Ethereum's Glamsterdam hard fork — its largest protocol change since the Merge — has entered final devnet testing with a Q4 2026 mainnet target. The upgrade bundles three headline EIPs: EIP-7732 (enshrined proposer-builder separation), EIP-7928 (block-level access lists enabling parallel executio...
"ePBS only prevents builder centralization from spilling over into the staking layer, but the issue of builder centralization itself remains." — Vitalik Buterin, Ethereum Co-Founder
Ethereum's Glamsterdam hard fork — its largest protocol change since the Merge — has entered final devnet testing with a Q4 2026 mainnet target. The upgrade bundles three headline EIPs: EIP-7732 (enshrined proposer-builder separation), EIP-7928 (block-level access lists enabling parallel execution), and EIP-7904 (gas repricing). Combined, these proposals target a 5× increase in gas capacity from 36 million to 200 million, a theoretical throughput ceiling of 10,000 transactions per second, up to 70% reduction in MEV extraction, and an estimated 78.6% cut in per-transaction L1 gas costs.
Devnet-7 held stable under elevated transaction loads through late July, according to All Core Devs Testing notes dated July 27. Devnet-8 — expected to be the final devnet before public testnets — was queued for early August with mandatory EIP-8070 (sparse blob pool) changes. Sepolia and Hoodi testnet activations will follow before any mainnet date is confirmed. The internal working target remains end of August, but core developers have consistently signaled Q4 2026 as the realistic window.
The upgrade arrives as Ethereum's staking ratio hits a record 33.98% (41.41 million ETH), L1 gas prices hover near 0.15 gwei, and daily active addresses fluctuate between 400,000 and 990,000. Glamsterdam does not solve Ethereum's revenue problem — the network generates roughly $65 million in annual fee revenue against $4–5 billion in staking issuance subsidies — but it restructures the economic plumbing through which that value flows.
Over 90% of Ethereum blocks are currently built via MEV-Boost, a third-party relay system operated outside the protocol. Validators outsource block construction to specialized builders who optimize transaction ordering for maximum extractable value. The relay layer — dominated by Flashbots and a small number of operators — represents a single point of centralization risk.
EIP-7732 moves the proposer-builder separation (PBS) mechanism directly into the Ethereum consensus layer. Under ePBS, block proposers commit to a block header from a builder, and the builder then reveals the full block body. The protocol enforces this handoff natively, eliminating dependence on external relays.
What changes operationally:
MEV reduction estimates. Ethereum Foundation researchers project that ePBS could reduce MEV extraction by up to 70%. Current annual MEV extraction on Ethereum exceeds $550 million, according to 2026 estimates from multiple MEV tracking services. The reduction comes not from eliminating MEV entirely but from constraining the builder's ability to exploit the relay layer's opacity. Sandwich attack value already fell from approximately $10 million per month in late 2024 to $2.5 million by late 2025 as more transactions moved through private, MEV-aware routes. ePBS accelerates this trend at the protocol level.
Limitations acknowledged by Buterin. The Ethereum co-founder has stated publicly that ePBS addresses builder centralization's spillover into the staking layer but does not solve builder centralization itself. Two to three builders currently construct the majority of Ethereum blocks. ePBS changes the mechanism by which they interact with proposers but does not alter the competitive dynamics of the builder market.
EIP-7928 introduces Block-Level Access Lists (BALs) — a per-block data structure that records every account and storage slot accessed during execution, along with post-execution state values. The proposal, authored by Toni Wahrstätter, Dankrad Feist, Francesco D'Amato, Jochem Brouwer, and Ignacio Hagopian, has received strong support from core developers.
How BALs enable parallelism. Today, Ethereum executes transactions sequentially within each block. A node must process transaction 1 before transaction 2 because it cannot know in advance which state each transaction will touch. BALs solve this by pre-declaring the state access pattern for the entire block:
Capacity target. BALs clear the path for increasing the gas limit from the current 36 million to a target of 200 million — a 5.5× increase. At full utilization, this translates to a theoretical throughput ceiling of approximately 10,000 TPS, up from the current practical limit of roughly 15–30 TPS on L1.
