Ethereum will turn on the Glamsterdam upgrade on the Sepolia testnet on Monday, October 6 at 13:53:36 UTC, the first public testnet fork for a package that includes protocol-level proposer-builder separation, block-level access lists and repriced gas rules. Sepolia will signal a block gas limit of 200 million at the fork, up from roughly 60 million, though that figure stays a test setting and no mainnet date has been set.
The Ethereum Foundation scheduled the fork for epoch 353,024 at slot 11,296,768. Node operators running Sepolia validators must upgrade their clients by Sunday, October 5, or they will drop off the testnet. The foundation’s roadmap page puts mainnet activation in the fourth quarter of 2026 with no confirmed date, and the Hoodi testnet fork date is still undecided.
| Milestone | Date |
|---|---|
| Sepolia client upgrades due | October 5, 2026 |
| Sepolia fork | October 6, 2026, 13:53:36 UTC |
| Hoodi fork | Not yet scheduled |
| Mainnet | Expected Q4 2026, no confirmed date |
What changes in Glamsterdam
Two proposals sit at the center of the upgrade. ePBS moves the block-building pipeline into protocol rules. Today proposers who want an external builder to assemble their block rely on off-chain agreements through services like MEV-Boost; under ePBS those commitments become protocol objects, so the arrangement is enforced by the chain rather than by trust. That closes a gap regulators and researchers have flagged for years: an outside-protocol market worth billions of dollars a year in priority fees had no direct representation in consensus rules.
Block-level access lists, the second piece, make each transaction declare the state it will read and write, which lets execution clients process transactions in parallel instead of one at a time. That is the mechanical change behind the scaling story. Blocks today execute serially because of state conflicts, so a bigger gas limit on its own gains little. With access lists declared up front, a client can split a block across threads and use the extra capacity.
The third piece is repricing. The roadmap notes describe EIP-8037, the “reservoir model”, which splits gas accounting between active computation and permanent state storage. Without that split, state growth at a 200 million gas limit would become unsustainable as the network scales, because running a computation and saving a contract to the database currently draw on the same budget.
What the rehearsals showed
Recent devnet rehearsals processed Glamsterdam’s new block-building rules while pushing gas limits up to 200 million, according to CoinDesk. The foundation credits ePBS, access-list optimizations and the new state-pricing rules with making a post-fork gas limit floor of 200 million credible. Testers also examined how the network propagates blocks that large before moving to the activation step itself.
A 200 million gas limit would more than triple the current throughput budget. Larger blocks mean more transactions before congestion lifts fees on the base layer. The tradeoff runs the other way for validators and node operators, who have to store and broadcast more data per block. Home stakers on slow connections are the group most exposed, and that concern is part of why the foundation treats the number as a test configuration rather than a settled mainnet parameter.
Where ether stands
Markets have mostly treated the fork as a technical milestone. Ether traded near $2,700 with buyers holding a directional edge on momentum trackers, and chart analysts put the next resistance near $2,749 on a four-hour close, with $2,800 and $2,900 beyond that. Bitcoin, meanwhile, retook $86,000 on renewed spot ETF demand, with $102.7 million of net inflows on October 1, giving the broader tape a calmer backdrop than ether has had for most of the month.
Separately, The Block reported the Ethereum Foundation launched zkAPI, an experimental tool that lets users pay for AI models without revealing their identity. It is a small launch next to a network upgrade, but it keeps attention on ether’s infrastructure pipeline in the same week the scaling fork reaches its first public testnet.
One testnet event that preceded the fork announcement is worth watching alongside it. MetaMask pulled 17,000 staking validators holding about 523,000 ETH after a breach redirected validator fee rewards, which pushed the exit queue to a nine-month high near 773,000 ETH with an estimated 13-day wait. Queue pressure of that size eases the load on block space in the short run, and it complicates planning for anyone who wants liquidity out before mainnet changes arrive.
What stakers should do
The practical ask for stakers is to test now. The foundation’s guidance says validators can check their setup by running a validator on testnets before mainnet, with testnet forks announced through the mailing list and blog. Sepolia landing on a Monday gives client teams the weekend to patch whatever the latest client releases surface in practice, and it leaves runway for a Hoodi pass before the mainnet fork window narrows.
Client diversity matters here more than in past upgrades. The access-list proposal touches execution clients hardest, and a bug found in one client rather than another is exactly what testnets are meant to surface. A chain checked at 200 million gas has far more room for that kind of problem to show.
Vitalik Buterin has signaled that the era of routine small forks is winding down, and CCN coverage framed Glamsterdam as a break from single-feature releases. What matters operationally is narrower: whether Sepolia holds the 200 million gas limit without fee spikes or chain reorganizations, and whether client teams agree on a Hoodi and mainnet schedule afterward. Both calls land in the next few weeks.
The Sepolia activation is set for October 6 at 13:53:36 UTC, with client upgrades due by October 5. Hoodi comes later, and mainnet sits in the fourth quarter with no confirmed date. The mainnet gas limit will be set only after developers see how the testnet handles blocks much larger than today’s, and anyone running a mainnet validator should follow the client release notes from here rather than assume the testnet numbers carry over.