Ethereum's Glamsterdam hard fork — the network's largest protocol overhaul since the 2022 Merge — completed its Sepolia testnet fork on August 3, 2026. Hoodi follows on August 17. Mainnet activation is scheduled for September 16, pending cross-client stability across multiple epochs on both publi...
"In 2026, no longer. 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 Glamsterdam hard fork — the network's largest protocol overhaul since the 2022 Merge — completed its Sepolia testnet fork on August 3, 2026. Hoodi follows on August 17. Mainnet activation is scheduled for September 16, pending cross-client stability across multiple epochs on both public testnets.
The upgrade bundles eight Ethereum Improvement Proposals around two structural changes: EIP-7732, which enshrines proposer-builder separation (ePBS) directly into the consensus layer, and EIP-7928, which introduces block-level access lists (BALs) to enable parallel transaction execution. A third package of gas repricing EIPs is projected to cut average L1 transaction costs by approximately 78.6%. Together, these changes aim to push the block gas limit from its current 60 million to a future target of 200 million, effectively shifting Ethereum's scaling narrative from "L2-only throughput" back toward a high-performance L1.
The stakes are material. ETH trades at approximately $1,914 as of August 9, 2026, with a market capitalization of $231 billion — well below its 2025 high near $5,000. Layer 2 networks now process 5x to 10x more daily transactions than mainnet, and the top two block builders construct roughly 73% of all blocks. Glamsterdam's success or failure will determine whether Ethereum can recapture fee revenue from its own rollup ecosystem and break the builder duopoly that currently extracts value from the network.
Glamsterdam is the consensus- and execution-layer fork that follows Fusaka (December 2025). It carries eight EIPs, divided into three functional clusters:
| Cluster | EIPs | Purpose | |---|---|---| | Consensus restructuring | EIP-7732, EIP-8061, EIP-8080 | Enshrined proposer-builder separation, validator workflow changes | | Execution parallelism | EIP-7928 | Block-level access lists for parallel state reads and validation | | Gas repricing | EIP-7904, EIP-8037, EIP-7976, EIP-7981, EIP-7778 | Realign gas costs with actual computational resources, remove refund distortions |
Fusaka raised the default block gas limit from 36 million to 60 million in December 2025, and subsequently increased the maximum blob count from 9 to 21 through two mini-upgrades. Glamsterdam does not directly raise the gas limit further but provides the technical prerequisites — specifically BALs and gas repricing — that make a future increase to 200 million feasible without compromising validator hardware requirements.
The problem. Since the Merge, Ethereum has relied on MEV-Boost, an out-of-protocol middleware maintained by Flashbots, to separate block proposers (validators) from block builders. This system works but introduces trust dependencies: validators must trust relays to deliver valid blocks, and relays must trust builders not to submit invalid payloads. As of mid-2026, the top two builders — Titan and Quasar — construct roughly 73% of all blocks. Earlier data showed concentration as high as 90% between Beaverbuild and Titan. The builder market has centralized, according to Flashbots, because major orderflow providers made exclusive deals with builders to internalize MEV through opaque off-chain agreements.
What EIP-7732 does. It moves proposer-builder separation into the protocol itself. Block builders assemble execution payloads and cryptographically seal their contents. Proposers select the highest-paying sealed block without seeing the transactions inside. Transactions are revealed only after the block is finalized. This eliminates the need for trusted relays and expands the data propagation window from approximately 2 seconds to roughly 9 seconds.
MEV impact. Developers project ePBS will reduce MEV extraction by up to 70%, primarily by eliminating the ability of builders and relays to engage in last-look transaction reordering. This is a direct transfer of value: MEV currently extracted by a concentrated set of builders would instead either remain with end users (through reduced sandwich attacks and front-running) or flow to validators through more competitive block auctions.
Flashbots launched BuilderNet in early 2026 as a stopgap — a decentralized block-building network using Trusted Execution Environments (TEEs), jointly operated by Flashbots, Beaverbuild, and Nethermind. EIP-7732 renders BuilderNet's relay-replacement function partially redundant by eliminating the relay layer entirely.
EIP-7928 introduces Block-Level Access Lists (BALs) — structured records of every account and storage location accessed during block execution, along with their post-execution values.
Why it matters. Current Ethereum execution is sequential: each transaction must complete before the next begins, because the EVM cannot predict which state a transaction will touch. BALs solve this by providing validators and execution clients with a pre-computed map of state access patterns at the block level.
This enables four parallel operations:
The practical result: once BALs are live, the network can safely accommodate a gas limit increase to 200 million without proportionally increasing hardware requirements for validators. At 200 million gas, Ethereum L1 could theoretically support approximately 10,000 transactions per second — a roughly 3.3x increase from the current 60 million gas limit capacity.
Five EIPs adjust the gas cost schedule:
The net effect is a projected 78.6% reduction in average L1 transaction fees for standard operations. For context: as of August 2026, Ethereum mainnet gas averages approximately 0.139 gwei, and a standard transfer costs a few cents. Post-Glamsterdam, basic operations could approach sub-cent territory.
