Hey there!

On the 6th of October at 13:53 UTC, Ethereum's next big upgrade, Glamsterdam, went live on Sepolia, one of the network's public test networks. Nothing changed on the real Ethereum network that day, but this is the first time the full upgrade has run somewhere outside the developers' own private test setups, and it gives a good preview of what is coming to mainnet.

The upgrade has an awkward name and a long list of technical proposals behind it. In this post I'll go through what it actually changes, in plain English, and who needs to do anything about it.

What is the Glamsterdam upgrade?

Ethereum is run by two pieces of software working together. The execution layer runs transactions and smart contracts (the programs that live on the blockchain). The consensus layer is how thousands of computers, called validators, agree on which blocks of transactions are valid. Each layer gets its own upgrade name: this time the execution side is called Amsterdam and the consensus side is called Gloas, so the combined upgrade is Glamsterdam.

According to the official testnet announcement, the upgrade bundles 25 EIPs. An EIP (Ethereum Improvement Proposal) is a written spec for one change to the network. It follows Fusaka, the previous upgrade, which went live on mainnet in December 2025.

The changes that matter most for understanding the upgrade fall into three groups: who builds blocks, how fast nodes can process them, and what things cost.

Change 1: Block building moves into the protocol (ePBS)

Every 12 seconds, someone has to put together the next block of transactions. Today that job is split informally. A validator is chosen to propose the block, but in practice most of them hand the actual building to specialised companies through outside software such as MEV-Boost, with third-party "relays" passing blocks between the two. The whole arrangement depends on trusting those relays to behave.

Enshrined proposer-builder separation, or ePBS (EIP-7732), makes that split an official part of Ethereum. The proposer picks the block, a builder assembles its contents, and the protocol itself handles the hand-off and the payment between them. The ethereum.org Glamsterdam page says this removes the need for middleware like MEV-Boost and stretches the window for sending block data around the network from about 2 seconds to about 9 seconds.

That extra time matters because a bigger block takes longer to send and check. Giving validators more breathing room is part of what makes larger blocks safe.

Change 2: Blocks come with a map (block-level access lists)

Right now, an Ethereum node processes the transactions in a block one at a time, in order. It has no way of knowing in advance which accounts or pieces of storage each transaction will touch, so it can't safely run two of them at once.

Block-level access lists (EIP-7928) fix that by attaching a list to every block of exactly which accounts and storage slots it reads and writes. ethereum.org describes it as a map of which parts of the database will be accessed before the work begins. If transaction A only touches Alice's wallet and transaction B only touches Bob's, a node can now see that up front and process both at the same time.

The same list helps new nodes catch up faster. Instead of replaying every past transaction, a syncing node can copy the final results it needs, which ethereum.org calls "executionless sync".

Change 3: Gas prices get rebalanced

Every action on Ethereum costs gas, a unit that measures how much work it takes. Your actual fee is the gas used multiplied by the current price per unit of gas. Glamsterdam changes how much gas some actions use, and the changes go in both directions.

Plain ETH transfers get cheaper. EIP-2780 cuts the base cost of a transaction, and ethereum.org says simple ETH transfers could become up to 71% cheaper in gas terms.

Creating and reading data on the chain gets more expensive. EIP-8037 raises the cost of creating new state (new accounts, new storage, deploying contract code), and EIP-8038 raises the cost of reading and writing existing storage. The Ethereum Foundation said in August that these are the first adjustments to state costs since the Berlin upgrade in 2021, and that the new schedule supports roughly a 3x increase in how much the network can process.

The logic is that storage is the part of Ethereum that keeps growing forever, because every node has to keep a copy of it. If the network is going to handle more transactions, the things that make that storage grow have to be priced closer to their real cost.

Together, these changes are meant to let blocks get much bigger. Blocks currently hold about 60 million gas, and developers have talked about 200 million becoming possible after Glamsterdam. That won't happen in one jump; validators raise the limit gradually.

A testnet that was already doing its job

Sepolia was originally pencilled in for August and slipped as developers worked through bugs on their private test networks. There was also a near miss right before the fork: according to reports, one consensus client, Prysm, shipped a version that defaulted to the old 60 million gas limit instead of the planned higher one, and a fixed release (7.2.1) came out about 16 and a half hours before the fork. Catching that sort of mismatch on a testnet, rather than on mainnet, is exactly why the testnets exist.

The order from here is Sepolia, then a second testnet called Hoodi, then mainnet. The announcement says the Hoodi and mainnet dates will be set by the client teams later. ethereum.org currently lists mainnet as expected in Q4 2026, but that date is not confirmed.

Do you need to do anything?

That depends on who you are.

• If you just hold ETH: no. ethereum.org is clear that you don't need to convert or upgrade your ETH and your balance stays the same. Anyone telling you to "upgrade" your ETH before the fork is trying to scam you.
• If you run a node: update both your execution and consensus clients before each network forks. The announcement lists the minimum versions for Sepolia.
• If you build on Ethereum: this is the group with real work to do. The repricing can break contracts and apps that assume a fixed amount of gas. The classic example is Solidity's transfer() and send(), which forward a fixed 2,300 gas. Other risks are hardcoded gas values in calls, code that changes its behaviour based on how much gas is left, and pre-signed transactions with a fixed gas limit. Wallets and frontends that cache gas estimates may also underestimate costs.

The Ethereum Foundation has published a repricing impact guide, linked from the announcement, with a search tool for checking whether a deployed contract is affected. In most cases the fix is simply allowing a higher gas limit. Testing your app on Sepolia now is the cheapest way to find out.

The takeaway

Glamsterdam doesn't add anything flashy for everyday users. What it does is make Ethereum's main network able to process a lot more: block building becomes part of the protocol instead of relying on outside relays, blocks come with a map that lets nodes process transactions in parallel, and gas costs are rebalanced so ETH transfers get cheaper while storage-heavy actions cost more.

If you hold ETH, there is nothing to do except ignore anyone asking you to "upgrade" it. If you build on Ethereum, the next few weeks on Sepolia are the time to check your gas assumptions, before the same changes reach mainnet.