Glamsterdam is Ethereum's biggest upgrade since the Merge. What ePBS, block-level access lists, and bigger blocks change for fees, wallets, and L2s.
By Team MoonPay
Published on Sep 15, 2026
Ethereum ships one or two big upgrades a year, and most of them pass without a ripple in your wallet. Glamsterdam is different in scale, not in what it asks of you. It rewires how blocks get built and how the network charges for storage, and one of the engineers building it calls it the largest fork since the Merge. You still don't have to do anything.
This article covers what Glamsterdam changes, when it lands, what you'll notice, and what stays exactly as it is.
In short: Glamsterdam is Ethereum's next hard fork, expected on mainnet in Q4 2026. It moves block building inside the protocol (ePBS), lets the network process transactions in parallel (block-level access lists), and reprices storage so blocks can get much bigger. Plain ETH transfers cost less gas. Your address, your keys, and your wallet app don't change, and there is no new token.
Where it stands (September 2026): final devnet testing, with a short-lived public testnet called Platåberget live since August. The Sepolia testnet fork is next, proposed for late September or early October. Mainnet is expected in Q4 2026, with no date confirmed. We'll update this box as dates get locked in.
Glamsterdam is two upgrades shipping as one. "Amsterdam" is the execution-layer half, named after the city that hosted a previous Devconnect. "Gloas" is the consensus-layer half, named after a star. Ethereum has paired a city with a star for every fork since the Merge, which is how the network ended up with Shapella, Dencun, Pectra, Fusaka, and now this.
It follows Fusaka, which activated on December 3, 2025, and Pectra, which landed in May 2025. Fusaka was mostly about data capacity for rollups. Glamsterdam turns back to the base layer itself: how blocks are made, how transactions are processed, and how the network charges for the data it has to keep forever.
The official upgrade page frames it around three goals. Speed up processing by letting the network handle many transactions at once instead of one after another. Expand capacity by splitting up the work of creating and verifying blocks so there's more time to move large amounts of data around. And stop the database from bloating by pricing new storage to match what it actually costs the computers that run Ethereum.
The full list of proposals lives in the meta-EIP, EIP-7773. In June, Ethereum Foundation devops engineer Parithosh Jayanthi told CoinDesk it is "probably the largest fork we've had since the Merge."
Glamsterdam has slipped twice. It was pencilled in for the first half of 2026, then Q3, and is now expected in Q4. The scope grew, and enshrining block building turned out to be harder than the early estimates.
The testing sequence is the thing to watch. Developers ran a series of numbered devnets from April onward, then in August opened Platåberget, a short-lived public testnet built just for this fork. At the August 20 consensus call they proposed September 28 for the Sepolia testnet fork and October 26 for Hoodi, the second public testnet. Both dates are provisional. The same call reported client problems on the newest devnet, and operators like Lido and Optimism asked for a few extra days of stable testing before any fork, so a slip of a week or two would surprise nobody.
Mainnet gets a date only after both testnets fork cleanly and run for a while. Nobody credible has published one yet. Anything you see with a specific mainnet day attached is a projection, not a decision.
One more thing about the scope: it's frozen in name only. The meta-EIP is still a draft, and proposals can be pulled during testing. One, EIP-7610, was removed in August after client teams decided the extra checks cost more than they were worth. For the current state, ethereum.org's page and Forkcast are the two sources that track the calls directly.
Ethereum core developers pick at most two "headliner" proposals per fork, one for each layer. Glamsterdam's are ePBS on the consensus layer and block-level access lists on the execution layer.
Today, the validator whose turn it is to propose a block usually doesn't build it. Specialized builders assemble the most profitable block they can, and the proposer picks the best offer. That hand-off happens outside the protocol, through relays and a sidecar program called MEV-Boost, and it works because everyone involved trusts a small set of relay operators to be honest. It also forces a rush. The current design gives nodes about 2 seconds to broadcast and execute a block, which caps how much data a block can carry.
EIP-7732 writes the hand-off into Ethereum itself. The proposer selects a builder's committed bid, the builder delivers the payload, and payment settles in-protocol without a relay in the middle. A new group of validators, the payload timeliness committee, checks that the payload actually showed up on time, and attestations for the consensus block and the execution payload are split so neither has to wait for the other. According to ethereum.org, that stretches the propagation window from about 2 seconds to roughly 9.
