Ethereum's largest protocol overhaul since The Merge is weeks from its first confirmed testnet activation, with the Sepolia fork pinned to September 28.
Ethereum's largest protocol overhaul since The Merge is weeks from its first confirmed testnet activation, with the Sepolia fork pinned to September 28.

Ethereum's Glamsterdam hard fork activates on Sepolia September 28, targeting a 78.6% cut in Layer-1 gas fees and 10,000 TPS. The date, confirmed in the minutes of All Core Devs Consensus call #185 on August 20, marks the first concrete milestone for the upgrade since developers launched Devnet-8 on August 13.
"Since there has been a big change in the block building process, third-party block builders should test their infrastructure against the new protocol specifications starting now," Kristin Kim, founder of Protocol Watch, said.
The upgrade bundles three EIPs. EIP-7928 introduces parallel execution to push throughput to 10,000 TPS. EIP-7904 reprices gas to cut Layer-1 fees by roughly 78.6%. EIP-7732 implements Enshrined Proposer-Builder Separation, targeting a 70% reduction in MEV extraction. Developers also confirmed a block gas limit increase from 60 million to 200 million, with EIP-8261 proposed to raise the limit automatically in stages after the fork.
ETH trades near $2,400, up 3.13%, with RSI at 86.13 — firmly overbought. Google Gemini AI projects a mid-case target of $4,150 by late 2026 if the fork delivers. A delay past Q4 could send ETH below $2,100, with a bear target at $1,850.
The September 28 date is provisional. The ACDC minutes note the decision was deferred to the next call on September 3, and six client teams — Lodestar, ethrex, Lighthouse, Nethermind, Teku, and Prysm — confirmed the epoch without dissent. Lido and Optimism both requested testing windows on a stable DevNet ahead of activation, with Lido's Dima Gusakov saying he would prefer both testnet dates pushed back by about a week.
The Hoodi testnet fork is set for October 26 at 17:42:48 UTC, contingent on stable DevNets. No mainnet date has been set. Earlier projections of a November 4 mainnet fork no longer hold, as that would place it only nine days after Hoodi — unusually tight for Ethereum's standard testnet-to-mainnet gap.
DevNet 8 experienced non-finality six hours before the August 20 call, caused by a caching bug affecting builder deposits at the fork transition. Three consensus clients — Lighthouse, Prysm, and Teku — were affected. Lodestar optimized deposit pre-processing to cut fork transition time from about 20 seconds to roughly 500 milliseconds.
While Glamsterdam heads to testnets, the selection for the follow-on upgrade Hegotá has started. EIP-833, fixing an off-by-one error in checkpoint root calculation, and EIP-8379, making the consensus client the sole sync source, are provisionally scheduled. EIP-7610 was cancelled from Glamsterdam, superseded by EIP-8253 targeting Hegotá.
Two proposals face sharp community disagreement: EIP-8363, which dynamically burns issuance rewards, and EIP-8141, which would let user accounts validate transactions via smart contract logic. Layer-2 projects Base and Offchain Labs oppose EIP-8141, with Base proposing EIP-8130 as an alternative. Client teams must publish priority assessments by mid-September.
Separately, Bitcoin Hyper has raised over $33 million in presale to build an execution environment on Bitcoin using the Solana Virtual Machine, offering yields up to 36% APY with a planned 2026 launch. The project targets high-speed transactions and smart contract functionality for Bitcoin's ecosystem.
The Glamsterdam fork represents a structural shift in what Ethereum's mainnet can handle. If the fee reduction and throughput gains materialize as specified, the range of applications that can run profitably on Layer 1 widens considerably — a direct challenge to the Layer-2 scaling narrative that has dominated Ethereum's roadmap since The Merge. The September 3 ACDC call will determine whether the Sepolia date holds.
This article is for informational purposes only and does not constitute investment advice.