Ethereum's Glamsterdam Upgrade Hits Public Testnet With the First Gas Repricing Since 2021

Ethereum's next upgrade has moved from developer devnets to a public testnet anyone can run against, and it carries a change the network has not made in five years.

Plataberget

The Ethereum Foundation launched a dedicated testnet, Plataberget, in August, and it forked to Glamsterdam on 20 August 2026. Internally it is identified as glam-devnet-8. Unlike a devnet spun up for a single week of testing, it is intended to run for "a few months", giving client teams and application developers a stable target.

The long-lived public testnets, Sepolia and Hoodi, will run through the Glamsterdam fork once devnet behaviour is stable. No dates have been announced for either, and none for mainnet. The Foundation's published target for Glamsterdam remains a timeframe rather than a date: Q4 2026.

What is in the upgrade

Glamsterdam is a paired fork — Gloas on the consensus layer, Amsterdam on the execution layer. Two changes carry the headline.

EIP-7732, enshrined proposer-builder separation, moves the split between proposing a block and building it into the protocol itself. Its practical effect is on timing: the propagation window widens from roughly two seconds to around nine, which is what makes the rest of the throughput work survivable on ordinary hardware.

EIP-7928, block-level access lists, publishes up front which state a block will touch, so execution can be parallelised rather than replayed serially.

Alongside them sits a longer list scheduled or in devnet: EIP-7610, 7688, 7778, 7843, 7976, 7981, 8024, 8246 and 8282.

The repricing

The quieter change may matter more day to day. EIP-8037 adjusts the cost of creating state — new accounts, new storage slots, deployed bytecode. EIP-8038 adjusts the cost of accessing it: SSTORE, SLOAD, cold account access, EXTCODESIZE and EXTCODECOPY.

This is the first repricing of state operations since Berlin, in April 2021.

Gas costs are fixed numbers written into the protocol, and the physical costs they were meant to approximate — disk, memory, bandwidth — move underneath them every year. When the two drift apart, the gap becomes an attack surface: an operation priced below what it actually costs a node to perform is an invitation to fill blocks with it. Repricing is maintenance, and skipped maintenance is why it is being done at scale now rather than incrementally.

The Foundation describes the new schedule as sized to support roughly a three-times increase in base throughput. That figure describes what the pricing permits, not what mainnet will be configured to run.

For developers the practical read is that relative costs shift. A contract that writes a great deal of new state gets relatively more expensive; a read-heavy one gets relatively cheaper. Anyone with hard-coded gas limits in deployed code should be testing against Plataberget rather than waiting for the mainnet announcement.

Where the network stands

Queried directly against a public beacon node on 27 August, the consensus layer showed 902,681 active validators holding 42,546,691 ETH — a little over a third of supply staked. The deposit queue held 31,120 pending entries.

Ether traded at 2,502.17 dollars on 27 August, up from 1,870.22 on 2 August. Layer-2 total value secured stood at 37.93 billion dollars, led by Base at 12.61 billion and Arbitrum One at 11.66 billion — though measured in ETH rather than dollars, layer-2 value has fallen over the same period, because ether rose faster than the deposits did.

Sources

• Plataberget testnet announcement, 17 August 2026 — blog.ethereum.org
• Glamsterdam repricing testing, 24 August 2026 — blog.ethereum.org
• Glamsterdam roadmap page — ethereum.org
• Validator and staking figures — public beacon node API, /eth/v1/beacon/states/head/validators, retrieved 27 August 2026.
• Layer-2 value secured — L2BEAT, retrieved 27 August 2026.
• Prices — Kraken public ticker API, 27 August 2026.