The new account creation cost is about to increase 7.3x. That's not a typo. Under EIP-8037, the state-gas dimension will charge 1,530 gas per byte of new state, and the replay tests show 2.7 million transactions will break. But the real story is not the numbers; it's the architecture of the breakage.
Tracing the logic gates back to the genesis block: Ethereum's next scheduled upgrade, Glamsterdam, is not a shard or a rollup. It's a re-pricing of the state itself. Two EIPs, 8037 and 8038, are currently in Review status, and they propose a fundamental split between execution gas and state gas. The goal is to triple network throughput by making state growth expensive enough to slow the bloating of the world state. The cost is measured in millions of potentially broken transactions, immutable contracts that cannot be patched, and an entire toolchain that must be re-engineered before Q4 2026.
I've spent the last decade reading EVM bytecode instead of marketing decks. This upgrade is the most consequential gas mechanism change since EIP-2929. And the more I dig into the replay data, the more I see a pattern: the Ethereum Foundation is doing everything right on process, but the ecosystem is sleepwalking into a compatibility cliff.
Let's start with the mechanics. EIP-8037 introduces a separate state-gas dimension, decoupled from execution gas. Every new byte of state—whether it's a new account, a new storage slot, or a new contract code—will be charged 1,530 gas per byte. The current cost for a new account is 25,000 gas. Under the new regime, that same account creation will cost 183,600 gas. That's a 7.3x increase. A new storage slot goes from 20,000 to 97,920 gas—4.9x. Deploying a 24 KiB contract with a new account? From 4,947,200 to 37,784,880 gas—7.6x. These aren't marginal adjustments; they're structural shifts in the economics of state creation.
The rationale is sound. State growth is the silent killer of node sync times. The Ethereum Foundation's own benchmarks, based on a March 2026 mainnet snapshot, show that the current pricing model doesn't reflect the true resource cost of state. By separating state gas, they can target an annual state growth of 120 GiB under a 150 million reference block limit, and 160 GiB in the worst case of a 200 million gas ceiling. That's a deliberate throttle on the state explosion.
But here's where the narrative diverges from the code. The replay tests are the most revealing artifact. The Foundation replayed 929,731,274 transactions from December 2024 to June 2026 under the proposed rules. Under EIP-8037 alone, 174 million transactions are repairable—meaning they'd fail but could be fixed with code changes. 2,687,652 transactions are potentially broken—they'd fail and cannot be automatically repaired. Under EIP-8038, the numbers are 84.7 million repairable and 3,036,537 potentially broken. These aren't abstract figures. They map to real protocols: Across, Socket/Bungee, CoW Protocol, 0x, and the entire ERC-4337 smart account stack.
Read the assembly, not just the documentation. The Foundation's outreach report explicitly names eth-infinitism's ERC-4337 EntryPoint as a priority. That's because smart accounts are the highest-risk category. The EntryPoint contract is immutable. If the gas re-pricing breaks its internal logic—specifically the way it handles gasleft() and validation loops—there's no patch. You need a new EntryPoint, new account implementations, new factories, and a migration path for every existing smart account. That's not a weekend project.
The counterfactual nature of these failures is the first blind spot. The replay tests simulate what would happen if the new rules were applied to historical transactions. But the tests assume the same code paths. In reality, many contracts will be updated before the fork. The Foundation's outreach is designed to catch the big names. But what about the long tail? The thousands of small DeFi protocols, NFT marketplaces, and experimental contracts that haven't been touched in years? They won't get the memo. They'll break silently, and their users will blame the network, not the code.
Here's the contrarian angle: the 3x throughput claim is an engineering target, not a guarantee. The Foundation is careful to say that the upgrade "could" triple network speed, but the actual capacity increase depends on how the new gas dimensions interact with block construction. The state-gas mechanism might actually reduce the number of state-heavy transactions per block, which could offset the throughput gains. The replay tests show that a 10x gas limit increase was used to stress-test the system—that's a hint that the Foundation expects gas demand to grow, but it's not a promise of immediate capacity.
