Ethereum Foundation Warns Wallets and Indexers May Break When Glamsterdam Lands
The Ethereum Foundation has warned that some wallets, block explorers, indexers and gas estimators may simply stop working correctly when the Glamsterdam upgrade activates, because of changes to how gas is calculated for transactions that create new state.
In a post published Monday, the Foundation's Protocol DevOps team said any tool relying on a hardcoded maximum gas limit will break and must be updated ahead of the fork. The team urged developers to test their systems on Plataberget, a new public testnet that launched on August 13 and is intended to operate for several months as a staging ground for the changes.
The warning is unusually direct by Ethereum standards, and it lands with a deadline attached. According to upgrade tracker Forkcast, the Glamsterdam fork is scheduled to hit the Platåberget public testnet on Thursday, with the Sepolia and Hoodi testnets to follow ahead of a mainnet date that is still undecided.
What actually changes
At the heart of the warning is EIP-8037, a proposal that introduces a separate state-gas dimension for operations that create new state on the network.
Under the new model, a plain ETH transfer to an existing account would still cost 21,000 gas, the number every Ethereum developer has memorized since the chain launched. But sending ETH to a brand-new account would incur an additional state-gas charge on top, because the transaction does something extra: it expands the state.
That sounds like a small accounting change. It is not, for tool builders. The Ethereum Foundation said developers should revisit any software that assumes 21,000 gas covers every ETH transfer, or that uses only a single gas dimension when estimating transaction costs. Wallets that estimate fees incorrectly could set gas limits too low and see transactions fail, while indexers and explorers that model gas the old way could report distorted data or break entirely.
The two-dimensional gas model is part of Ethereum's longer-term effort to rein in state growth, one of the network's persistent scaling headaches. Making state creation carry its own explicit cost gives the protocol a lever to manage the size of the chain's database, but it ripples outward into every piece of software that touches transaction simulation or fee estimation.
Glamsterdam's bigger picture
Glamsterdam is one of the most consequential Ethereum upgrades in years, and the gas change is only one part of it. The fork also includes enshrined proposer-builder separation, a landmark change that moves the relationship between block proposers and block builders into the protocol itself, along with block-level access lists and increased limits for contract and initialization-code sizes.
The upgrade follows the Foundation's recent Glamsterdam milestones announcement, which also named new protocol leads, part of a broader institutional reshuffle at the organization as it pushes through its most technical roadmap since the Merge.
For everyday users, the Thursday testnet activation should be invisible if the tooling ecosystem does its homework. That is precisely why the Foundation is being loud now. A broken gas estimator or a wallet that cannot predict fees correctly turns a routine upgrade into a user-facing incident, and the blame tends to land on Ethereum rather than on the unmaintained library that caused it.
A testnet built for the long haul
The Plataberget testnet, launched August 13, is designed to give developers months of runway to hammer against the new rules rather than a brief pre-fork window. The Foundation's messaging makes clear it expects infrastructure teams to use it, not to wait for mainnet to expose problems.
History offers a cautionary precedent. Past upgrades that changed transaction semantics or added new transaction types caught slices of the ecosystem off guard, from hardware wallets that could not display new fields to block explorers that misrendered transactions. The more fundamental the change, the wider the blast radius of unprepared tooling.
With a two-dimensional gas model, Ethereum is making exactly that kind of fundamental change, and the Foundation's warning amounts to a checklist for anyone building on the chain: find your hardcoded gas assumptions, test against Plataberget, and ship before Thursday.
For now, the network itself is calm, with Ethereum trading around the 1,900 USD area ahead of the fork. The immediate risk is not market volatility but operational breakage, the unglamorous kind of failure that decides whether an upgrade is remembered as a milestone or a weekend of support tickets.
eip-8037 adding a separate state gas dimension is a bigger deal than people realize, every indexer has to rewrite its assumptions
Rewriting indexer assumptions for a two dimensional gas model is a bigger lift than most teams budgeted for. Expect a wave of broken explorers Thursday.
the wave already started, saw two explorers return wrong gas for new account creation on plataberget last week
Gas estimators failing on account creation txs is going to flood every wallet support channel Thursday afternoon. Most users never read fork notes, the update banner has to do all the work.
Plataberget running for months before mainnet is smart. The deadline is still Thursday and tools always wait for the last minute.
testnet launched aug 13 with mainnet on thursday, barely any soak time imo
sepolia and hoodi coming after mainnet is the wild part to me. youd normally soak on testnets first, feels backwards
backwards is generous. mainnet before sepolia is node teams trusting core devs to be perfect, works until it doesnt
the order is the part nobody defends. forkcast literally shows mainnet thursday and sepolia after, so the soak argument only works if you ignore the schedule
agree on the thin soak time but Plataberget ran alongside devnets for weeks before aug 13. less rushed than it looks
weeks on plataberget only counts for teams that actually tested on it. half the wallet devs out there are learning the state gas charge exists from this article
^ this. two dimensional gas isnt a config tweak, every estimate endpoint needs a schema change. teams that skipped plataberget are about to find out thursday
skipping a months long public testnet to save a sprint of work is wild. plataberget has been live since aug 13, that is plenty of runway to break your estimator
exactly this. my teams gas oracle still returns a flat 21000 for empty account sends and the fix ticket sat untouched since july. thursday is going to be educational
a fix ticket untouched since july while the fork date was public is the whole timeline in one anecdote. thursday support queues will be a museum of this
if your gas estimator hardcodes a max limit in 2026 thats on you tbh. the eip 8037 spec has been public for months
spec being public doesnt help when half the explorer infra out there runs code nobody touched since 2021. the 21k assumption is everywhere
exactly, the 21k constant is baked into embedded libs that shipped years before eip-8037 existed. spec being public fixes nothing if nobody maintains the fork
maintenance debt is the whole story. thursday breaks whatever nobody bothered to redeploy since the merge era, spec public or not
merge era code is exactly the risk. half the indexers out there still special case 21k because the constant predates eip-8037 by years and nobody audited it since
21k for a plain transfer staying sacred while new accounts cost extra is the quiet part of eip-8037. every gas tutorial ever written is now half wrong lol
Every gas explainer on the internet: step 1, a transfer costs 21000. That assumption is in docs, tutorials, Stack Overflow answers, everywhere. The cleanup is going to take years.
Wallets estimating a two dimensional gas model overnight is the real risk. Expect failed transactions with confusing errors all Thursday afternoon.
state gas as its own dimension is honestly the cleanest fix for state growth they could have picked. painful for tool devs but the chain database cannot just grow forever for free
forkcast showing mainnet thursday and sepolia after is the detail that changes everything. soak arguments mean nothing when the schedule runs backwards