Ethereum Glamsterdam: On October 6 Sepolia Forks, Not the Mainnet, the Next Step for Your ETH

Update (November 1, 2026): Glamsterdam has been running on the Sepolia testnet since October 6, 2026. The next step, activation on the validator testnet Hoodi on October 26 at 17:42 UTC (epoch 132,352), is not happening for now: on October 22 the Ethereum Foundation announced that the core developers had postponed the upgrade there in order to look more closely at how the new storage pricing from EIP-8037 interacts with existing smart contracts on the mainnet. There is no new date for Hoodi. A mainnet date has not been set either; the roadmap page still names the fourth quarter of 2026 with the note that no date is confirmed. According to the Foundation, ETH holders have nothing to do.

Ethereum Glamsterdam did not start on the mainnet on October 6, 2026, but on the Sepolia testnet. For your ETH on an exchange or in a wallet that day therefore meant precisely nothing: no swap of balances, no new address, no pressure to act. When Glamsterdam will reach the real Ethereum network is open. To this day the Ethereum Foundation names only the quarter.

That mattered because on September 30 several financial outlets claimed something else. Anyone searching for the upgrade that day found headlines saying Glamsterdam had already launched. This piece separates the two: what actually happened on October 6, what Glamsterdam changes technically, and how you will recognise the real mainnet date when it comes.

Sepolia fork on October 6, 2026: a testnet, not a mainnet launch

The developers activated the Glamsterdam fork on Sepolia on October 6, 2026. Sepolia is one of the public testnets of Ethereum: a complete image of the network running the same software, but in which no money of any value moves. The ETH on Sepolia is play money and is handed out from a faucet, not bought.

Testnet in this context means an independent blockchain whose sole purpose is to try out protocol changes under real conditions before they reach the main network. If something goes wrong on Sepolia, nobody loses assets. That is exactly the point. Anyone running a node there had to update both the execution client and the consensus client before activation; the Ethereum Foundation has pointed that out explicitly.

A fork on a testnet is the normal case before every larger Ethereum upgrade and not an event that touches the price or the availability of ETH. Our earlier article on the originally scheduled Sepolia date still named September 28; that date passed, the activation came on October 6.

The mainnet date is not fixed: the Ethereum Foundation names only the fourth quarter

On the official Glamsterdam page of the Ethereum Foundation, the upgrade status reads Testing on devnets, the expected mainnet activation reads Q4 2026, and behind it, in the same field, sits the addition Date not yet confirmed. At the end of September the only next milestone listed there was the Sepolia fork on October 6, 2026; since then it has shown the Hoodi fork, which the Foundation postponed on October 22.

For Hoodi, the testnet tailored to validators that usually comes after Sepolia, there is currently no date either: the fork scheduled for October 26 was postponed on October 22, and a new date is still pending. That leaves at least one intermediate step missing before a mainnet date can be set at all. Anyone reading a specific calendar day for the mainnet activation today is reading an estimate, not a commitment.

Mainnet is the productive Ethereum network in which real ETH moves and on which the exchanges, wallets and applications depend. Only an activation there touches balances, fees and applications. Everything before it is preparation. In our overview of crypto dates in the fourth quarter the Sepolia fork therefore carries a date and the mainnet activation does not.

Devnet, testnet, mainnet: the three stages in the Ethereum roadmap

An Ethereum upgrade passes through three stages, and the order explains why October 6 is not a launch date.

Devnets are short-lived developer networks set up specifically for an upgrade and switched off again afterwards. This is where Glamsterdam stood, according to the Ethereum Foundation, until the Sepolia fork. At this stage the interplay of the different client programs is tested, meaning the independent software implementations with which participants run the network.

Public testnets such as Sepolia and Hoodi are the second stage. Networks like these run permanently, carry real applications in test form and have a public audience. A fault shows up here before it costs money.

