Glamsterdam-finalized: Ethereum will fork off Sepolia on September 28, 2026

Glamsterdam-first confirmed date

September 28, 2026 14:44:48 UTC, the next major upgrade to Ethereum will be activated on the Sepolia test network. This time point comes from the minutes of the developer meeting on August 20, 2026, and is accurate to the specific epoch and slot. The six client teams unanimously agreed during the meeting chat, and no one objected.

Two prerequisites must be stated at the same time, otherwise the plan will become a commitment. First of all, this date is tentative: it is clearly stated in the minutes that the final decision will be postponed until the next meeting. Secondly, this only applies to test networks and has nothing to do with the main network. There is still no specific date on the main website. Since August 20, the numbers circulating on the Internet are no longer reliable.

This article will show you where this date came from, how to recalculate it on its own, what factors may change it, and what it means to you as an ETH holder. For information on what changes Glamestead actually includes and the filtering mechanism for the next upgrade, please refer to our article "Ethereum Upgrade After Glamestead: What Hegotá Decided". This article only focuses on the timetable.

When will Glamestead come? First date appears in core developer notes

Ethereum core developers coordinate upgrades through public video conferencing, known as the "All Core Devs Calls". The ACDC meeting is a consensus layer version meeting, which is the part where the consensus client team negotiates time, vulnerabilities, and testing the network. There are minutes of the meeting, and the minutes and chat records will then be posted in the project planning warehouse.

The key sentence appeared in the ACDC #185 meeting minutes on August 20, 2026. Under the heading "Forking Status", the record reads: Sepolia Glamstam Forking Proposal: Era 351232, Slot 11239424, September 28, 2026-No objection was raised; decision was made to postpone until the next ACDC. The same line appears again in the target list of the same document, with the annotation "Pending next ACDC confirmation".

Chat logs allow you to track this moment minute by minute. At 27 minutes and 50 seconds, Parithosh Jayanthi, who was in charge of the Ethereum Foundation's test network coordination, wrote the full value into the chat: Sepolia: Era 351232, timeslot 11239424, Unix timestamp (1790606688), Monday, September 28, 2026, 14:44:48. Over the next 60 seconds, six participants responded to the message with boat emoticons-the usual way of expressing agreement in these meetings: representatives from Lodestar, Ethrex, Lighthouse, Nethermind, Teku and Prysm. Soon after, Barnabas Busa recorded how to interpret this situation: No objection = agree.

This makes this number different from other dates circulated. It is not an observer's estimate or a team's wish list, but a proposal from the coordinator, and the client team does not object in the record. However, this is still not binding.

What eras, slots and Unix timestamps mean for the Sepolia fork

A slot is a fixed 12-second window during which exactly one block can be raised on Ethereum. A era is a set of 32 such time slots, a window of 6 minutes and 24 seconds. Upgrade depends on the beginning of the era, not a certain time on the calendar. This is why three numbers are always listed at the same time in the minutes.

You can understand the arithmetic behind it in two steps. The time slot number is equal to the era number multiplied by 32: 351,232 multiplied by 32 yields 11,239,424, which is exactly the number in the minutes. The timestamp comes from the start time of the Sepolia beacon chain plus 12 seconds per slot. The result is a Unix value of 1,790,606,688, which corresponds to 14:44:48 UTC on Monday, September 28, 2026. In German daylight savings time is 16:44:48.

In effect, this means an issue that headlines often overlook: clock time is a derivative value, not an agreed time. It only works when the network generates blocks on time. If a block fails to appear, the actual activation time is delayed without anyone having to change the date.

Sepolia, Hoodi, Mainnet: Why does Ethereum hard fork happen three times

Hard fork Is a backward incompatible change to protocol rules: software that does not accept updates will not be able to follow the chain later. Because mistakes here are costly, each change will be rolled out in stages. First on the internal test network, then on the public test network, and finally on the main network.

A test network is an independent chain running the same software but using worthless tokens. Sepolia is a test network for application developers to check their contracts;Hoodi is a newer network that targets validators and pledge infrastructure. Both have their own items in the meeting.

For Hoodi, the same source gives the era 132352, time slot 4,235,264, and the Unix value 1,793,036,568, which is 17:42:48 UTC on Monday, October 26, 2026. However, this line is more unstable than Sepolia's: the minutes attach the qualification of "depends on a stable development network", and in the chat, Lighthouse representative kingy_sigp replied: "As long as we have a stable development network, this date is fine for us."

For the main network , there is not a single line in the minutes. There is no mainnet date for both decisions and goal lists. Anyone who reads any main-network date today will read predictions that are not backed by decisions.

DevNet8 failure: builder deposit cache error leads to final loss of property

A Developer Network It is a short-term test network set up by developers specifically for a certain upgrade phase and usually only runs for a few days. Six hours before the August 20 meeting, the Gloas phase was activated on DevNet 8, and problems occurred during the process.

minutes record the event and its cause in one sentence: activation triggered non-finality , which was just restored by the time of the meeting due to a cache error that affected the builder's deposit during the fork transition. Three consensus clients were affected: Lighthouse, Prysm and Teku.

