Ethereum's Glamsterdam upgrade — a combined execution-layer (Amsterdam) and consensus-layer (Gloas) fork — cleared its final devnet milestone on September 16 when Devnet-11 processed 84,000 validators across multiple client teams, raised the block gas limit from 60 million to 200 million, and mai...
"I can just spin up a thousand builders, rotate them, offer very high bids, and not produce payloads. Any teenager can do this." — Potuz, Ethereum Consensus Developer, All Core Developers Call, September 17, 2026
Ethereum's Glamsterdam upgrade — a combined execution-layer (Amsterdam) and consensus-layer (Gloas) fork — cleared its final devnet milestone on September 16 when Devnet-11 processed 84,000 validators across multiple client teams, raised the block gas limit from 60 million to 200 million, and maintained finality throughout. The Sepolia public testnet fork is scheduled for October 6. Mainnet activation is targeted for Q4 2026, though developers have not committed to a fixed date.
The upgrade's most consequential component is EIP-7732: enshrined proposer-builder separation (ePBS). It moves block construction — currently handled off-chain by MEV-Boost relays — into Ethereum's protocol layer. As of September 21, 2026, three builders (Titan at 56.1%, Quasar at 16.8%, BuilderNet at 15.3%) produce approximately 88% of all Ethereum blocks through MEV-Boost infrastructure that operates outside the protocol with no on-chain enforcement. ePBS replaces this arrangement with a staked, verifiable builder role inside the consensus mechanism.
The shift carries risk. A January 2026 academic paper warned that ePBS could "significantly amplify profit and content centralisation" among builders even as it reduces validator-side concentration. At a September 17 developer call, consensus developer Potuz warned that testnet environments lack the economic constraints of mainnet, leaving ePBS vulnerable to cheap builder-griefing attacks during the Sepolia trial. The mainnet version will require builders to stake real ETH, but the optimal deposit size and banning parameters remain under active debate.
Glamsterdam's Devnet-11 launched on September 16 with 84,000 validators running across Nethermind, Geth, and other execution-layer clients. The network completed its Gloas transition — switching to ePBS block-building rules — and then raised the gas limit from 60 million to 200 million without losing finality. Nethermind passed all 2,302 performance tests, processing 570.7 billion gas in 3 minutes and 15 seconds, according to The Defiant.
An earlier attempt, Devnet-9, launched September 1 with 1,000 validators and failed to finalize, prompting the shift to the larger Devnet-11 configuration.
The current timeline:
| Milestone | Date | Status | |-----------|------|--------| | Devnet-11 Gloas transition | September 16, 2026 | Completed | | Client software freeze | September 29, 2026 | Pending | | Sepolia testnet fork | October 6, 2026 | Scheduled | | Hoodi testnet fork | October 27, 2026 (tentative) | Planned | | Mainnet activation | Q4 2026 | No fixed date |
Developers shortened the review window between the client freeze and the Sepolia fork to 7 days, down from the standard 14. The compressed timeline reflects confidence from the Devnet-11 results but has drawn some criticism from client teams preferring more review time.
Glamsterdam is tracked under Meta EIP-7773 and includes ten Ethereum Improvement Proposals spanning block construction, gas economics, and execution capabilities:
Headline EIPs:
Gas and State Repricing:
Execution and Developer Tools:
Additionally, EIP-2780 is expected to reduce the cost of standard ETH transfers by up to 71%, and the combined repricing package is projected to cut overall L1 fees by approximately 78%, according to KuCoin and Cointribune reporting.
The current block production pipeline works as follows: validators run MEV-Boost, a sidecar application that connects to relays. Relays sit between builders (who assemble profitable blocks) and proposers (validators who include them in the chain). Validators trust relays to deliver payloads on time; builders trust relays not to front-run their transaction ordering. Neither trust relationship is enforced by the protocol.
According to relayscan.io data from September 21, 2026, the relay market breaks down as:
| Relay | 24h Market Share | Payloads | |-------|-----------------|----------| | Ultrasound | 32.70% | 4,386 | | Titan Relay | 28.04% | 3,761 | | bloXroute Regulated | 26.31% | 3,529 | | Aestus | 9.18% | 1,231 |
Approximately 88% of Ethereum blocks currently use MEV-Boost, according to Everstake. The system works, but it operates entirely outside Ethereum's consensus rules.
EIP-7732 changes the architecture: a builder becomes a staked consensus-layer actor. The builder submits a sealed block commitment (a bid) on-chain. The proposer selects the highest bid without seeing the transaction contents. The block is revealed only after the commitment is locked. A new validator duty — the Payload Timeliness Committee (PTC) — monitors whether builders actually deliver the payload they committed to.
