Ethereum's Glamsterdam hard fork — the network's largest protocol change since the September 2022 Merge — entered its final testing phase in August 2026 with the launch of the Platåberget public testnet. The Sepolia fork is provisionally scheduled for September 28, 2026, with mainnet activation t...
"In Glamsterdam, Ethereum is getting ePBS, which lets proposers outsource to a free permissionless market of block builders." — Vitalik Buterin, Ethereum Co-Founder
Ethereum's Glamsterdam hard fork — the network's largest protocol change since the September 2022 Merge — entered its final testing phase in August 2026 with the launch of the Platåberget public testnet. The Sepolia fork is provisionally scheduled for September 28, 2026, with mainnet activation targeted for Q4 2026. The upgrade bundles more than 20 Ethereum Improvement Proposals under Meta EIP-7773, headlined by EIP-7732 (Enshrined Proposer-Builder Separation) and EIP-7928 (Block-Level Access Lists).
The combined effect is a structural overhaul of how Ethereum produces, validates, and prices blocks. If deployed on schedule, Glamsterdam will triple the network's gas capacity from 60 million to a 200-million target floor, cut standard ETH transfer costs by up to 71%, expand the data propagation window from 2 to 9 seconds, and lay the technical foundation for parallel transaction execution — a prerequisite for the network's stated goal of 10,000 L1 TPS. At the same time, the upgrade removes Ethereum's dependence on third-party MEV relay infrastructure, a market in which two builders currently construct 73% of all blocks.
The Glamsterdam upgrade was initially targeted for H1 2026 but slipped as scope expanded. The current progression:
| Milestone | Date | Status | |-----------|------|--------| | Devnet-4 multi-client testing | June 2026 | Completed | | Final devnet phase | June 2026 | Completed | | Platåberget public testnet launch | August 17, 2026 | Completed | | Platåberget Glamsterdam fork | August 20, 2026 | Completed | | Sepolia testnet fork | September 28, 2026 (provisional) | Pending | | Hoodi testnet fork | TBD | Pending | | Mainnet activation | Q4 2026 | Pending |
The Ethereum Foundation opened the Platåberget testnet specifically for wallet providers, infrastructure teams, and smart contract developers to test code against the new gas model before mainnet deployment.
The centerpiece of Glamsterdam is the enshrinement of Proposer-Builder Separation (ePBS) directly into Ethereum's consensus layer. Under the current system, over 90% of Ethereum validators run MEV-Boost, a third-party middleware developed by Flashbots, to outsource block construction to specialized builders. This architecture, while functional, depends on trusted relays and has produced severe market concentration.
According to Relayscan data for the week ending July 23, 2026, Titan built approximately 51% of all MEV-Boost blocks, Quasar built 23%, BuilderNet 11%, and Eureka 10%. The Herfindahl-Hirschman Index for the builder market measured approximately 3,892 — more than double the 1,800 threshold the U.S. Department of Justice uses to define a "highly concentrated" market. For context, in late 2022, no single builder controlled more than 20% of market share.
EIP-7732 addresses this by making the trustless exchange of a block payload for payment a protocol-level operation. The mechanism works as follows:
This eliminates the relay as a required intermediary. The data propagation window expands from approximately 2 seconds to 9 seconds, accommodating larger blocks and reducing the advantage held by low-latency builders. According to research from the Flashbots Collective, ePBS could reduce MEV-related losses for users by up to 70%.
Relays will not disappear overnight. The protocol becomes what core developers describe as "the canonical neutral relay," but existing relay operators may continue to offer value-added services. The structural difference is that relay failure or censorship no longer constitutes a systemic risk.
The second headline EIP introduces Block-Level Access Lists (BALs), attaching a structured record of every account and storage slot a block touches to each block header. The implications are threefold:
Parallel execution. By knowing which transactions touch which state, clients can safely execute non-conflicting transactions simultaneously rather than sequentially. This is the core technical enabler for scaling L1 throughput toward 10,000 TPS — approximately 10x Ethereum's current average of 25.78 TPS recorded in Q1 2026.
Prefetching. Clients can begin loading required state from disk before block execution starts, reducing I/O bottlenecks.
Executionless sync. New nodes joining the network can verify blocks without replaying every transaction, accelerating sync times.
EIP-8159 (eth/71) provides the companion networking protocol, enabling nodes to share access lists across the peer-to-peer layer. The practical effect is that the infrastructure bottleneck shifts from CPU-bound sequential execution to a parallelizable workload — a necessary condition for the 200M gas limit target.
Since Ethereum's genesis in 2015, a standard ETH transfer has cost exactly 21,000 gas. Glamsterdam replaces this constant with resource-based pricing through EIP-2780.
The old flat fee decomposes into measured components:
| Component | Gas Cost | Description | |-----------|----------|-------------| | TX_BASE_COST | 12,000 | Signature recovery, sender access, block inclusion | | COLD_ACCOUNT_ACCESS | 3,000 | Touching the recipient account | | TX_VALUE_COST | 6,000 | Recipient balance write and transfer log | | Total (existing account) | 21,000 | Unchanged for standard transfers |
For many transaction types, the net effect is a reduction. Zero-value transactions fall to 15,000 gas. Self-transfers cost 12,000 gas. The Ethereum Foundation estimates that standard ETH transfers between existing accounts will remain at 21,000 gas, but transfers that create a new account will now incur an additional 183,600 units of state gas — a deliberate repricing to make state growth economically sustainable.
