the expense of posting calldata and proofs from L2s to mainnet — the primary cost component in the rollup economic model. This directly compresses the "sequencer margin," the spread between what users pay on L2 and what the sequencer pays to settle on L1. For rollups like Arbitrum and Optimism, lower settlement costs could either flow to users as fee reductions or accrue as increased protocol margin.
Second, a more capable L1 creates competitive pressure. If Ethereum mainnet can process 10,000 TPS at 78% lower gas costs, the value proposition of routing transactions through L2s narrows. Base currently holds over $4.2 billion in DeFi value locked; Arbitrum holds approximately $1.3 billion. Whether this capital shifts back toward L1 depends on realized throughput and latency — metrics that remain theoretical until mainnet activation.
On smart contract compatibility, the gas repricing in EIP-8037 and EIP-8038 introduces backward-compatibility risk for contracts deployed under prior gas assumptions. Contracts that hardcode gas stipends or rely on specific gasleft() values may fail under the new pricing schedule. The Ethereum Foundation has published testing guidance, but the burden of auditing falls on individual protocol teams.
ETH traded at $2,706 as of October 1, 2026 — approximately 45% below its August 2025 peak of $4,946. Approximately 37 million ETH (30.6% of circulating supply) is locked in staking contracts, with over 3 million additional ETH in the validator entry queue.
Ethereum co-founder Vitalik Buterin has stated that the subsequent fork, Hegota, could be the network's last "normal" hard fork before development shifts toward recursive STARKs, automated formal verification, optimized consensus, and quantum-safe cryptography. If accurate, Glamsterdam represents the final upgrade in Ethereum's current architectural paradigm — a fact with implications for how the market prices the transition risk.
Glamsterdam is architecturally significant. It attempts two things simultaneously: a 10x throughput increase through parallel execution, and a structural shift in how blocks are built by moving proposer-builder separation into the protocol itself. The first addresses Ethereum's persistent capacity constraint. The second addresses a concentration problem where four entities control 90%+ of block production.
The risk profile is equally clear. Gas repricing may break deployed contracts. Builder abuse could compromise testnet quality. The 7-day Sepolia review window is half the standard. And the competitive dynamics between a faster L1 and existing L2s remain unresolved.
Whether the upgrade delivers on its throughput targets is an empirical question that only mainnet data can answer. The October 6 Sepolia activation is the first public data point.