Ethereum's Glamsterdam hard fork — the network's largest protocol overhaul since The Merge in September 2022 — entered public testing on August 20, 2026, via the Platåberget testnet. Sepolia and Hoodi testnet deployments are anticipated in September, with mainnet activation targeted for Q4 2026. ...
"Protocol is now a more united and leaner organization with more focused teams…ensuring the EF's resources are allocated toward maximal impact." — Tim Beiko, Former Ethereum Foundation Protocol Lead
Ethereum's Glamsterdam hard fork — the network's largest protocol overhaul since The Merge in September 2022 — entered public testing on August 20, 2026, via the Platåberget testnet. Sepolia and Hoodi testnet deployments are anticipated in September, with mainnet activation targeted for Q4 2026. The upgrade bundles ten Ethereum Improvement Proposals (EIPs), headlined by EIP-7732 (Enshrined Proposer-Builder Separation) and EIP-7928 (Block-Level Access Lists), which together restructure how Ethereum builds blocks and executes transactions.
The fork arrives at a critical juncture. Ethereum's gross revenue fell 69.3% year-over-year to $127 million in H1 2026, according to 21Shares, as activity migrated to Layer 2 networks. The network's annualized L1 fee revenue sits near $65 million — down from $30 million per day at peak in 2021-2022. Glamsterdam represents a strategic pivot: rather than continuing to cede execution entirely to rollups, Ethereum is scaling its base layer to recapture value, targeting a 200-million gas limit (up from ~60 million) and a theoretical ceiling of 10,000 transactions per second.
The Ethereum Foundation's Protocol DevOps team launched the Platåberget public testnet on August 17, 2026, with the Glamsterdam fork activating on the network three days later. The testnet is permissionless — validators, developers, and users can join via one-click configuration at plataberget.dev, with a block explorer at dora.plataberget.ethpandaops.io and faucet at faucet.plataberget.ethpandaops.io.
Platåberget runs on Glamsterdam devnet-8 specifications, incorporating all ten EIPs scoped for inclusion. Six Ethereum client teams have published container images: Geth, Nethermind, and Reth on the execution layer; Lighthouse, Prysm, and Teku on the consensus layer.
The progression path is Platåberget → Sepolia and Hoodi testnets (expected September 2026) → Ethereum mainnet (Q4 2026). The original timeline targeted June 2026 for mainnet; the delay stems from ePBS implementation complexity and cross-client interoperability testing that required additional devnet cycles, including the Soldøgn interop event concluded May 2, 2026, in Svalbard, Norway.
The Ethereum Foundation simultaneously announced a leadership transition in its Protocol cluster. Tim Beiko and Barnabé Monnot are departing the Foundation, and Alex Stokes is taking a sabbatical. Will Corcoran, Kev Wedderburn, and a developer identified as Fredrik will serve as new Protocol leads.
Glamsterdam modifies Ethereum at two structural layers simultaneously. The consensus layer component (Gloas) reshapes how blocks are proposed and validated. The execution layer component (Amsterdam) overhauls how transactions are processed and priced. Together, they represent the most extensive set of coordinated protocol changes since Ethereum transitioned from proof-of-work to proof-of-stake.
The ten EIPs break into three functional groups:
EIP-7732 moves Proposer-Builder Separation from its current off-chain implementation directly into the Ethereum protocol. This is the single most consequential change in the upgrade.
Current state: Approximately 90% of Ethereum blocks are constructed via MEV-Boost, an off-chain system where validators outsource block building to specialized builders through third-party relay operators. Relay market share as of April 2026: Ultrasound Money at 33.92%, Titan at 24.19%, bloXroute at 14.67%, and Aestus at 10.03%, according to Relayscan data. Flashbots' own relay — despite having created the MEV-Boost standard — held just 3.44% of the 7-day window as of late October 2025.
Post-Glamsterdam: ePBS eliminates the need for these trusted relay intermediaries. Block building becomes a protocol-native process with three changes:
The estimated impact on MEV extraction is a reduction of up to 70%, according to protocol researchers. For end users, this translates to more transparent transaction ordering, reduced swap slippage, and a lower hidden tax on DeFi activity. The Ethereum ecosystem currently sustains $1-5 billion in annual MEV extraction, per the webthreepedia economic value analysis.
The relay operators — Ultrasound Money, Titan, bloXroute, Flashbots, and others — face existential disruption. Their role as trusted intermediaries between proposers and builders becomes redundant once ePBS is enshrined. This does not eliminate MEV itself, but it changes how MEV is captured and redistributed, shifting control from off-chain infrastructure operators to protocol-level mechanisms.
EIP-7928 introduces Block-Level Access Lists (BALs), a data structure that maps which accounts and storage slots each transaction in a block will touch. This information, shared via a new peer-to-peer protocol (EIP-8159, eth/71), enables Ethereum clients to execute transactions in parallel across multiple CPU cores rather than sequentially.
Performance implications: The current 60-million gas limit processes transactions one at a time. With BALs providing upfront knowledge of state dependencies, non-conflicting transactions can be executed simultaneously. This unlocks the path to a 200-million gas limit — more than triple current capacity — without requiring proportionally more powerful hardware.
The 200-million target is not enforced by the fork itself. Validators set the gas limit via standard gas-vote signaling and would increase it incrementally as nodes demonstrate they can handle larger blocks without degraded propagation. Without parallel execution and the associated gas repricings, raising the gas limit to 200 million would produce unsustainable state growth exceeding 380 GiB annually, according to protocol specifications. EIP-8037's Cost Per State Byte mechanism targets a sustainable 120 GiB per year.