Mainnet is the third and final stage. The date for it is set only once the testnets have run through cleanly. That sequence is the reason the Foundation names a quarter and not a date: between October 6 and a mainnet activation lies at least one further testnet fork whose own date is still open.

ePBS: proposer-builder separation moves into the protocol

The most conspicuous change in Glamsterdam is called enshrined proposer-builder separation, ePBS for short, filed technically as EIP-7732. Today an Ethereum block is built in two roles: a proposer, meaning a validator whose turn it is, proposes the block; a builder assembles the content. That handover has so far not been part of the protocol but runs over intermediary software from third parties, which both sides have to trust.

ePBS pulls that handover into the protocol itself. The builder's payment is thereby settled without trust, and intermediary software such as MEV-Boost is no longer needed for the basic case. At the same time the Foundation points out that builders and proposers can continue to use external services voluntarily, as long as they need functions the protocol does not cover.

The practical gain sits in one number: the window in which a block can propagate through the network grows, according to the Ethereum Foundation, from around two to about nine seconds. That allows considerably larger volumes of data to be carried per block without overloading the network. For staking providers the change means adjustment work on their own architecture; the Foundation describes the staking user experience afterwards as improved, because the selection of the builder is regulated more cleanly.

Block-level access lists: the map for parallel processing

The second large change is block-level access lists, BALs for short, technically EIP-7928. Today Ethereum works through transactions strictly one after another. The reason is banal: the network does not know in advance which data a transaction will touch. If it were to execute several at the same time, two of them might want to change the same place.

A block-level access list is a kind of table of contents that records, before execution, which accounts and storage slots a block touches and which values end up there. From that list you can read off which transactions do not get in each other's way. Such transactions can then be processed simultaneously. The block header carries only a checksum of that list, so that the block itself stays lean.

A second benefit concerns new participants: because the list also contains the final values, a freshly started node can adopt the network state without recomputing every transaction again. The accompanying EIP-8159, filed in the network protocol as eth/71, makes sure the nodes can send each other these lists, and it is mandatory for all execution clients.

EIP-8037: new gas prices for permanent storage in the reservoir model

The third construction site is unspectacular and, for application developers, the most consequential. Data that sits permanently on Ethereum has to be held by every computer in the network for an indefinite time. The fees for it have so far reflected that effort only imprecisely. Rolling out a large smart contract is considerably cheaper per byte than writing to individual storage slots.

EIP-8037 ties the cost to the actual volume of data and sets a fixed price per byte. The aim is a predictable growth of permanently stored data of around 120 gibibytes a year, so that Ethereum can continue to run on commodity hardware. On top of that comes a separate pot: computation and permanent storage no longer share the same gas limit. Developers can therefore roll out larger applications, as long as they fill the storage pot sufficiently.

For existing applications this is the point with the greatest need for testing. Contracts working with hard-wired gas amounts may react differently after the change; the Ethereum Foundation has pointed that out explicitly, according to Cointelegraph. That is exactly what the testnet stage is for. This is the point at which the developers paused on October 22: they postponed the Hoodi fork to examine more closely how EIP-8037 interacts with existing mainnet contracts.

ETH on an exchange: the Sepolia fork does not touch your balance

The most common worry with an Ethereum upgrade is whether balances have to be swapped, frozen or moved. With Glamsterdam the answer for October 6 was unambiguous: nothing changed on the mainnet that day, because the upgrade was not activated there.

The later mainnet activation is a protocol upgrade too and not a currency conversion. No new tokens arise, ETH keeps its addresses, and holders have nothing to apply for or confirm. What can actually happen with a mainnet fork are short maintenance windows at individual providers: some trading venues halt deposits and withdrawals for a few hours while their own nodes run through the fork. Trading itself usually stays open.

Anyone wanting to know how their own provider handles this will find the announcement in the status area or the blog of the trading venue, usually a few days before the date. How the larger trading venues differ on fees, custody and authorisation is shown by our comparison of crypto exchanges. As long as no mainnet date is set, you need to do nothing there.