And then there's the extrapolation problem. The state growth projections are based on an incentive model, not on measured behavior. The 387 GiB figure that gets thrown around is an extrapolation from a 30 million to 60 million gas limit scenario, but the response is non-linear. The Foundation's own data shows that the relationship between gas limit and state growth is not a simple multiplier. So the 120 GiB target is a hope, not a forecast.
The deeper issue is the immutability trap. The upgrade will render a class of contracts permanently broken. The Foundation's solution is to encourage new implementations, but that's a cost that falls on developers and users. For a protocol like CoW Protocol, which relies on a specific gas pattern for its settlement logic, the fix might be straightforward. But for a contract that's been abandoned since 2021, with no maintainer and no funds, the upgrade is a death sentence. These zombie contracts will remain on-chain, consuming state, but they'll be inert. The network will carry them as dead weight, and the state-gas tax will make it even more expensive to interact with them.
I've seen this pattern before. In 2020, when I was analyzing Synthetix's oracle vulnerabilities, the community dismissed the risk because the exploit hadn't happened yet. The same logic applies here: the replay tests show that 0.3% of transactions will break, and the market shrugs. But 0.3% of 929 million is still 2.7 million transactions. And those transactions are concentrated in the most critical infrastructure—bridges, DEX aggregators, and smart accounts. A single bridge failure can cascade into a liquidity crisis. The cross-chain bridge sector has already lost over $2.5 billion to hacks; now we're adding a self-inflicted compatibility risk.
The infrastructure layer is the silent victim. Wallets, RPC providers, indexers, node tools, and gas estimators all need to understand the new state-gas rules. The Foundation has announced the Platåberget testnet, but the warning is clear: old tools may not work. That's a massive coordination problem. Every wallet that does gas estimation, every indexer that tracks state changes, every bundler that packages user operations—they all need to be updated. The ERC-4337 bundlers are particularly exposed because they need to distinguish between repairable failures and validation protection failures. If they get it wrong, they'll drop valid transactions or accept invalid ones.
Let me give you a concrete example from my own audit work. I've been reviewing the gas logic in several ERC-4337 EntryPoint implementations. The current code uses gasleft() to determine how much gas is available for validation. Under EIP-8037, the state-gas dimension is separate, but the EVM's gasleft() opcode still returns the total gas remaining. If a bundler doesn't account for the new state-gas costs, it might underestimate the gas required for a user operation, leading to out-of-gas failures. The fix is not trivial; it requires changes to the EntryPoint's internal accounting and to the bundler's estimation logic. This is exactly the kind of subtle breakage that replay tests can't fully capture because they don't model the dynamic behavior of gas estimation.
The Foundation's transparency is commendable. They've published a public dashboard, they've done the replay tests, and they're doing outreach. But transparency doesn't fix the fundamental asymmetry: the upgrade is being designed by a small group of core developers, and the impact is borne by the entire ecosystem. The EIP process is open, but the timeline is fixed. Q4 2026 is not far away. The Review stage could last months, and the final EIPs might change. But the direction is clear: state is becoming more expensive, and the cost of creating new state is going to be a significant factor in every contract's gas budget.
What does this mean for the average developer? If you're building a new contract, you need to design for state efficiency from day one. The old patterns of creating a new storage slot for every user action are no longer viable. You'll need to batch state updates, use more efficient data structures, and consider whether you really need to store that data on-chain. The upgrade is effectively a tax on state, and the tax rate is high.
For the protocols that are already live, the clock is ticking. The Foundation's outreach report identifies the key players, but it's not exhaustive. If you're running a DeFi protocol that uses a bridge or a smart account, you need to run your own replay tests. The public dashboard is a starting point, but it doesn't cover every contract. You need to simulate your own historical transactions under the new rules and see what breaks. I've done this for a few clients, and the results are often surprising. A simple change in the way a contract reads a storage slot can trigger a 4.9x cost increase, which might push a transaction over the block gas limit.
