I’ve been watching Ethereum upgrades for years, and every so often one arrives that forces the entire ecosystem to stop and double-check its assumptions. Glamsterdam feels like one of those moments. The core developers have issued a clear warning: wallets, indexers and gas estimators that still treat the classic 21,000-gas ETH transfer as an unbreakable rule are likely to break when the upgrade activates. That single change alone is enough to make a lot of people sit up straighter.
Why Glamsterdam Changes Everything Developers Thought They Knew
Most of us got comfortable with the idea that a basic ETH transfer always costs 21,000 gas. It became muscle memory. Wallet software, portfolio trackers, even some smart-contract libraries baked that number in as a safe maximum. Glamsterdam deliberately breaks that comfort. The upgrade introduces a separate state-gas dimension, meaning the cost of creating new state is no longer hidden inside the old fixed price. Suddenly a transfer to an address that has never held ETH before will cost more than a transfer to an existing account. That difference is not theoretical. It is already visible on the Platåberget testnet.
In my view the most interesting part is not the higher cost itself. It is the fact that software can no longer assume a single fixed gas ceiling or a single fixed price for common operations. The Protocol DevOps team put it bluntly: any tool relying on a hardcoded maximum gas limit will break. That includes a surprising number of production systems still running in the wild.
The 21,000-Gas Assumption Is No Longer Safe
A transfer to an existing account will still land at 21,000 gas. The base transaction cost, cold account access and value-transfer cost remain familiar. But the moment the recipient address has never been seen on-chain, an extra state-gas charge appears. That charge is metered per state byte and applied at runtime. Applications that simply send 21,000 gas and hope for the best will start failing once the upgrade reaches mainnet.
I’ve already seen early test reports where gas estimators returned incomplete figures because they only understood one gas dimension. The second dimension for state creation simply did not exist in their code. Those estimators will need a rewrite, not a patch.
EIP-8037 And The New Price Of Permanent State
The heart of the change is EIP-8037. It treats the creation of new state as a first-class cost. Creating an account, deploying code or writing a fresh storage slot now carries a fixed cost per state byte. The design goal is clear: limit annual state growth to roughly 60 GiB even if the block gas limit sits around 300 million. Under the parameters discussed earlier this year, new account creation could become roughly 8.5 times more expensive and contract deployment costs could rise about tenfold.
That sounds dramatic, yet the intention is long-term health of the chain. State is the permanent memory of Ethereum. Every new account and every new storage slot stays forever. Making that permanence more expensive forces developers to think twice before creating it. I find the approach sensible, even if the short-term friction is real.
Creating new state will no longer be an invisible side effect of ordinary transactions. It becomes a deliberate, metered cost.
Large contracts that deposit a lot of code will feel the change most. The upgrade raises the maximum deployed contract size from 24 KiB to 64 KiB and the maximum initcode size from 48 KiB to 128 KiB, so the capacity is there. The price simply rises in line with the amount of permanent data added.
Wallets, Indexers And Gas Estimators Under Pressure
The Protocol DevOps team listed the three categories most at risk: wallets, indexers and gas estimators. Any system that sets a hard upper bound on gas or assumes every ETH transfer costs exactly 21,000 is a candidate for failure. I’ve spoken with a few infrastructure teams already scrambling to update their estimation libraries. Some of them had treated the 21,000 figure as a constant for years. That constant is about to become a variable.
Indexers face a different problem. They often rely on predictable gas patterns to classify transaction types quickly. When those patterns change, classification logic can start mislabeling ordinary transfers as failed or under-funded. The result is noisy dashboards and confused users.
- Wallets that hard-code gas limits for simple transfers
- Gas estimators that only model one gas dimension
- Indexers that classify transactions by expected gas usage
- Custom scripts that assume a fixed block gas ceiling
- Portfolio tools that display estimated fees based on outdated constants
None of these failures will be dramatic on day one. They will appear as intermittent errors, incorrect fee quotes or silent under-funding of transactions. That is often worse than a clean break, because the root cause is harder to spot.