Overhead. According to the EIP specification, BAL data adds approximately 35 KiB per block at 36 million gas, with worst-case sizes remaining below calldata worst-case sizes. The overhead is designed to be marginal relative to the execution gains.
Developer reception. Wahrstätter noted on X (formerly Twitter) that "the Glamsterdam headliner-proposal Block-level Access Lists (EIP-7928) received much love from core devs recently." The proposal passed through the Fellowship of Ethereum Magicians review process and was included in the Soldøgn interop devnet that completed May 2, 2026.
EIP-7904 recalibrates gas costs to reflect actual computational resource consumption on modern hardware. Many current gas prices were established years ago and no longer correspond to real execution costs.
Projected impact. The repricing package is estimated to reduce gas costs by approximately 78.6% for both simple ETH transfers and complex smart contract interactions. A simple ETH transfer currently costs approximately $0.10–$0.20 on L1 at 0.15 gwei gas prices. Post-Glamsterdam, base costs per operation drop, though actual user-facing fees depend on demand. Increased capacity does not guarantee lower fees if usage scales proportionally.
Interaction with gas limit increase. The repricing works in tandem with the 200 million gas limit target from EIP-7928. Lower per-operation costs combined with higher block capacity means more transactions can fit per block at lower individual cost — assuming demand does not immediately saturate the new capacity.
Soldøgn interop (May 2, 2026). The multi-client interop devnet validated all planned Glamsterdam EIPs. Developers stress-tested builder behavior under elevated transaction loads, iterated on gas semantics affecting wallets, rollups, and MEV infrastructure.
Devnet-7 (late July 2026). The All Core Devs Testing notes from July 27 confirmed that Glamsterdam-Devnet-7 "held up well over the weekend with increased transaction loads." This cleared the path for Devnet-8.
Devnet-8 (early August 2026). Expected to be the final devnet iteration before public testnets. Mandatory changes include EIP-8070 (sparse blob pool) Engine API modifications. According to the ACDC #184 meeting notes from August 6, progress is tracked against a short list of non-negotiable implementation items.
Public testnets. Sepolia and Hoodi (or equivalent) activations will follow Devnet-8 stabilization. These are the clearest leading indicators for mainnet timing.
Mainnet. The internal target of late August 2026 is described by developers as aspirational. Q4 2026 — broadly September through December — represents the realistic activation window, contingent on cross-client testing completion and implementation parity across all execution and consensus layer clients. Stakers and node operators must update both consensus-layer and execution-layer client software before activation. ETH holders require no action.
Parallel roadmap note. Ethereum developers have also renamed the subsequent upgrade from "Heze-Bogotá" to "Hegotá," with scope adjustments ongoing. Glamsterdam and Hegotá together constitute Ethereum's 2026 upgrade calendar.
Annual MEV extraction on Ethereum currently exceeds $550 million. ePBS restructures the extraction pipeline:
| Stakeholder | Pre-Glamsterdam | Post-Glamsterdam (Projected) | |---|---|---| | MEV searchers | 60–70% of extracted value | Reduced via on-chain builder accountability | | Block builders | 15–25% | Constrained by protocol-level rules | | Validators (proposers) | 10–15% via MEV-Boost | Direct protocol-level MEV share | | Relay operators | 1–5% infrastructure fees | Role diminished; relays become optional |
The shift does not eliminate MEV. It moves the extraction mechanism from an opaque, off-protocol system to an observable, on-chain one. Solo stakers currently earn 3.3–4% all-in APR including MEV. Post-ePBS MEV distribution to validators will be determined by in-protocol auction dynamics rather than relay-mediated negotiations.
With 41.41 million ETH staked across approximately 897,000 active validators, the base staking APR sits at 2.78%. EIP-8361, proposed separately from Glamsterdam, suggests tapering issuance as the staking ratio rises — which would halve yield to approximately 1.09% if implemented. Glamsterdam itself does not alter issuance mechanics, but ePBS changes how execution-layer rewards (priority fees + MEV) reach validators.
