Ethereum developers fork the Sepolia testnet on Tuesday, October 6 at 13:53:36 UTC, and the upgrade matters even though nothing happens to your ETH that day. The Glamsterdam activation adds enshrined proposer-builder separation and block-level access lists, the two changes that let the network test a gas limit three times larger than today’s. The mainnet date is still not set.
The fork lands at epoch 353,024, slot 11,296,768, according to the Ethereum Foundation’s announcement. Hoodi and mainnet activation slots remain listed as TBD, and the roadmap page pegs mainnet only as expected sometime in the fourth quarter of 2026 with no confirmed date. Sepolia is where the client teams find out whether their code survives contact with real-world load.
The point of Glamsterdam is throughput. Today the block gas limit sits around 60 million. Glamsterdam introduces EIP-8261, a gas limit schedule, which lets client defaults recommend 200 million per block after the Gloas consensus change activates. That is roughly triple current capacity. Sepolia’s planned preference is 200 million, but the number is a test target, not a confirmed mainnet floor. Developers are deriving state pricing at a 150 million reference value, which suggests the eventual mainnet number may start lower than the marketing figure.
Nothing switches instantly either. EIP-8261 changes no consensus rules. Validators who want to propose 200 million blocks must configure it themselves. The Prysm 7.2.0 client defaults to 60 million after the fork, and the –suggested-gas-limit flag stops having any effect after Gloas, so stakers have to use version 2 proposer settings or the keymanager API. Teku 26.9.1 behaves the same way and requires an explicit –validators-builder-registration-default-gas-limit=200000000. The 1559 elasticity rule of plus or minus 1/1024 still means the realized limit drifts block by block, so validators who set a lower preference will simply drag the network average down over successive blocks.
Why block access lists unlock the higher limit
The piece that makes a bigger limit workable is EIP-7928, block access lists. Each block now carries a map, written into the block header as a hash, listing every account and storage key its transactions will touch along with the post-execution state values. Validators no longer re-derive database access from scratch; they can read the map and verify instead of guessing. Average compressed lists run about 72 KiB at a 60 million limit, and their size scales with the gas limit rather than a fixed cap, so the overhead grows as capacity does.
Because the lists are big, the network needs new ways to move them. EIP-7975 adds partial block receipt lists on the eth/70 protocol, letting nodes split the payloads into smaller pieces, and eth/71 adds a block access list exchange mechanism. Analyst summaries note the wider data-propagation window, roughly from two seconds to around nine seconds, gives validators the headroom to accept larger blocks without racing the clock. Parallel execution is the payoff: distinct transactions touching different storage can be validated side by side instead of serially.
Enshrined PBS changes who builds blocks
The second headliner, EIP-7732, enshrines proposer-builder separation in the protocol. Today that separation is a market convention run through MEV-Boost relays, trusted middleware with its own censorship and take-rate debates. Glamsterdam moves the auction into consensus, so block proposers commit to a builder’s block header and the constraint is enforced by the fork itself. It removes a middleware dependency and, in principle, opens builder competition to anyone without running a relay.
Price context: an upgrade nobody prices
ETH trades near 2,500 dollars, about 45 percent below its 4,946 dollar peak, and coverage of the fork leans heavily on the same caveat: October 6 is a testnet date and demands no action from holders. No wallet migration, no token swap, no deadline. Where the money story sits is elsewhere: BitMine, the largest corporate Ether holder, has staked 5,067,309 ETH, 84 percent of its 6 million token stack, projecting 257 to 358 million dollars in annualized staking revenue depending on the week it published the figure. And the MetaMask security incident pushed around 17,000 validators, roughly 523,000 ETH, into the exit queue, which now sits at a nine-month high as those validators unwind.
| Network | Epoch | Slot | UTC date | Status |
|---|---|---|---|---|
| Sepolia | 353,024 | 11,296,768 | October 6, 2026, 13:53:36 | Scheduled |
| Hoodi | TBD | TBD | TBD | Pending |
| Mainnet | TBD | TBD | Q4 2026 expected | No date |
Small print matters for node operators. Beyond the gas settings, Glamsterdam includes EIP-7954, a higher maximum contract size, and EIP-7976, an increase in the calldata floor cost, both aimed at keeping blob and calldata economics honest as the limit grows. Anyone running a Sepolia execution or consensus client must upgrade both layers before October 6, and client releases for the fork are already published.
There is a oyun too. EIP-8363, a proposal to burn part of staking rewards, was pulled from the Hegota upgrade, the next fork after Glamsterdam. Contributors argued the fork-scoping process was the wrong venue to settle issuance policy, and the fight was deferred rather than won. Hegota is aimed at 2027, so the issuance debate moves there.
What happens next is mostly boring and that is the point. Sepolia forks Tuesday, some validators propose at 200 million and others at their defaults, testers watch for failures in state creation pricing, and the foundation updates the Hoodi slot once it sees results. The mainnet window in Q4 gives the client teams a couple of months if the testnet behaves, and scrapes the schedule if it does not.