The data propagation window extends from approximately 2 seconds under the current system to approximately 9 seconds under ePBS, giving builders more time to finalize blocks. Existing MEV-Boost and relay infrastructure continues to function — the transition is opt-in, not forced.
The builder market is concentrated. Relayscan.io data from September 21, 2026 shows:
| Builder | 24h Block Share | Blocks | |---------|----------------|--------| | Titan | 56.10% | 3,678 | | Quasar | 16.78% | 1,100 | | BuilderNet | 15.25% | 1,000 |
Three builders control 88% of block production. The Herfindahl-Hirschman Index stands at approximately 2,140, placing the market in "moderately concentrated" territory according to GitHub analysis by reldothescribe. Titan alone generated 48.21 ETH in profit over 24 hours, while smaller builder bombora.build reported negative profit of -2.13 ETH — illustrating the scale advantages that entrench dominant builders.
A January 2026 academic paper (Yang, Öz, Wu, Zhang — published in Proceedings of the ACM on Measurement and Analysis of Computing Systems, Volume 10, Issue 2) found that ePBS could reduce validator-side concentration but "significantly amplify profit and content centralisation" among builders. The researchers' simulation models showed that geographic distribution of builders matters more under ePBS than under the current relay system.
The Bitfinex blog cited a study finding that over 50% of high-value Ethereum transactions were routed through private channels as of mid-2025, a phenomenon that ePBS makes more transparent but does not eliminate.
The gas limit expansion works in phases. Core contributors have aligned on a 200M floor — a 3.3x increase from the current 60M — but this depends on BALs (EIP-7928) enabling parallel execution. Without parallel processing, a 200M gas limit would overwhelm sequential client execution.
EIP-8037 was critical to unlocking the higher limit. By introducing a fixed cost per state byte and a separate gas reservoir for state growth, it gives client teams a sustainability ceiling. Without state-growth controls, tripling capacity would accelerate blockchain bloat — a problem that has historically constrained gas limit increases.
The projected impact on fees:
These projections assume demand remains constant. If usage rises proportionally to match new capacity, base fees could stabilize near current levels. Ethereum's fee-burn mechanism (EIP-1559) means higher throughput at lower per-transaction fees does not automatically reduce total ETH burned — it depends on aggregate gas consumption.
According to Figment's institutional staking analysis, MEV-Boost currently adds 10–30% to staking rewards, depending on market conditions. ePBS changes the validator's relationship to this revenue stream in several ways.
What changes:
What stays the same:
What is uncertain:
Node operators and stakers must update both consensus-layer and execution-layer clients before activation. The PTC duty is new client software logic, not a separate application.
Builder griefing on testnet. Consensus developer Potuz's September 17 warning is specific to Sepolia, where free test ETH eliminates the economic cost of creating fake builder identities, bidding high, and withholding payloads. On mainnet, builders must stake real ETH, creating economic deterrence. The open question is whether the staking requirement is high enough to prevent well-funded attackers from disrupting block production.
Builder centralization under ePBS. Academic research suggests ePBS may worsen builder concentration. If the top three builders maintain 88% market share after the transition, the upgrade will have moved centralization risk from one layer (relays) to another (builders) without reducing it.
Compressed testing timeline. The 7-day review window between client freeze (September 29) and Sepolia fork (October 6) is half the standard period. Client teams have expressed concern that this leaves insufficient time for edge-case testing.
L2 interaction. The longer propagation window (9 seconds vs. 2 seconds) could enable higher blob capacity in future forks, benefiting Layer 2 rollups. But Glamsterdam itself does not change blob pricing or availability — that is deferred to the subsequent Hegota upgrade, planned for H1 2027.
Repricing breaks assumptions. Three separate gas-repricing EIPs change the economics of existing smart contracts. While contracts remain compatible, dApp developers face recalibrated cost structures. According to Everstake, this requires testing but no code changes for most deployed contracts.
Glamsterdam is Ethereum's largest structural upgrade since The Merge. It addresses a genuine architectural gap: the network's block production depends on off-chain infrastructure that 88% of validators use but that no protocol rule governs. Moving builder commitments on-chain is a direct response to this dependency.
The capacity expansion — a 3.3x gas limit increase — is enabled by parallel execution tooling that did not exist when the gas limit last reached community consensus. The fee repricing package redistributes costs to align with actual resource consumption rather than legacy pricing from 2015-era opcode designs.
The risk profile is concentrated in builder economics. ePBS does not guarantee a more competitive builder market. If scale advantages persist, the same entities that dominate block production today will dominate it on-chain tomorrow — with the difference being that their dominance becomes protocol-visible rather than relay-visible. Whether transparency alone produces competitive pressure is an empirical question that Glamsterdam's mainnet deployment will answer.