This repricing also serves a secondary purpose: by making L1 transactions cheaper, it reduces the economic incentive to migrate activity exclusively to L2s. Ethereum currently averages 0.139 gwei on L1, while L2s like Base charge approximately $0.05 per transaction. If Glamsterdam compresses L1 fees to comparable levels, some activity could migrate back.
| Milestone | Date | Status | |---|---|---| | Soldøgn interop devnet conclusion | May 2, 2026 | Completed | | Stable multi-client devnet | May–July 2026 | Running | | Sepolia testnet fork | August 3, 2026 | Completed | | Hoodi testnet fork | August 17, 2026 | Scheduled | | Mainnet activation (target) | September 16, 2026 | Aspirational |
Historical precedent suggests caution. Past Ethereum forks have run two to four months of public-testnet seasoning before mainnet activation. If that cadence holds, mainnet could slip to Q4 2026.
Three risks could push activation beyond September:
ePBS implementation complexity. EIP-7732 represents the most significant change to Ethereum's consensus mechanism since the Merge. Cross-client implementation parity must hold across all major execution and consensus clients (Geth, Nethermind, Besu, Erigon on execution; Prysm, Lighthouse, Teku, Nimbus, Lodestar on consensus).
Gas repricing under load. Testing the repricing package under realistic mainnet-scale conditions remains outstanding. Devnet traffic patterns do not replicate mainnet's MEV-heavy, DeFi-concentrated transaction mix.
Validator coordination. With 34% of ETH supply now staked — approximately 41.4 million ETH across the network — the upgrade requires coordinated client updates from a validator set that has grown significantly since the Merge.
Counterpoint: the Ethereum Foundation delivered both Pectra (May 2025) and Fusaka (December 2025) on schedule, suggesting improved execution cadence relative to earlier years.
Validators. Mixed outcome. ePBS eliminates validator dependence on trusted relays and creates more competitive block auctions, potentially increasing proposer payments. However, the 70% reduction in MEV extraction reduces the total pool of MEV available. Net validator revenue impact depends on whether increased block auction competition offsets reduced MEV volume. Current base staking APR sits at approximately 2.66%, with staking participation at an all-time high of 34%.
Block builders. Clear losers. The builder duopoly — Titan and Quasar controlling ~73% of blocks — depends on exclusive orderflow agreements and relay relationships. ePBS eliminates relay trust requirements and sealed-block mechanics prevent builders from engaging in last-look MEV extraction. Builder margins compress.
End users. Clear beneficiaries. The 78.6% projected fee reduction combined with 70% MEV reduction translates to lower transaction costs and reduced value extraction from sandwich attacks and front-running.
L2 sequencers. Ambiguous. Lower L1 posting costs reduce rollup operating expenses, improving margins. However, if L1 fees approach L2 levels, some marginal activity migrates back to mainnet, reducing L2 transaction volumes.
ETH as an asset. If L1 fee revenue increases through higher transaction volume (enabled by lower per-transaction costs and higher gas limits), more ETH is burned through EIP-1559's base fee mechanism. This is structurally deflationary for ETH supply. Whether this translates to price appreciation depends on whether the volume increase exceeds the per-transaction fee decrease — a relationship that remains empirically untested at the projected scale.
The L2 landscape has consolidated significantly. Base and Arbitrum One together hold roughly 80% of all rollup TVL — Base at approximately $11.49 billion (40% share) and Arbitrum One at $10.12 billion (39%). Combined daily L2 transactions exceed mainnet by 5:1 to 10:1.
Glamsterdam introduces two dynamics that affect this equilibrium:
First, cheaper L1 settlement. L2 sequencers post batch data to Ethereum L1. Lower gas costs reduce this overhead, improving rollup economics. Arbitrum, Optimism, and Base all benefit proportionally.
Second, L1 competitive pressure. If Glamsterdam achieves its fee and throughput targets, Ethereum L1 becomes competitive with L2s for certain transaction types — particularly simple transfers and swaps where L2 complexity adds latency without meaningful cost savings. This does not threaten rollups' core value proposition for high-frequency applications but narrows their advantage for casual users.
The longer-term trajectory, assuming a future 200 million gas limit, implies Ethereum L1 could handle sufficient throughput to serve as both a settlement layer and a direct execution environment — the "L1 scaling" thesis that Vitalik Buterin outlined in his March 2026 roadmap post.
Glamsterdam is not a feature release. It is a structural reorganization of how Ethereum produces blocks and processes transactions. The protocol moves from a system where two builders control 73% of block production through opaque off-chain deals, to one where proposer-builder separation is enforced at the consensus layer. It replaces sequential execution with parallel processing. It reprices gas to reflect actual computational costs rather than historical approximations.
The question is not whether these changes are technically sound — the devnet has been stable since May. The question is whether Ethereum can execute a coordinated upgrade across nine major client implementations and 34% of a $231 billion network's supply within the projected timeline. The Foundation's improved delivery cadence since 2025 suggests it can. The complexity of ePBS — the most significant consensus change since proof-of-stake — suggests caution.
If Glamsterdam lands on schedule and performs as designed, it shifts the economic value distribution within the Ethereum ecosystem materially: away from MEV extractors and toward users and validators. That redistribution, not the throughput numbers, is the upgrade's most consequential outcome.