Ethereum's Glamsterdam hard fork — the network's largest protocol change since the 2022 Merge — activated on the Sepolia testnet on October 6, 2026, at epoch 353,024 (13:53:36 UTC). The upgrade combines the Amsterdam execution layer with the Gloas consensus layer and carries two headline EIPs: EI...
"I can just spin up a thousand builders, rotate them, offer very high bids, and not produce payloads." — Potuz, Ethereum Consensus Developer, All Core Developers Consensus Call (September 2026)
Ethereum's Glamsterdam hard fork — the network's largest protocol change since the 2022 Merge — activated on the Sepolia testnet on October 6, 2026, at epoch 353,024 (13:53:36 UTC). The upgrade combines the Amsterdam execution layer with the Gloas consensus layer and carries two headline EIPs: EIP-7732, which enshrines proposer-builder separation (ePBS) directly into consensus, and EIP-7928, which introduces block-level access lists (BALs) enabling parallel transaction execution. The gas limit rises from 60 million to 200 million — a 3.3x increase.
Early Sepolia data shows blocks using 52 million to 92 million gas (26–46% of the new limit), with 99.97% of eligible testnet stake voting toward finality. Geth's processing time for a 40-million-gas block dropped from approximately 115 milliseconds to 30 milliseconds. The Hoodi testnet is scheduled for October 26, per the October 8 All Core Developers Execution call (ACDE #247). No mainnet date is confirmed, though Q4 2026 remains the target.
This report compares the current off-chain MEV-Boost relay architecture against Glamsterdam's protocol-native ePBS system, and examines how BALs restructure sequential block execution into a parallelized model. The upgrade does not merely add capacity — it rewires how blocks are built, who builds them, and how quickly nodes can verify them.
Ethereum's block production today relies on MEV-Boost, an off-chain protocol maintained outside the core consensus layer. As of late 2025, MEV-Boost facilitated the production of over 92.5% of Ethereum blocks. The system introduces a trusted intermediary — the relay — that holds a builder's transaction payload and a proposer's signed commitment, releasing both atomically.
The relay market has shifted considerably. As of August 2026, neutral (non-filtering) relays handled roughly 70% of daily block production. Ultra Sound Relay held 35.1% of blocks, Titan Relay followed at 29.8%, and Aestus at 7.2%. Flashbots' relay, once dominant, produced just 2.4% of blocks.
However, centralization persists at the builder level. The top two builders — Titan and Quasar — constructed approximately 73% of all blocks as of mid-2026. This concentration creates several systemic risks:
The current system works, but it depends on middleware that sits outside the protocol's security guarantees.
EIP-7732 moves the proposer-builder handoff into Ethereum's consensus protocol. Under the current MEV-Boost model, the exchange between proposer and builder is mediated by a relay. Under ePBS, the protocol itself enforces the atomic swap: a builder commits a bid, the proposer signs a commitment to that bid, and the payload is revealed in a subsequent step governed by consensus rules.
Key structural changes:
| Component | MEV-Boost (Current) | ePBS (Glamsterdam) | |-----------|---------------------|---------------------| | Relay dependency | Required — trusted third party | Eliminated — protocol-native | | Payload commitment | Off-chain, relay-mediated | On-chain, consensus-enforced | | Builder accountability | Reputational only | Cryptoeconomic — slashing possible | | Censorship resistance | Relay-dependent | Protocol-level guarantees | | Validator software | MEV-Boost sidecar required | Native client support |
The elimination of relays removes an entire class of infrastructure from the critical path. Validators no longer need to run MEV-Boost alongside their consensus client. Block building becomes a protocol-level auction rather than an off-chain negotiation.
For builders, the change introduces protocol-level accountability. A builder who wins an auction and withholds a payload faces consensus-level penalties rather than mere reputational damage. This is a meaningful shift: under MEV-Boost, a builder who withholds a payload wastes the proposer's slot but faces no direct financial penalty beyond being blacklisted by individual relays.
The broader implication is that MEV management moves from social coordination to protocol enforcement. The relay operators — Flashbots, Ultra Sound, Titan, Aestus, and others — lose their intermediary role. Whether this reduces builder centralization at the construction level remains an open question. ePBS restructures how builders interact with the protocol, but it does not directly address why two builders win most auctions (exclusive order flow, superior algorithms, and private mempool access).
EIP-7928 introduces block-level access lists (BALs), which fundamentally change how Ethereum nodes process transactions within a block.
Under the current sequential model, each transaction in a block is executed one after another. A node must process transaction N before it can determine whether transaction N+1 touches the same state. This serialized execution bottleneck limits throughput regardless of available hardware.