A client is software for nodes to participate in the network. Ethereum is deliberately implemented in parallel by multiple independent teams so that mistakes in one piece of software do not drag down the entire network. This is the mechanism that works here: the error affected three implementations, the network was restored, and countermeasures were proposed during the meeting.

The minutes specifically mentioned two of these measures. In Lodestar, deposit pre-processing has been optimized to shorten the duration of the fork transition from approximately 20 seconds to approximately 500 milliseconds; the change is already in the main branch. In Prysm, the builder deposit cache was removed from the branch before the fork, and an additional optimization allows recovery to be shortened to a single slot even without the cache.

What does finality mean and why losing it slows down Glamsterdams schedule

Finality It describes the state where a block is considered settled and can only be reversed through a large amount of pledge. If ultimately lost, the chain does continue to run and create blocks, but no block is closed. This is the most embarrassing situation for exchanges, bridges, or pledge agreements, because there is no secure way to determine which version is correct.

The chat contains an argument about this judgment, which is worth reading because it clearly separates the two perspectives. Barnabas Busa argued from a protocol perspective: "We may also encounter non-finality on the main network. Why is this a problem? We have no consensus issues." Dima Gusakov from pledge provider Lido responded from an operator's perspective, saying that his team was unable to check how its oracle and tools passed through hard forks precisely because the authenticity was lost.

The severity of the situation can be seen in a comment by the same Barnabas Busa later in the meeting: "I think it's safe to say that if we have 4-5 months of non-finality, it's over." This comment refers to a completely different proposal on data retention, but it shows the magnitude of thinking here.

Before testing the network, there is testing the network: Sepolia and Hoodi's dates will be closer only after the short-term development network is stable.

Lido and Optimism require a test window, which is the only objection on record

There are no formal objections to this date, but it is not unconditional. The two operators made clear their requirements during the meeting and recorded them in the minutes by name.

Lido, according to Dima Gusakov, needs to be run on a stable development network for at least one day, and preferably two days, to check for the oracle and tools before Glamestead is activated. Optimism, via Chris Berry, requested a similar window and made it clear in the chat: "If we knew, we could do it for 1-2 days." There is also a request to announce the fork of the development network one week in advance so that verifiers can go online in time.

Gusakov was more explicit on the date itself. Shortly after Hoodi's line, he wrote in a chat: "I would rather have both dates a little later. Let's say about a week to make more room for pending Dev.com tests." Two other participants, Miguel Tenorio and Chris Berry, marked the message and agreed. This is not a veto, nor does it block the proposal, but it is the only recorded voice that believes September 28 is too early.

Anyone who wants to set this date now has information on both sides. Those who supported it were the approval of six client teams and the conclusion that "no objection was raised" in the minutes. The opposition is that the network's largest pledge operators want more buffers, and Hoodi's line clearly relies on a stable development network.

September 3 is the date behind the date: ACDC meeting #186 will make a decision

The key wording in the minutes is the half sentence "Decided to postpone until the next ACDC." That makes September 28 a target that still needs to be voted on. So the real question is: When will this goal become a decision?

These meetings are held every two weeks. The 184th meeting was held on August 6, and the 185th meeting was held on August 20. So the next meeting is September 3, 2026 . Until then, the date remained a widely supported proposal; after that, it was either confirmed, postponed or split.

This has a practical use for you as a reader. If you read between September 3 and September 5 that Glamsterdamm now has a date, that's a confirmation of the offer, not news. If there is a date different from September 28, it means that there was a change in the meeting, and it is worth checking the minutes at this time.

Why circulated main-network dates (such as November 4) are no longer reliable

English calendar aggregators have been listing three numbers for Glamestead for weeks: September 21, Hoodi October 5, and main-network November 4, 2026. On relevant pages, most of these entries are marked as predictions based on early goals and have no decision-making support.

As of August 20, all three figures were outdated. Sepolia's numbers were delayed by one week and Hoodi's by three weeks. The main network fork on November 4 was only nine days later than the Hoodi fork on October 26. On Ethereum, it usually takes several weeks to separate the second public test network from the main network, and the activation is observed and evaluated over these weeks. Nine days were extremely stressful for this step.

The German reports in mid-August were still in the pre-meeting state. There, GlamAmsterdam is an upgrade that has been delayed until the fourth quarter of 2026 and is lagging behind. This was correct as of August 19, but has not been complete since then because of the lack of lines to test the network. If you find yourself sorting decisions based on dates picked from titles, it's worth comparing to original sources. Anyway, people comparing providers can find the current terms in our Best Cryptocurrency Exchange Comparison .

EIP-7610 has been removed from Glamestead rather than postponed

EIP is an Ethereum improvement proposal that is a formal proposal for a single technology change. Which EIPs can be upgraded are decided by the core developers in these meetings, and the list is variable before activation.

During the August 20 meeting, one participant asked about the status of EIP-7610, and the proposal had been removed from Glamsburg's scope: whether it would be postponed or cancelled completely. EIP editor Jochem Brouwer's answer in the chat was clear: "Answer clearly and cancel. I will also ask the author to formally withdraw this EIP." This is also mentioned in the minutes under "Decisions" and the reason is given: EIP-7610 has been replaced by EIP-8253, which is targeted at Hegotá, a subsequent upgrade to Glamestead.