BALs also enable a feature called executionless state reconstruction, allowing new nodes to sync the chain without re-executing every historical transaction — a meaningful improvement for node operator onboarding.
Beyond the two headline proposals, eight additional EIPs are scoped for inclusion:
| EIP | Change | Practical Impact | |-----|--------|-----------------| | 8037 | State Creation Gas Cost Increase | Introduces separate gas metering for state-creating operations via a dedicated state_gas_reservoir. Anti-state-bloat mechanism. | | 7976 | Increase Calldata Floor Cost | Raises the gas cost of calldata to reflect its actual bandwidth burden. | | 7981 | Increase Access List Cost | Recalibrates EIP-2930 access list pricing upward. | | 7954 | Increase Maximum Contract Size | Expands deployed contract limit from ~24 KiB to 64 KiB; initcode from 48 KiB to 128 KiB. | | 7778 | Block Gas Accounting Without Refunds | Removes gas refund mechanics from block-level accounting. | | 8024 | Backward-Compatible SWAPN, DUPN, EXCHANGE | Adds new EVM stack manipulation opcodes. | | 7843 | SLOTNUM Opcode | Exposes the current slot number to smart contracts. | | 7708 | ETH Transfers Emit Log | Native ETH transfers and burns now generate ERC-20-style event logs. |
A critical warning from the Ethereum Foundation: "Any tool that relies on a hardcapped maximum gas limit — think wallets, indexers and gas estimators — will break and needs to be updated." Additionally, plain ETH transfers will no longer universally cost 21,000 gas. Transfers to new accounts will incur additional state gas charges.
Glamsterdam is not merely a technical upgrade. It is a response to a structural economic shift that has eroded Ethereum's L1 value capture.
Revenue decline by the numbers:
Activity metrics tell a different story:
The paradox is clear: Ethereum hosts more activity than ever, but captures less revenue per unit of activity. The Dencun upgrade (March 2024) introduced proto-danksharding via EIP-4844, cutting L2 data costs by 90-99%. This was beneficial for L2 economics but devastated L1 fee income. Glamsterdam's L1 scaling ambitions aim to rebalance this equation by making the base layer competitive for direct execution again.
At $2,477 per ETH as of September 5, 2026, the market has not priced in a recovery in L1 fee revenue. ETH faces repeated rejection near $2,550, with support at $2,438.
Validators and stakers must update both Consensus Layer and Execution Layer clients before the fork activates. The new PTC attestation duty introduced by ePBS adds a requirement that did not exist previously. Validators who fail to upgrade will fork onto an incompatible chain.
Node operators need to verify storage IOPS capacity for parallel execution workloads and prioritize network latency for PTC participation. The shift to parallel transaction processing changes the hardware profile from CPU-clock-speed-dependent to core-count-and-IOPS-dependent.
dApp developers benefit from backward compatibility — existing contracts continue to function. However, three gas repricings (state creation, calldata, access lists) change the cost structure for deploying contracts and interacting with state. Applications that create significant new state will see higher costs. Applications performing computation on existing state will see relatively lower costs as the gas limit expands.
L2 rollups benefit from the longer propagation window enabled by ePBS, which allows higher blob capacity in future forks. Over time, this reduces L2 data posting costs. However, a more capable L1 also creates competitive pressure: if Ethereum L1 can process 10,000 TPS at low cost, the value proposition of some L2s narrows.
Two proposals initially considered for Glamsterdam were deferred:
EIP-7782 (6-Second Slots): Would have halved Ethereum's slot time from 12 seconds to 6 seconds. Shelved due to conflicts with zero-knowledge proving timelines — current ZK provers cannot generate proofs fast enough for 6-second blocks — and incomplete client implementations. The feature may return in a future fork.
FOCIL / EIP-7805 (Fork-Choice Inclusion Lists): A censorship-resistance mechanism that would force validators to include certain transactions. Moved to Hegotá (the next planned upgrade, targeting H1 2027) to avoid interaction complexity with the ePBS implementation. The interaction between ePBS and FOCIL requires additional specification work.
Glamsterdam is Ethereum's attempt to answer a question the network has deferred since the rollup-centric roadmap was articulated in 2020: can Ethereum's base layer remain economically relevant if the majority of execution happens elsewhere?
The data suggests urgency. At $127 million in H1 2026 revenue — roughly what a mid-tier SaaS company generates — Ethereum's fee income does not match its $297 billion market capitalization or its role as settlement infrastructure for $156 billion in stablecoins and $16 billion in tokenized real-world assets. The network processes record transaction volumes while capturing diminishing revenue per transaction.
Glamsterdam does not solve this problem alone. A 200-million gas limit and parallel execution create capacity, but capacity without demand is infrastructure without revenue. The upgrade's economic success depends on whether applications choose to execute on L1 rather than L2s when the cost differential narrows — a behavioral question, not a technical one.
What Glamsterdam does accomplish is removing two long-standing architectural constraints. ePBS eliminates Ethereum's dependency on trusted off-chain relay operators for 90% of its block production — a systemic risk that has persisted since MEV-Boost's adoption in 2022. Block-Level Access Lists end the sequential execution bottleneck that has capped L1 throughput since genesis.
Whether these changes translate into recovered fee revenue will depend on the gas limit signaling decisions of validators, the migration patterns of applications, and the competitive response from L2 networks and alternative L1s. The testnet data from Platåberget, Sepolia, and Hoodi over the coming weeks will provide the first empirical signal of whether the 200-million gas target is operationally viable.