BALs solve this by requiring block producers to include a state access map with each block. This map declares which accounts, storage slots, and contract states each transaction will read or modify, along with post-execution state differentials. With this information, verifying nodes can:
The performance data from Sepolia is notable. Geth's execution time for a 40-million-gas block dropped from approximately 115 milliseconds to approximately 30 milliseconds — a 74% reduction — largely attributable to state prefetching enabled by BALs.
This has asymmetric implications. Block producers (builders) must still execute transactions sequentially on high-performance hardware to generate accurate access lists. Block verifiers (the vast majority of nodes) benefit from parallelism, reducing their hardware requirements and improving decentralization at the verification layer.
The design draws a line between proving and verifying. The expensive work of sequencing and simulating transactions remains with specialized builders. The cheaper work of verification gets distributed across the network with lower computational barriers.
Glamsterdam raises the block gas limit from 60 million to 200 million. This 3.3x increase is not a simple capacity expansion — it arrives alongside gas repricing changes (EIP-7904, EIP-8037, EIP-8038) that alter what operations cost.
Key repricing changes:
While state creation and access costs increase, the overall fee impact is projected to be a net 78.6% reduction in L1 transaction fees due to the dramatically larger block capacity. Simple ETH transfers may become up to 71% cheaper.
For Layer 2 rollups — which carry approximately 95% of Ethereum's transactions as of Q2 2026 — lower calldata costs reduce settlement expenses. This flows through to lower user-facing fees on Arbitrum, Optimism, Base, and other L2s that post data to Ethereum L1.
Ethereum's average L1 throughput in Q2 2026 stood at 25.9 TPS. Glamsterdam's combined capacity and parallelism improvements target an eventual path toward 10,000 TPS, though this figure represents a ceiling that depends on adoption, block utilization, and future protocol work — not an immediate outcome.
The Sepolia activation on October 6, 2026, produced the following observable results:
The low utilization rate is expected: Sepolia lacks the organic transaction demand of mainnet. The relevant signal is that blocks can be produced, finalized, and verified at the higher gas limit without consensus failures.
Builder withholding attacks. Ethereum consensus developer Potuz flagged a testnet-specific risk during the September 17 All Core Developers Consensus call: under ePBS, a malicious actor on Sepolia (where test ETH is free) could spin up thousands of fake builders, submit artificially high bids, win auctions, and then withhold payloads — leaving blocks empty. On mainnet, where ETH has real value, the economic cost of such attacks is significantly higher, but the attack surface exists.
Builder centralization persistence. ePBS removes relay dependency but does not address the upstream causes of builder concentration: exclusive order flow deals, private mempools, and algorithmic advantages. Two builders controlling 73% of block production is a structural issue that protocol changes alone may not resolve.
State growth management. The 200 million gas limit significantly expands block capacity. Even with repriced state creation costs, the rate of state growth at full utilization could strain node operators. EIP-8037 introduces per-state-byte cost metrics to bound this, but the effectiveness depends on actual usage patterns post-mainnet.
Client diversity. The emergency Prysm fix before Sepolia activation highlights the fragility of client-specific implementations during major upgrades. Multi-client testing on Hoodi will be a critical checkpoint.
| Milestone | Date | Status | |-----------|------|--------| | Sepolia testnet activation | October 6, 2026 | Completed | | Client update target for Hoodi | October 16, 2026 | Pending | | Hoodi testnet activation | October 26, 2026 | Scheduled | | Mainnet activation | Q4 2026 (tentative) | Unconfirmed |
No mainnet date has been committed. Developers have indicated December 2, 2026, as a tentative target, pending Hoodi results and client readiness.
Vitalik Buterin has framed the broader context: the fork following Glamsterdam — Hegota, planned for 2027 — is "likely to be Ethereum's last 'normal' fork, with features and technology that would be recognizable to someone in 2015." Post-Hegota, Ethereum's upgrade path shifts to recursive STARKs, automated formal verification, and quantum-safe consensus algorithms.
Glamsterdam is architecturally significant because it moves two critical functions — block auction settlement and state access declaration — from off-chain coordination into the protocol itself. The relay layer that grew organically to manage MEV becomes unnecessary at the protocol level. The sequential execution model that constrained throughput regardless of hardware gives way to verifier-side parallelism.
The data from Sepolia is preliminary but directionally consistent with design goals: larger blocks, faster verification, maintained finality. The open risks — builder centralization, state growth, client bugs — are known and under active testing. The Hoodi testnet on October 26 will determine whether the upgrade is ready for mainnet before year-end.
For the economic value distribution across Ethereum's ecosystem, the implications are structural. Lower fees reduce revenue per transaction for validators and builders. Higher throughput increases total addressable volume. The elimination of relays removes an entire category of intermediary from the value chain. Whether builders, validators, or users capture the net economic benefit depends on market dynamics that the protocol change sets in motion but does not determine.