For anyone who isn't running a validator, the gain is indirect but real: fewer trusted middlemen between validators and the blocks they propose, and a much wider window that makes bigger blocks and more blob space for layer 2s safe to add later. Builders and proposers can still use relays for features the protocol doesn't cover yet, so the relay business doesn't vanish overnight. It just stops being load-bearing.
Ethereum processes the transactions in a block one at a time, because until a transaction runs, nobody knows which accounts and storage slots it will touch. Run two of them at once and they might both try to change the same balance.
EIP-7928 fixes that with a list attached to every block that spells out, in advance, every account and storage slot the block's transactions will access, along with the values they'll end up at. The block header carries a fingerprint of the list so it can't be tampered with. Once nodes know which transactions don't overlap, they can read the data for many of them from disk at the same time and, eventually, execute them in parallel. A companion change, EIP-8159, upgrades the peer-to-peer protocol so nodes can share these lists with each other.
There's a second benefit that matters for decentralization. Because the list includes the final results, a node syncing to the network can copy those results instead of replaying every transaction from scratch. Faster syncing keeps running a node from home practical as blocks grow.
Ethereum's block gas limit sits around 60 million today. Glamsterdam is built to make a floor of 200 million safe, and developers are currently testing pricing at a 150 million reference limit. Bigger blocks mean more room per block. But letting blocks grow three times over without changing anything else would balloon the database every node has to store, so the fork reprices the operations that create permanent data.
EIP-8037 sets a fixed cost per byte of new state, tuned so the database grows at a predictable 120 GiB a year, and it splits the accounting so storage costs come out of their own reservoir instead of competing with computation for the same gas. EIP-8038 raises the price of reading existing state, because some lookups had become cheap enough to be a denial-of-service risk.
Then the cut in the other direction. EIP-2780 lowers the flat base cost every transaction pays, so a plain ETH transfer between two existing accounts becomes up to 71% cheaper, per ethereum.org. Sending to a brand-new address carries a surcharge, since the network has to create a permanent record for it.
Two caveats so nobody reads "71%" as a promise. First, that figure is the gas cost of the transfer, not the fee you'll see quoted. The fee is gas multiplied by the going gas price, and the gas price is set by demand. Second, bigger blocks lower the cost per transaction only if demand doesn't rise to fill the extra space. Ethereum.org's own wording is that Glamsterdam "will most likely reduce fees for everyday users." Contract deployments and anything that writes a lot of new state can get more expensive, which is the point.
Glamsterdam carries more than a dozen smaller proposals. These are the ones with a visible effect.
EIP-8061 lets exit capacity scale with the total amount of ETH staked instead of sharing a fixed cap with new activations, and gives validator consolidations their own lane. At current staking levels that's roughly four times the exit capacity and twice the consolidation capacity. The trade-off is that the window a node can stay offline before it needs a fresh trusted checkpoint drops from about 15.7 days to about 7.
EIP-7997 places a standard factory contract at the same address (0x12) on every participating EVM chain, so developers can deploy a smart wallet or app to an identical address on mainnet, Base, Arbitrum, and anywhere else that adopts it. Different addresses on different networks is one of the classic ways people lose funds. This is a quiet step toward chain abstraction.
Token transfers have always emitted a standard log; plain ETH moving between contracts hasn't, which is why exchanges and bridges run custom tracing to spot deposits. EIP-7708 makes every non-zero ETH transfer or burn emit a log. Expect deposit detection to get more reliable across the board.
EIP-7954 raises the maximum contract size from roughly 24 KiB to 64 KiB.
EIP-8045 stops already-slashed validators from being picked to propose, which removes guaranteed empty slots during a bad event.
Paginated receipt lists so nodes don't choke syncing large blocks (EIP-7975), a new opcode that reads the current slot number (EIP-7843), and a set of calldata and access-list floor prices that make block-stuffing more expensive (EIP-7976, EIP-7981).
Nothing changes on the day. Afterward, plain transfers between existing accounts cost less gas. There's nothing to do.
