Ethereum’s next big upgrade, Glamsterdam, is now live on the Sepolia testnet and has already cleared its first formal checkpoint. After the Oct. 6 fork, Sepolia pushed its gas limit from 60 million to the upgrade’s 200 million goal in roughly 11 hours. Through all of that, the chain kept finalizing blocks as usual and logged no major faults.
Glamsterdam bundles the consensus-layer change called Gloas with the execution-layer change Amsterdam. Right now, this is testnet-only. But the direction is obvious: a big jump in capacity for Ethereum mainnet.
Ethlabs researcher Barnabé Monnot summed up the fork bluntly after rollout: "This is a big deal. Finishing this upgrade means Layer 1 scale increases by at least more than 3x, the contract size limit rises by 2.5x, plus a large batch of useful new features." He said the package would expand Layer 1 scale by more than 3x, lift the contract size cap by 2.5x, and bring in a wide set of practical features.
Ethereum Foundation researcher Toni Wahrstätter made much the same argument. He called Glamsterdam Ethereum’s biggest throughput increase since launch. By design, the upgrade aims for a gas limit of about 200 million, versus the current 60 million, or about 3.3x in one move.
The article says Ethereum mainnet sat at a 30 million gas limit for years before validators lifted it to 60 million in three steps in 2025. Even the 2021 London upgrade, despite doubling maximum block capacity, left the target value at 15 million.
What changes in Glamsterdam
The fork contains 18 Ethereum Improvement Proposals, or EIPs. Still, three of them do most of the heavy lifting on scaling.
ePBS gives validators more time
The first major piece is enshrined proposer-builder separation, or ePBS, under EIP-7732.
Right now, most Ethereum blocks are put together by specialized builders and handed to validators through trusted middlemen called relays. Under EIP-7732, proposers first commit to a signed bid from a builder. The builder then reveals the full block after that.
That moves block inspection work out of the roughly four-second attestation window. Under the proposal, most validators would get about nine seconds instead.
Block-level access lists open the door to parallel processing
The second pillar is block-level access lists, or BALs, under EIP-7928.
Today, Ethereum clients usually process block transactions one at a time because they do not know ahead of time which parts of state each transaction will touch. EIP-7928 changes that. It attaches a map to each block listing all touched state and the resulting post-execution values.
Once clients have that, they can prefetch data and validate transactions in parallel. The article uses a factory comparison: ePBS gives quality control more time, and BALs make sure the workshop has everything ready so several production lines can run at once.
Larger contracts and higher pricing for new state
Another big group of changes deals with contract size and gas repricing. EIP-7954 raises the contract size cap from 24 KiB to 64 KiB, loosening a long-running limit that forced larger applications to split their code apart.
At the same time, the fork makes creating new state more expensive. Under EIP-8037, creating a new storage slot rises from 20,000 gas to 97,920 gas. Under EIP-2780, a standard ETH transfer to an existing account stays at 21,000 gas.
So yes, onchain activity that creates huge numbers of new accounts or storage entries, including airdrops, mints, and new wallets, will get pricier. Ordinary ETH transfers will not change in the same way.
What happened on Sepolia
The first Glamsterdam block on Sepolia was produced on Oct. 6 with a gas limit of 60 million, matching Ethereum mainnet today. About 11 hours later, Sepolia hit the 200 million target and has remained there.
Each block proposer can move the gas limit by only about 0.1% at a time. If every proposer had raised it in every block, Sepolia could have made the climb in around four hours. In practice, it took more than twice as long. One reason the article gives: two major consensus clients, Prysm and Teku, kept validators at 60 million after the fork unless operators explicitly chose to switch to 200 million.
Wahrstätter later posted early figures suggesting the changes were already showing results. Compared with pre-fork performance, clients processed each unit of gas faster. For a 40 million-gas block, Geth had previously needed about 115 milliseconds to execute it. After the fork, that number dropped to about 30 milliseconds.
But the article also warns that these figures cover less than a day of observation and just one node for each client. Even so, it still called the early read encouraging.
The path to mainnet
After Sepolia, Glamsterdam is scheduled for deployment on Ethereum’s Hoodi testnet. The tentative activation date being discussed is Oct. 27, though that date is not final.
Mainnet is still being aimed at some point in the fourth quarter of 2026. But the schedule has slipped before. Glamsterdam was first planned for the first half of 2026, and the Sepolia fork itself was delayed from Aug. 3 to Oct. 6. More timetable changes could still happen.
What else the upgrade changes
The larger block design also brings operational side effects. The Ethereum Foundation has warned that wallets, indexers, and gas estimation tools that assume a fixed maximum gas limit need to update or risk breaking.
Because state growth is being repriced, activity that creates many new accounts or storage entries will face a higher cost, while routine transfers will not. The article presents those effects as new limits that arrive with the extra room.
At the widest level, the main shift is simple: Ethereum Layer 1 is on course to get a lot roomier. The article points to Fake World Assets, which briefly became mainnet’s biggest gas consumer on its busiest day in August. With Glamsterdam in place, spikes like that would have much more capacity to spill into, instead of pushing fees higher and squeezing everyone else out.
For now, Glamsterdam has cleared its first testnet hurdle. Next up: whether Hoodi activates on the current tentative timeline, and whether the mainnet target of Q4 2026 still stands.