Stakers and node operators: Sepolia requires both clients up to date

One group was very much in demand on October 6: everyone running a node or a validator on Sepolia. There both client programs had to be updated before activation, the execution client and the consensus client. Anyone who missed that dropped out of the valid chain after the fork. Anyone running a node on Hoodi should, according to the Ethereum Foundation, now install client releases whose release notes explicitly confirm that the October 26 activation has been removed.

The same task is coming on the mainnet as soon as the date there is set, and then for all validators. Which version levels fit together and where the pitfalls lie we have written up in the article on client readiness for Glamsterdam. The procedure stays the same, only the version numbers move on.

Anyone staking ETH not themselves but through a provider has nothing to do with the client question. There the operator handles the update, and the reward keeps running. More relevant for that group is the ePBS change, because staking providers have to adapt their infrastructure to it.

Three provable signals for the real mainnet date

A mainnet date does not appear out of nowhere. It comes about in a fixed order, and a news report can be measured against it.

First, the Hoodi fork. As long as no date is set for Hoodi, the testnet tailored to validators, a mainnet date is not plausible. That step comes between Sepolia and the main network. Since the postponement of October 22, that is exactly the situation again.

Second, the status line on the roadmap page of the Ethereum Foundation. It carries the upgrade status, the expected mainnet activation and the next milestone. Once the addition Date not yet confirmed turns into a date, the date is set. As long as it sits there, there is none.

Third, the epoch number. An Ethereum fork is not activated by clock time but at a fixed epoch, meaning a numbered section of the chain. A solid announcement names that number. A report without an epoch and without client versions describes a plan, not an activation.

Headlines of September 30: what is wrong with “starts today”

On September 30, 2026, several German-language financial outlets wrote that the Glamsterdam upgrade was starting that day, in some cases tied to a price target. Neither claim holds up against the primary source: no fork took place on September 30, and the mainnet activation has no date.

A second point concerns dates circulating online. At the end of September individual specialist sites carried a specific date in November 2026 for the mainnet activation of Glamsterdam as confirmed. The roadmap page of the Ethereum Foundation contradicted that on the same day with the note that no date is confirmed. We follow the primary source here and therefore name no calendar day.

For classifying upgrade reports, one simple question helps: does the text name a network? Reports describing Glamsterdam without the addition Sepolia, Hoodi or mainnet leave out the decisive difference. Out of that comes the impression of a launch that does not exist in that form.

After Glamsterdam comes Hegotá: the developers are aiming at 2027

Glamsterdam is not the end of the roadmap. The following upgrade carries the name Hegotá and is, according to Cointelegraph, targeted at 2027. Ethereum co-founder Vitalik Buterin has described it as possibly the last ordinary fork before development shifts more strongly towards cryptography topics such as recursive STARKs, automated formal verification and quantum-safe methods.

The name follows the same logic as Glamsterdam, which is put together from Amsterdam for the execution layer and Gloas for the consensus layer. Anyone looking for the wider arc will find it in our article on what has already been decided for Hegotá.

For classifying Glamsterdam that means: the changes to block production, parallel processing and storage prices are preparatory work for higher gas limits. The Ethereum Foundation names a lower bound of 200 million gas per block as the target figure, while testing currently runs with a reference value of 150 million.

Glamsterdam: What to take away

October 6, 2026, was a testnet date. For ETH holders it was a date to note, not a date to act on, and the same goes for the postponement on Hoodi. Three steps are worth taking anyway.

1. Settle where your ETH sits. With a later mainnet activation a provider may briefly halt deposits and withdrawals, while self-custody remains untouched by it. Which devices come into question and what they cost is in the hardware wallet comparison.
2. Look at your staking chain. If you stake yourself, both clients belong on the current version before every fork. If you stake through a provider, a glance at its status page is enough; the differences between providers are shown by the overview of staking platforms.
3. Measure future reports against the primary source. A mainnet date is only one once it stands on the roadmap page with an epoch and client versions. Anyone looking for a trading venue in the meantime is best off comparing fees and authorisation through the overview of crypto exchanges.

(As of September 30, 2026, updated on November 1, 2026. This article is not investment advice. Prices and fee structures change; check the terms with the provider before you buy.)