Nothing changes immediately. The wider propagation window ePBS creates is what allows Ethereum to add blob space in later forks, and blob space is the lever that lowers L2 costs. Nothing to do here either.
Solo stakers need Glamsterdam-ready versions of both their consensus and execution clients installed before the fork, and the testnets are the place to check that setup. Exits get faster once the fork is live, and the builder market changes underneath you. If you stake through a pool or a provider, the operator handles all of it; read their upgrade notice and that's the extent of your job.
The gas model changes. Tooling that treats gas limits as fixed values will break, per the Ethereum Foundation's protocol devops team. ETH transfers start emitting logs, contracts can be bigger, and the same-address factory arrives. Run your full test suite against Sepolia and Hoodi before mainnet.
Nothing changes and there's nothing to do.
Your address, your private keys, your balance, and your wallet app all stay the same. There is no token swap, no migration, and no airdrop. The ethereum.org FAQ says it plainly: anyone instructing you to "upgrade" your ETH is trying to scam you.
Named upgrades attract a predictable con. A message, a post, or a fake site tells you a deadline is coming and you need to migrate, claim, or verify your ETH by connecting a wallet or signing a transaction. The deadline is the tell. Real upgrades don't need anything from you, so anything that does is not the upgrade. The one honest task is for stakers and node operators, who update their software and nothing else.
Layer 2 fees don't move on the day either. Base, Arbitrum, and Optimism keep paying for blob space the way they do now. Glamsterdam's job is to make more of that space safe to add in later forks.
Three forks in about eighteen months, each with a different job.
On timing, Pectra activated on May 7, 2025, Fusaka followed on December 3, 2025, and Glamsterdam is expected in Q4 2026.
On what each one changed, Pectra gave ordinary wallets smart account features through EIP-7702, raised the maximum validator balance, and added blob capacity. Fusaka shipped PeerDAS and started ramping blob capacity for rollups. Glamsterdam brings ePBS, block-level access lists, storage repricing, and the path to a 200 million gas limit.
On what users noticed, Pectra showed up as passkeys, batched transactions, and sponsored gas inside wallets. Fusaka showed up as cheaper layer 2 fees over time. Glamsterdam should show up as cheaper plain transfers and faster staking exits.
On what you had to do, the answer is the same for all three: nothing.
Hegotá, the fork after Glamsterdam, is already being scoped. Its consensus-layer headliner is FOCIL, a mechanism that lets many validators force transactions into blocks so no single builder can censor them, and its execution-layer work continues the push toward nodes that don't have to store the full state.
Running alongside all of this is the Ethereum Foundation's post-quantum program. It runs pq.ethereum.org, a hub for the migration to quantum-resistant signatures, with weekly test networks and milestones mapped across four future hard forks. Glamsterdam doesn't touch that problem. It's the fork that makes the base layer big enough to keep building on.
When is Glamsterdam? Mainnet is expected in Q4 2026 with no confirmed date. The Sepolia testnet fork is the next milestone, proposed for late September or early October, with Hoodi to follow.
Do I need to do anything with my ETH? No. Balances, addresses, and keys are unaffected. Only node operators and solo stakers have a task, which is updating their client software.
Will gas fees go down? For plain ETH transfers between existing accounts, the gas cost drops by up to 71%. What you pay also depends on the gas price, which demand sets. Creating new accounts or contracts gets more expensive.
Does Glamsterdam affect Base and Arbitrum? Not on the day. L2s keep working as they do now. The longer propagation window ePBS creates is what lets Ethereum add blob capacity in later forks, which is how L2 costs come down.
Is there a Glamsterdam token or airdrop? No. Any site or message offering one is a scam.
Glamsterdam is a base-layer upgrade, and where you hold ETH decides how much of it you feel. Mainnet gets cheaper plain transfers. Layer 2s stay the cheap place for everything else.
MoonPay lets you buy ETH directly on Ethereum mainnet or on layer 2 networks like Base, Arbitrum, and Optimism, with a card, bank transfer, Apple Pay, or Google Pay. Pick the network where you'll actually use it and skip the bridging step entirely.
Glamsterdam changes how Ethereum builds blocks. It doesn't change what you have to do to use it.