Platåberget Offers A Real Testing Window
The good news is that developers are not being asked to wait until the last minute. Platåberget launched on August 13 and is designed to stay online for several months. Glamsterdam is scheduled to activate on that testnet on August 20. Public validator and builder deposits are already open. Solo stakers, distributed validator projects, custom software operators and large staking providers have all been invited to participate.
Unlike the short-lived development networks used earlier in the process, Platåberget gives teams time to observe post-upgrade behaviour under realistic conditions. Client images for both consensus and execution layers are available. Lighthouse, Lodestar, Nimbus, Prysm, Teku, Grandine, Besu, Geth, Erigon, Nethermind, Reth and others are listed. Tagged releases remain optional while teams prepare their own builds.
I’ve always preferred longer-lived testnets for this kind of work. You need time to catch the edge cases that only appear after a few days of continuous operation. Platåberget looks well suited to that purpose.
Beyond Gas: Enshrined Proposer-Builder Separation
Gas repricing is only one half of Glamsterdam. The fork also brings enshrined proposer-builder separation, or ePBS, into the core protocol. The split between block building and block proposing becomes part of the consensus rules rather than an external market. A new builder API flow and payload-timeliness checks arrive with it. Infrastructure that sits between proposers and builders will need to adapt.
Block-Level Access Lists form another major piece. These lists record the state locations accessed during execution and the post-transaction state changes. The data is stored separately from the block body and exchanged between execution-layer peers through the eth/71 networking protocol. The practical benefit is that nodes can preload required data and, in some cases, process transactions in parallel when they do not touch the same state.
I see this as a quiet but important step toward higher throughput without sacrificing the security model. Parallel execution only works when the access patterns are known in advance. Block-Level Access Lists give clients that information.
What Application Developers Should Do Right Now
The advice from the Protocol DevOps team is straightforward: test on Platåberget before the upgrade moves to Sepolia and Hoodi. That testing should cover more than happy-path transfers. Teams need to exercise the paths that create new accounts, deploy fresh contracts and write previously unused storage slots. Gas estimators must return correct figures under the new dual-dimension model. Wallets must present accurate fee quotes even when the recipient address is brand new.
- Deploy your estimation libraries against Platåberget and compare results with mainnet behaviour
- Simulate transfers to both existing and never-before-seen addresses
- Verify that under-funded transactions are rejected cleanly rather than silently failing
- Check indexer classification logic against the new gas patterns
- Review any hard-coded gas ceilings in custom scripts or monitoring tools
Perhaps the most interesting aspect is how many systems still treat gas as a static property rather than a dynamic resource. Glamsterdam forces that mental model to update. In my experience the teams that treat upgrades as routine maintenance rather than emergency events come through these transitions with far less stress.
The Longer Road After Platåberget
Once feedback from Platåberget is incorporated into specifications and client software, a non-finality devnet is expected to follow within the month. That network will deliberately test difficult consensus scenarios. Sepolia and Hoodi will receive the upgrade only after the development networks remain stable. Mainnet activation will follow successful upgrades on the long-lived testnets. No fixed mainnet date has been announced, and that is deliberate. Deployment depends on test results and client readiness.
Meanwhile work continues on the next scheduled fork, currently referred to as Hegotá. Developers are reviewing dozens of proposals for a 2027 timeframe. Some of those proposals already look at further gas and state pricing adjustments in case the block gas limit continues to rise. The trajectory is clear: Ethereum is preparing for higher capacity while keeping state growth under control.
Why The Timing Matters For Everyday Users
Most users will never read an EIP. They simply expect their wallet to quote a sensible fee and their transfer to succeed. When that expectation breaks, the experience feels broken even if the underlying protocol is working as designed. That is why the current warning carries weight. The developers are not merely announcing a technical change. They are asking the wider ecosystem to prepare the interfaces that ordinary people actually touch.