The difference between postponements and cancellations sounds academic, but it has an impact on the schedule. A postponed EIP will appear on the same list again in the next round. A cancelled EIP has ended and a different proposal will replace it.

Hegotá welcomes the first round of screening: EIP-833 and EIP-8379

While Glamestead moves towards testing the network, screening for the next upgrade also begins at the same meeting. According to the minutes, two proposals are considered tentative arrangements: EIP-833, which fixes off-one errors in checkpoint root calculations to improve the accuracy of target voting; and EIP-8379, which aims to make consensus clients the only source of synchronization, thereby reducing duplicate block propagation.

The third proposal, EIP-12188 on shortening the consensus block retention window, was neither on the list nor in the trash: it was returned to research and discussion. In response, the minutes recorded one participant questioning his motives, while another showed cautious openness.

What's more interesting for schedules are the organizational decisions that go with it. All client teams must release preliminary assessments of the Hegotá proposal by mid-September , divided into four levels from S to D. Barnabas Busa defines a level in the chat: S means strongly recommend inclusion, A means recommend inclusion once outstanding issues are resolved, B means expand goals, and D means decline. By DevCon, this should form a broad consensus on scope.

Forks start when a specific era is reached. The clock time in the report is only a conversion value for that era.

What September 28 means for you as an ETH holder

Short answer: Nothing. September 28 involves Sepolia, which is a testing network. The ETH used there has no market value, is distributed free through faucets, and cannot be exchanged for real ETH. Forks on Sepolia will not affect your balance, your address, or your tax status.

There are no snapshots, no redemption, and no new tokens for this upgrade. GlamAmsterdam is a protocol upgrade of an existing chain, not a chain split, nor a migration. If someone tells you the opposite, it's selling you something.

This is the most common scam target around fork dates. Every time a date is announced, messages urging action appear: so-called wallet transfers, forms to unlock something, deadlines. The rules are simple and there are no exceptions: Ethereum upgrades will never require you to enter mnemonics, transfer funds, or verify accounts.

Pledges, your own nodes and exchange balances: Who really needs to take action when forking

Hard forking only creates a need for action in one case, that is, those running their own software. This circle is controllable.

If you are running your own node or validator

, you must update the client software to a version that is aware of the new rules before activating. For test networks, this applies to the respective dates; for main networks, later. Those who miss out will not be able to follow the chain after the fork. Developer Network's experience on August 20 also showed that not to use the first version, but to use the version with fixed transition errors.

If you pledge

through the provider, then the provider will process the update, which is why Lido requested a test window during the meeting. You just need to know in advance who your provider is and how it communicates. A provider that remains silent about upgrades won't be more aggressive at the next interruption.

If your ETH is on an exchange or wallet

then you don't need to do anything. The exchange updates its own nodes, and wallets without its own nodes will automatically follow. Experience has shown that individual exchanges will suspend deposits and withdrawals for several hours around large main-network forks. This is a routine operation, will be announced in advance, and will not affect test network forking in any case.

How to check the Glamestead date by yourself in five minutes

You don't have to rely on any headlines. The core developer's documentation is public and the access path is the same as this article.

Meeting materials are stored in the project's planning warehouse, sorted by date. For the August 20, 2026 meeting, there are two documents: summary with decisions and goals and full chat history . In the first file, you can find rows about fork status, decisions, and goals; in the second file, you have timestamps, exact numbers, and team responses.

Make the conversion yourself. Multiply the era number by 32 to get the time slot. Multiply the time slot by 12 seconds and add the start time of the associated chain to get the Unix timestamp. If your results are consistent with the stated time, the numbers are internally consistent. If there is a deviation, it means that someone did not perform arithmetic operations when copying it.

To get a rating overview of individual proposals, the project also runs a public ranking tool, which was explicitly mentioned during the meeting. Starting in mid-September, you can see how the client team classifies Hegotá proposals.

Read Ethereum Upgrade Dates: Summary of Key Points

September 28, 2026 is the first reliable date for Glamestead. It is suitable for test networks and is yet to be confirmed on September 3, and the status of the main network has not yet been clarified. With this in mind, the next few weeks will be much calmer than the markets.

Mark September 3 on the calendar instead of September 28. That is the day when the next meeting will decide whether the date is established. Prior to this, any report using the word "decided" was inaccurate. If you are still comparing terms, our Exchange Compare is the fastest way.

Determine in advance who will update you. You are the one who owns your own node, and the provider who uses the pledge is the provider. Those who are still selecting providers should observe how they communicate around upgrades; differences are explained in our Pledge Comparison .

Treat every call to action as a fraud attempt. Agreement upgrades never require the relocation of your funds. Holding your own key is the best protection against this scam; relevant devices are listed in our Hardware Wallet Comparison .

(As of August 21, 2026. This article does not constitute investment advice. Price and fee structures will change; please check terms with your provider before purchasing.)