Ethereum’s largest protocol upgrade since the Merge gets its first public trial this week. The Glamsterdam fork activates on the Sepolia testnet on Tuesday, October 6, 2026, at epoch 353,024, slot 11,296,768, which lands at 13:53:36 UTC. The Ethereum Foundation announced the activation date in late September after months of devnet testing, and mainnet is targeted for the fourth quarter of 2026 with no confirmed date yet.
Glamsterdam carries two headline changes chosen through Ethereum’s first formal round of upgrade headliners. The consensus-side feature is Enshrined Proposer-Builder Separation, known as ePBS and specified in EIP-7732. The execution-side feature is Block-Level Access Lists, or BALs, specified in EIP-7928. They ship together with a supporting cluster of smaller proposals covering gas repricing, state costs and quality-of-life changes.
Both features have been discussed for years, and neither survived to a devnet in one piece the first time it was attempted. Their arrival together on a public testnet marks the end of the longest-running design argument in Ethereum’s current roadmap.
What ePBS actually changes
Today, the validator chosen to propose a block hands the job of building that block’s contents to an external builder through MEV-Boost, a piece of software outside the protocol. Proposers trust middleware, builders trust validators not to punish them, and the foundation runs a relay infrastructure that neither side fully controls. ePBS pulls this handshake directly into the consensus layer. A proposer includes a builder’s signed commitment to an execution payload, the builder later reveals the payload, and the protocol itself guarantees the builder gets paid.
The practical effect is a wider window for data propagation. Under the current design, a block’s payload has to arrive within about two seconds of the proposal for the network to stay in step. With ePBS, that window stretches to roughly nine seconds, since validators can attest to the consensus block before the payload lands. Ethereum Foundation researchers say the larger window is what unblocks the next round of throughput increases on layer 1 and a heavier schedule of data blobs for layer 2 networks.
It also removes a political dependency. MEV-Boost relays have been a chokepoint for censoring compliance during the past two years, and moving the payment guarantee into the protocol shrinks the role a relay can play. It does not eliminate out-of-protocol relays, but it makes them optional plumbing instead of load-bearing infrastructure.
Block-Level Access Lists and parallel execution
EIP-7928 attaches a map to every block. The map lists every account and storage slot the block’s transactions will touch, along with the state values they produce after execution. Clients read that list up front and can then execute the block’s transactions in parallel instead of streaming through them sequentially.
This is the feature Ethereum researchers have been chasing since the first proposal for a wheel-and-spoke access list model in late 2024. Execution race retrieval through EIP-7843 ran into contention overhead on busy devnets and was shelved. BALs avoid the race entirely: because the access list is written into the block itself and checked against a root hash, every node receives the same deterministic input.
Ethereum Foundation researcher Toni Wahrstatter, one of the main authors of the proposal, said in a post on X that “Glamsterdam with EIP-7928 is approaching the final mile, and things look great.” The assessment refers to devnet testing, not a shipped product, but it tracks the project’s own schedule. The Ethereum Foundation’s roadmap page for the upgrade lists the Sepolia fork as the next milestone and mainnet activation in the fourth quarter.
What else ships in the fork
The bundle carries five smaller EIPs alongside the two headliners. EIP-7843 adds a SLOTNUM opcode for smart contracts to read the current slot. EIP-7954 raises the maximum contract size cap. EIP-7976 raises the calldata floor cost to make oversized call bundles more expensive. EIP-7981 raises the cost of access-list usage to reflect state pressure. EIP-8037 raises the gas cost of creating new state objects.
| Proposal | Scope | One-line effect |
|---|---|---|
| EIP-7732 | Consensus | Enshrines proposer-builder separation, widens the payload window to about 9 seconds |
| EIP-7928 | Execution | Block access lists enable parallel execution and executionless sync |
| EIP-7843 | Execution | SLOTNUM opcode for contracts to read the current slot |
| EIP-7954 | Execution | Raises the maximum contract size cap |
| EIP-7976 | Execution | Raises the calldata floor cost |
| EIP-7981 | Execution | Raises access-list usage cost |
| EIP-8037 | Execution | Raises state-creation gas cost |
Why the gas-limit debate refuses to settle
BALs change the calculus for raising the block gas limit, and the debate around it has been one of the more visible tensions in the lead-up to the fork. A higher gas limit with parallel execution means more work per block and higher hardware demands for home stakers. Layer 2 advocates want the limit up aggressively. Solo stakers have pushed back on the hardware curve for months.
The Ethereum Foundation’s guidance has settled on supporting a gas-limit floor of 200 million after the fork, which is the position public researchers have been communicating. That is more than three times today’s capacity, but not the levels some layer 2 teams wanted for the first Glamsterdam-adjacent bump. Expect the argument to keep moving as BAL performance data arrives from Sepolia.
The upgrade bundles two features that each failed at least one earlier devnet attempt in a different form. Their arrival together on a public testnet is the clearest sign yet that the engineering risk has moved from design to deployment.
What happens this week
The Sepolia activation is a replay of the network’s last full testnet cycle before Holesky, with the same structure: a coordinate swap, activation at a named epoch, and a waiting period for the client teams to confirm health before mainnet gets a date. Client release tables published with the Foundation’s announcement list the versions validators and node operators must run before the fork activates.
Sepolia node operators who do not stake should update both execution and consensus layer clients before October 6, 13:53:36 UTC. Stakers on Sepolia have the same obligation plus the usual caveat that a missed activation means their validator will continue attesting on the pre-fork chain until it cycles off the fork boundary.
Mainnet timing is the open question. Hoodi activation has not been scheduled publicly. If the Sepolia fork runs clean on October 6, mainnet deployment inside the fourth quarter remains the working assumption of the researchers involved. A bumpy testnet would push the date toward the second half of Q4 without changing the protocol content.
What would count as a problem? A fork boundary that fails to finalize would trigger the path Ethereum has rehearsed before: client teams dig into logs and advocate a recovery on the basis of chain state, as happened in earlier testnet incidents. Less dramatic issues include trigger misfires, where the fork activates cleanly but clients need a follow-up release, or client disagreements on edge cases around the new BAL root format.
For markets, the event is a governance milestone rather than a price event. Ethereum has historically seen modest volatility around hard forks without any reliable directional pattern once the uncertainty window closes. Traders tracking the upgrade cycle should treat Sepolia as an engineering checkpoint and the eventual Hoodi and mainnet schedule as the catalysts that matter.