Glamsterdam's ePBS introduces a longer payload-propagation window, which creates room for higher blob capacity in future Blob Parameter Only (BPO) forks. This directly affects L2 rollup data-posting costs:
Three networks — Base, Arbitrum, and Optimism — now process nearly 90% of all L2 transactions. Cheaper L1 settlement disproportionately benefits these incumbents over smaller rollups with thinner margins.
| Metric | Value | Source | |---|---|---| | Staking ratio | 33.98% (41.41M ETH) | As of August 4, 2026 | | Active validators | ~897,000 native staking | KuCoin Research, August 2026 | | Base staking APR | 2.78% | Network data | | Gas price | ~0.15 gwei | June–August 2026 average | | Simple transfer cost | $0.10–$0.20 (L1) | Network data | | Daily active addresses | 400,000–990,000 range | Etherscan, growthepie | | Annual fee revenue | ~$65 million | On-chain data | | Annual staking issuance | ~$4–5 billion | Ethereum inflation model | | MEV-Boost adoption | >90% of blocks | Flashbots data | | Annual MEV extraction | >$550 million | MEV tracking services |
Glamsterdam bundles three structural changes — ePBS, parallel execution via BALs, and gas repricing — into Ethereum's largest hard fork since the Merge. Devnet-7 held stable under load; Devnet-8 is in progress with a Q4 2026 mainnet target.
ePBS moves block building on-chain, reducing Ethereum's dependence on external MEV-Boost relays. Researchers estimate up to 70% reduction in MEV extraction. The change constrains how value leaks from users to MEV infrastructure but does not eliminate builder concentration.
Block-Level Access Lists (EIP-7928) enable parallel transaction execution by pre-declaring state access patterns per block. This clears the path for a 200 million gas limit (up from 36 million) and a theoretical 10,000 TPS ceiling.
Gas repricing (EIP-7904) cuts per-operation costs by an estimated 78.6%. Actual user-facing fees depend on demand. More capacity does not guarantee lower fees if activity scales to fill it.
L2 rollups benefit indirectly through cheaper L1 data-posting and future blob capacity expansion. Base, Arbitrum, and Optimism — which process 90% of L2 transactions — capture the largest margin improvement.
The revenue gap persists. Ethereum generates $65 million in annual fee revenue against $4–5 billion in staking subsidies. Glamsterdam restructures value flows but does not close this gap. The network remains subsidy-dependent at current activity levels.
Glamsterdam represents Ethereum's most ambitious execution-layer overhaul since the consensus switch to proof-of-stake. The three headline EIPs address longstanding structural issues: builder centralization via ePBS, sequential execution bottlenecks via BALs, and misaligned gas pricing via repricing. The technical scope is significant.
The economic implications are more nuanced. A 200 million gas limit and 78.6% cost reduction sound transformative. But Ethereum L1 currently operates at approximately 0.15 gwei — near-zero gas prices reflecting that most activity has migrated to Layer 2 networks. The upgrade expands capacity that is, at present, underutilized on L1. Whether Glamsterdam attracts activity back to the base layer or simply reduces costs for the rollups that post data to it depends on application demand that the protocol cannot manufacture.
MEV restructuring through ePBS is the most immediately consequential change. Moving block construction on-chain creates protocol-level visibility into an extraction pipeline that currently operates through opaque relay infrastructure. The projected 70% MEV reduction benefits end users — anyone trading, borrowing, or providing liquidity on Ethereum L1 — directly. But as Buterin has acknowledged, ePBS addresses the symptom (relay-mediated extraction) more than the cause (builder market concentration).
The devnet progression from Soldøgn through Devnet-7 to the imminent Devnet-8 suggests Glamsterdam is on the later side of its development timeline rather than at risk of indefinite delay. Public testnet activations on Sepolia and Hoodi remain the key milestones to monitor. Q4 2026 mainnet activation appears probable absent unexpected client implementation failures.
For the broader Ethereum economic model, Glamsterdam changes the pipes but not the water flow. The network's fundamental challenge — generating $65 million in fees while distributing $4–5 billion in subsidies — requires demand growth, not infrastructure expansion. Glamsterdam provides the infrastructure. Demand remains the open variable.