The contrarian takeaway is that the upgrade might not actually triple throughput. It might just shift the bottleneck from execution to state. If the state-gas dimension becomes the limiting factor, then the network's capacity will be determined by how much state can be created per block, not by how much computation can be executed. That could lead to a situation where simple transfers are cheap, but complex DeFi operations become prohibitively expensive. The 3x claim is based on the assumption that execution gas is the current bottleneck, but the data suggests that state growth is the real constraint. The upgrade might just move the problem, not solve it.
And then there's the political dimension. The Ethereum Foundation is trying to control state growth to keep node requirements low, which is good for decentralization. But the side effect is that it makes the network less friendly to new applications that require heavy state. This is a trade-off that the community hasn't fully debated. The upgrade is being framed as a performance improvement, but it's really a resource allocation policy. The question is: who gets to decide what state is worth keeping? The answer is the gas price, and that's a market mechanism, but it's a market that's being redesigned by a few core developers.
I'm not saying the upgrade is bad. I'm saying it's a fundamental change that deserves more scrutiny than it's getting. The replay tests are a good start, but they're not sufficient. The Foundation needs to do more than publish a dashboard; it needs to actively work with every major protocol to ensure they're ready. The outreach report is a step, but it's a small step. The long tail of protocols is where the real risk lies.
In my experience, the most dangerous failures are the ones that don't show up in tests. The counterfactual failures are predictable; the real surprises come from interactions between different contracts. A DEX aggregator that calls a bridge that calls a smart account—each one might be fine individually, but the combination could break under the new gas rules. The replay tests don't capture these composite interactions because they replay individual transactions, not entire workflows.
The upgrade is coming. The question is not whether it will happen, but how much collateral damage it will cause. The Foundation has set a target of Q4 2026, which gives the ecosystem about six months to prepare. That's not a lot of time. The infrastructure providers need to update their tools, the protocol teams need to update their contracts, and the users need to be educated about the changes. It's a massive coordination problem, and the coordination is happening in public, but it's still a race against the clock.
Here's my forward-looking judgment: the upgrade will go through, but it will be messy. There will be a period of instability after the fork, with some protocols failing and others scrambling to fix issues. The market will see this as a negative, and there will be FUD. But the long-term effect will be positive, because the state-growth control will make Ethereum more sustainable. The network will be faster, more efficient, and more decentralized. The cost is the short-term pain.
The ones who will thrive are the ones who read the assembly, not just the documentation. The ones who understand the gas mechanics, who have already run their own replay tests, who have updated their contracts and their tools. The ones who are prepared for the state-gas tax. The ones who see this as an opportunity to optimize their code and gain a competitive advantage.
The rest will be caught off guard. They'll be the ones complaining on Twitter about broken transactions and high fees. They'll be the ones who didn't bother to read the EIPs, who thought the upgrade was just another routine improvement. They'll be the ones who learn the hard way that in Ethereum, the only constant is change, and the change is always more expensive than you think.
Tracing the logic gates back to the genesis block, I see a pattern: every major upgrade has broken something. The DAO fork broke the immutability narrative. EIP-2929 broke gas estimation. EIP-1559 broke the fee market. And now Glamsterdam will break a generation of contracts. The question is not whether it will break, but whether we're ready to fix it. I've been auditing smart contracts for a decade, and I've never seen a more predictable disaster. The data is there. The replay tests are there. The outreach is there. The only missing piece is the will to act.
So, to every protocol developer reading this: run your own replay tests. Check your gas usage. Update your contracts. And to every user: be prepared for some friction. The upgrade is a necessary step, but it's a step that will leave some behind. The ones who survive will be the ones who understand that the code is the truth, and the truth is about to change.