I’ve found that the upgrades that generate the most friction are rarely the ones that change consensus rules. They are the ones that quietly alter the assumptions buried inside application code. Glamsterdam is one of those. The 21,000-gas number has been treated as a constant for so long that many systems forgot it was only ever an approximation.
The separate metering of state creation also changes incentives. Applications that frequently create new accounts or write new storage will feel higher costs. Applications that reuse existing state will not. Over time that should push design patterns toward more efficient state usage. Whether that shift happens smoothly depends on how thoroughly the ecosystem tests before mainnet.
Practical Realities Of Dual Gas Dimensions
Gas estimators that only understand a single dimension will return incomplete results once state gas is introduced. The total cost of a transaction becomes the sum of the traditional gas and the new state-gas charge. Tools that ignore the second number will under-estimate. Users who trust those under-estimates may submit transactions that fail. The failure mode is not always obvious. Sometimes the transaction simply never lands. Sometimes it lands with a partial state change that leaves the application in an inconsistent condition.
I keep coming back to the same practical question: how many production systems still treat gas as a constant rather than a calculated value? The answer is higher than most teams would like to admit. Glamsterdam will surface those systems quickly.
Contract Size Increases And What They Enable
Raising the maximum deployed contract size to 64 KiB and the initcode size to 128 KiB removes a long-standing constraint. Complex DeFi protocols that previously had to split logic across multiple contracts can now keep more of that logic in a single deployment. The higher state-gas cost offsets some of that freedom, but the capacity is there. Forward-compatible consensus data structures are also part of the package, which should reduce future friction when further upgrades arrive.
The combination of larger contracts and more expensive state creation creates an interesting design tension. Teams will need to decide which data truly belongs on-chain and which can live off-chain with proofs. That conversation has been happening for years. Glamsterdam makes the economic side of the conversation sharper.
Infrastructure And Staking Providers
Solo stakers and large staking providers are both invited to deposit on Platåberget. The testnet is relatively small but open. The longer lifespan gives operators time to test validator and builder deposit workflows under the new rules. Distributed validator technology projects should pay particular attention to the ePBS changes, because the separation of proposer and builder roles becomes protocol-native rather than market-driven.
I’ve always believed that the health of Ethereum depends as much on the operational layer as on the protocol layer. When staking infrastructure is tested early, the later stages of an upgrade tend to go more smoothly. Platåberget is structured to encourage exactly that kind of early participation.
Looking Ahead Without Fixed Dates
No mainnet activation date has been set, and that absence is intentional. The foundation has repeatedly said that deployment depends on test results and client readiness. After Platåberget, the non-finality devnet will stress consensus edge cases. Only then will Sepolia and Hoodi receive the upgrade. Mainnet follows successful long-lived testnet upgrades. The process is deliberately paced.
Meanwhile the conversation about future capacity continues. Some proposals already explore how state-creation pricing should adjust if the block gas limit moves toward 600 million. Glamsterdam sets a reference point. Later upgrades can recalibrate from that point. The direction of travel is higher throughput with controlled state growth.
In the end the upgrade is less about any single EIP and more about forcing the ecosystem to treat gas and state as dynamic resources rather than fixed constants. That shift is overdue. The friction will be real for teams that left hard-coded assumptions in place. The teams that treat testing as routine rather than optional will barely notice the difference.
Glamsterdam is a reminder that Ethereum still evolves in public, with open testnets and clear warnings. The tools that break will be the ones that stopped evolving with it. The tools that survive will be the ones that treated the 21,000-gas number as a temporary convenience rather than a permanent law. That distinction is worth keeping in mind as the August 20 activation on Platåberget approaches.
The next few weeks will show how thoroughly the ecosystem has prepared. Early testing is free. Late surprises are expensive. The developers have given the community both the warning and the testnet. The rest is up to the people who build the interfaces most users actually touch.