The wallet ecosystem impact is non-trivial. According to CoinDesk reporting from August 18, 2026, any wallet, developer tool, or dApp still hardcoding the 21,000 gas limit or relying on a single-dimensional gas model will malfunction once Glamsterdam goes live.
EIP-8037 establishes fixed cost-per-state-byte (CPSB) pricing, targeting a maximum state growth rate of 120 GiB per year. This is the economic guardrail that makes a 200M gas limit viable: computation becomes cheaper (EIP-7904 estimates a 78.6% reduction in average L1 fees from opcode repricing alone), but permanent state creation becomes more expensive.
The 200M gas limit is not a hard parameter set by the fork. It is a design target that becomes safe to reach through validator gas-vote signaling after Glamsterdam's BALs and state pricing go live. Current gas capacity sits at approximately 60 million following the Fusaka upgrade in December 2025.
For context, Ethereum's median daily gas limit has already doubled from 30 million in January 2024 to 60 million by December 2025 through successive upgrades. Glamsterdam clears the path for another 3.3x increase.
Beyond the headlines, Glamsterdam ships several changes with significant downstream effects:
EIP-7954 raises the maximum contract bytecode size from approximately 24 KiB to 64 KiB and the initcode limit from 48 KiB to 128 KiB. This unblocks deployment of larger, more complex smart contracts that currently require splitting across multiple addresses.
EIP-7708 mandates that all ETH transfers and burns emit a standard log event, analogous to ERC-20 Transfer events. This closes a long-standing gap in Ethereum's event infrastructure: native ETH movements have historically been invisible to standard event-monitoring tools, forcing indexers and analytics platforms to rely on trace-level data.
EIP-8045 excludes slashed validators from future block proposal duties, tightening network security.
EIP-8061 increases exit and consolidation churn capacity by approximately 4x and 2x respectively, reducing queue times for validators seeking to exit or consolidate stakes.
EIP-7997 deploys a universal deterministic factory contract at address 0x12, enabling identical smart contract addresses across multiple EVM-compatible chains — a practical improvement for cross-chain deployment workflows.
The economic effects of Glamsterdam operate on multiple layers:
Fee reduction for users. The combination of opcode repricing (EIP-7904) and increased gas capacity is projected to reduce average L1 fees by 71-79%, according to Ethereum Foundation estimates and KuCoin research. This narrows the fee differential between L1 and L2 networks, potentially pulling some transaction volume back to mainnet.
MEV redistribution. ePBS eliminates the relay operator as a rent-extracting intermediary. The Flashbots Collective estimates up to 70% reduction in MEV-related losses. However, the builder market itself remains concentrated — ePBS removes relay risk but does not directly address builder oligopoly. Fork-Choice Inclusion Lists (FOCIL/EIP-7805), which would have addressed proposer censorship resistance, were deferred from Glamsterdam's scope.
State creation costs. Higher gas prices for permanent state writes create an implicit tax on protocols that generate significant on-chain state. DeFi protocols, NFT platforms, and any application deploying large contract suites will face higher deployment costs, offset by lower execution costs.
Infrastructure upgrade burden. Wallets, indexers, gas estimation tools, and RPC providers must update software to accommodate the new gas model. The 21,000-gas assumption is embedded in code across the ecosystem, dating back over a decade.
Complexity risk. With 20+ EIPs touching consensus, execution, and networking layers simultaneously, Glamsterdam represents a broad surface area for potential bugs. The Ethereum Foundation's multi-testnet approach (Platåberget → Sepolia → Hoodi → mainnet) is designed to mitigate this, but the upgrade's scope is the primary reason for its delay from H1 to Q4 2026.
FOCIL deferral. EIP-7805 (Fork-Choice Inclusion Lists), which would have given proposers a tool to resist builder censorship, was not included. This means ePBS addresses relay centralization but leaves censorship resistance as an open item for a future fork.
Builder concentration persistence. While ePBS removes the relay layer, the underlying dynamic — exclusive order flow agreements between searchers and dominant builders — remains unaddressed at the protocol level. BuilderNet, launched by Flashbots, Beaverbuild, and Nethermind to share order flows via TEE hardware, produced 25.5% of blocks in January 2026 but is a voluntary, off-protocol solution.
Wallet compatibility. The gas model change is backward-incompatible for tools hardcoding the 21,000 constant. The Platåberget testnet provides a testing window, but Ethereum's fragmented wallet ecosystem makes universal adoption before mainnet uncertain.
Glamsterdam represents a structural redesign of Ethereum's block production and gas economics, not an incremental parameter adjustment. The ePBS enshrinement addresses a governance vulnerability — the network's near-total dependence on third-party relay infrastructure — while BALs provide the execution architecture for the next phase of L1 scaling.
The economic tradeoff is explicit: computation becomes cheaper, state creation becomes more expensive. This pricing shift favors high-throughput, low-state applications and imposes higher costs on state-heavy deployments. Whether the 200M gas limit materializes depends on validator signaling after the fork — but the technical and economic prerequisites will be in place.
The remaining question is timing. Glamsterdam has already slipped six months. The Sepolia fork, provisionally set for September 28, will provide the first large-scale public test of the full EIP bundle. If testing proceeds without major issues, Q4 2026 mainnet activation remains plausible. If not, the upgrade could push into early 2027 — extending the period in which Ethereum's block production depends on an increasingly concentrated relay market.