Ethereum’s Glamsterdam hard fork reached its final devnet stage Tuesday, locking in the EIP bundle that core developers expect to carry the network through public testnets and on to mainnet activation in the second half of 2026. The release is being framed as the largest protocol change since the Merge.

The upgrade ships ten Ethereum Improvement Proposals tracked under the Glamsterdam Meta EIP-7773, with two headliners doing the structural heavy lifting: EIP-7732, which enshrines Proposer-Builder Separation (ePBS) directly in the protocol, and EIP-7928, which introduces Block-Level Access (BALs) Lists so validators can process unrelated transactions in parallel.

The combination clears the path for a 200 million gas-limit floor, roughly tripling current L1 capacity from the 60 million range and unlocking what proponents say is up to 10,000 TPS-equivalent throughput under realistic workloads.

The Full EIP Bundle

The devnet-0 spec published by the EF’s pandaops team lists the included proposals. Beyond ePBS and BALs, the package contains EIP-7708 (ETH transfers and burns emit a log), EIP-7778 (block gas accounting without refunds), EIP-7843 (a SLOTNUM opcode), EIP-7954 (raising the maximum contract size from roughly 24 KiB to 32 KiB), EIP-7975 (eth/70 partial block receipt lists), EIP-8024 (backward-compatible SWAPN, DUPN and EXCHANGE opcodes), EIP-8037 (state-creation gas-cost increase), and EIP-8159 (eth/71 Block Access List Exchange).

The bundle resolves a debate that ran through several All Core Devs calls this spring over whether ePBS and BALs were too ambitious to ship together. The May 2026 finalization of EIP-8037, which sets a fixed cost per state byte and dedicates a separate gas reservoir for state growth, was the final piece that gave client teams a sustainability ceiling under which a 200M gas limit could be raised without bloating the database past 120 GiB per year.

The Two Headliners

ePBS pulls block-building duties into the consensus layer, separating the validator that proposes a block from the builder that constructs the execution payload. The handoff is currently mediated by off-protocol relays like MEV-Boost, which the ethereum.org documentation notes will become optional rather than required once the protocol natively settles builder payments. The change also widens the data-propagation window from two seconds to roughly nine, which is what unlocks the higher gas limit without forcing validators to rush block validation.

Block-Level Access Lists give every block an upfront map of which accounts and storage slots its transactions will touch, plus the post-execution state values. That lets nodes prefetch data in parallel and process non-overlapping transactions concurrently, rather than replaying them serially. BALs also enable executionless sync, where new nodes can update their state from the access-list digest without replaying the full transaction history. The projected throughput gains were laid out in earlier coverage of the framework when the design first crystallized.

Changes for Users

For end users, the most visible change is EIP-2780, which cuts the intrinsic transaction-gas floor and is projected to make standard ETH transfers between existing accounts up to 71% cheaper. EIP-7708 also makes ETH transfers emit a log, which exchanges and wallets have wanted for years because it removes the need for custom transaction tracing.

For validators, ePBS rewrites the builder-selection process and adds a Payload Timeliness Committee that attests separately to consensus blocks and execution-payload timeliness. Staking pools will need architectural updates to monitor the new flow trustlessly, but the user-facing exit process improves through EIP-8080, which lets standard exits borrow unused capacity in the consolidation queue at a three-for-two rate.

For Layer 2s, the wider propagation window means Ethereum can carry more blobs per block, expanding the data-availability budget that rollups draw from. That continues the Fusaka direction of decoupling rollup data costs from L1 execution congestion, alongside parallel research tracks like the post-quantum key registry laid out earlier this month.

No Mainnet Date

A mainnet target slot is not on the table yet. Client teams use the public testnet phase, which follows successful devnet rotation, to set the activation date. Holesky and Hoodi will fork before mainnet, and only after multi-client stability holds for several epochs across those networks.

Past forks have run two to four months of public-testnet seasoning; on that cadence, mainnet would land between September and December 2026.

The 200 million gas limit is the design target for what Glamsterdam unblocks, not a value the fork itself enforces. Validators set the limit via standard gas-vote signaling, which they currently coordinate around the 60 million range, and would step the limit up only as nodes prove they can handle the larger blocks without degraded propagation.