Ethereum is preparing the Glamsterdam upgrade, its biggest hard fork since the Merge. Gas costs should fall significantly, with mainnet activation targeted for the third quarter of 2026.
Glamsterdam is a coordinated hard fork that updates Ethereum's consensus layer and execution layer at the same time. The name combines the two central components: "Gloas" stands for the consensus layer, while "Amsterdam" stands for the execution layer. The upgrade follows Fusaka from December 2025 and Pectra from May 2025, both of which laid the technical groundwork. Before that, the Merge in September 2022 was the last comparably profound layer-1 upgrade, when Ethereum moved to proof of stake. Developers therefore rank Glamsterdam as likely the biggest fork since the Merge. Originally planned for an earlier phase, the "Soldøgn" devnet completed successfully in May 2026, which means the teams have tested all planned EIPs. Before mainnet activation, however, the public testnets Sepolia and Hoodi still follow.
Enshrined proposer-builder separation: Why ePBS reorders block production
Today, more than 88% of all Ethereum blocks are created outside the protocol. The reason is MEV-Boost, an off-chain software that connects block proposers with specialized block builders through external relays. These builders optimize the order of transactions to extract additional returns from so-called maximal extractable value (MEV). However, the heavy dependence on a few external relays is considered a centralization risk for a network that is supposed to operate in a decentralized way. As a result, exactly this concentration of block production sits at the center of criticism of today's architecture.
EIP-7732 addresses this point and integrates the separation of block proposer and block builder directly into the Ethereum protocol. Protocol-backed bidder commitments therefore replace the off-chain relays and work without trusted intermediaries. At the same time, the upgrade extends the data retrieval window for validators from two to roughly nine seconds. This longer window gives validators more time to fetch and process the ePBS payload. Consequently, it lowers the technical barrier to validating a block yourself instead of relying on external providers.
Supporting EIPs close the remaining gaps. EIP-7805 introduces fork-choice-enforced inclusion lists, which shield validators against censorship of individual transactions. EIP-8045 furthermore excludes slashed validators from proposing new blocks. The developers are currently testing the components as a bundle.
Subscribe to our newsletter
The best articles of the week, directly delivered into your mailbox.
Ethereum Glamsterdam upgrade and parallel processing
The second core building block is EIP-7928, known as block-level access lists. Each block declares in advance which accounts and smart contract data it will access. Nodes can therefore preload the required data in parallel and validate transactions in parallel, instead of processing them sequentially. This parallelization additionally allows a significant increase in the gas limit, from the current 60 million to a targeted 200 million gas per block, which corresponds to a factor of roughly 3.3. The gas limit determines how much computational work a single block can absorb, and thus counts as a central lever for throughput.
The performance targets are accordingly ambitious. According to model calculations, throughput should rise to roughly 10,000 transactions per second. Simple ETH transfers should also become cheaper, by up to 71% based on the same calculations, and complex smart contract calls likewise lose noticeable cost. However, these figures rest on theoretical models under optimal conditions and are not guaranteed results. Ultimately, the actual relief in practice depends on network utilization and the type of transactions.
Elsewhere, a deliberate trade-off emerges. EIP-7904 adjusts gas costs based on client benchmarks: computational operations become cheaper, while storing state data becomes more expensive. EIP-8011 furthermore splits gas metering into the dimensions of computation, storage, and bandwidth, while EIPs 8032, 8037, and 8038 further raise the costs for state creation and state access. This shifts the incentives to store data directly on the blockchain. The repricing logic should therefore spread the load on nodes more evenly across the different resources.
Testnet roadmap: From Soldøgn to the mainnet launch in the third quarter of 2026
The "Soldøgn" devnet marks the first completed phase. The developers finished it successfully in early May 2026 and likewise validated all planned Glamsterdam EIPs in a private test network. The technical foundation is therefore in place, even as the hardening of the code continues. Such devnets serve to test how the changes interact before real value is at stake.
As the next step, the public testnets Sepolia and Hoodi follow before the upgrade reaches mainnet. However, the Ethereum Foundation has not communicated a fixed date, which is why all timing figures remain developer estimates. Originally, sources named June 2026, yet after the completion of "Soldøgn" and the still-pending public testnets, the third quarter of 2026 appears more realistic. The specific target is the end of August 2026. Later, the next upgrade should follow with Hegotá, which the developers are planning for the second half of 2026.
What ETH holders, validators, and developers need to know now
For ETH holders, nothing changes. Balances and wallets remain unchanged, and no measures whatsoever are required. A hard fork updates the rules of the protocol, not the holdings of users. Therefore, anyone who merely holds ETH needs neither to install software nor to carry out transactions.
The situation looks different for validators and node operators. Before mainnet activation, they must update both their consensus-layer and their execution-layer clients. Additionally, they should check their IOPS capacity, since parallel processing generates significantly more simultaneous disk accesses. Anyone who ignores these requirements risks losing connection to the network after the fork. The higher hardware requirements in particular are likely to become relevant for smaller operators.
Smart contract developers, finally, face shifted design incentives. Because the costs for state creation rise, storing data directly on the blockchain becomes more expensive. The strategic direction is notable: unlike Dencun in March 2024, which primarily made layer-2 cheaper through blob transactions, Glamsterdam instead addresses layer-1 scaling directly. As a result, Ethereum's base layer moves back to the center of development after years.
