A lot of wallets still assume 21,000 gas is enough for a basic ETH transfer. It’s practically muscle memory at this point. But that habit is about to get you in trouble.
Glamsterdam, Ethereum’s next big fork, is reshaping how gas is priced. The number 21,000 doesn’t go away, but what it means does. And if you send ETH to a fresh address, the old mental model won’t save you.
This isn’t theory. Developers have finalized repricing, public testnets are queued up, and the block target is stepping into a new range.
Ethereum’s core teams are pushing a repricing package that dismantles the flat “21k covers a transfer” assumption and folds in a real charge for state growth. In parallel, they’re preparing the chain for much higher per-block gas capacity.
The 21k number was a budget. Not a right. Glamsterdam makes that explicit by unbundling costs and putting a price on new state.
On the capacity side, the official Glamsterdam roadmap was updated on August 6, 2026 with a 150 million reference block gas limit and a configurable path toward a 200 million gas-limit floor, positioning a safer L1 scaling route (ethereum.org (Glamsterdam roadmap)). One day earlier, core devs agreed to launch the first public testnet, named Platåberget, permissionlessly, starting pre-fork and activating Glamsterdam about a week after genesis, with a setup described as roughly 50,000 validators across about 50 nodes (Etherspot (blog)).
Follow-up calls confirmed the repricing figures as final and adopted the 200M gas target path, with notes citing a repricing package around EIP-2780, EIP-8037, and EIP-8038 (KuCoin syndicating EtherWorld ACDE coverage). The 21k era isn’t exactly ending, but the old shortcut will be unsafe by default.
How we got stuck on 21,000 gas
For years, “send ETH, spend 21,000 gas” was the simple story. Wallets hardcoded it. Exchanges used it as a sanity check. Even devs treated 21k like a constant of nature.
That worked because the fee model bundled several costs into one flat number. You could ignore the details and still get a predictable experience for a plain value transfer. But the chain has grown up. We’ve added more opcodes, more state, more caching rules, and a lot more users. Cheap assumptions tend to age badly in that kind of environment.
The problem with a flat intrinsic is that it hides the real driver of long-term costs: state growth. New accounts and fresh storage are not free to carry. They add weight to the network. So when usage spikes in ways that grow state quickly, you want pricing that reflects it, or you get misaligned incentives.
Why this matters now
Glamsterdam is the moment when Ethereum makes that alignment more explicit. The base fee market (EIP-1559) remains, but the guts of what a “simple transfer” costs will depend on what you’re touching and whether you’re growing the state. Which means the old 21k mental model breaks the moment you target a fresh address without a careful estimator.
Inside Glamsterdam’s gas repricing
The spec decomposes the old intrinsic into visible parts. Think of it like pulling the cover off a machine you used to treat as a black box.
Per the repricing meta/spec around EIP-8007 and EIP-2780, the legacy flat 21,000 intrinsic gas is expressed as explicit primitives:
- TX_BASE_COST: 12,000
- COLD_ACCOUNT_ACCESS: 3,000
- TX_VALUE_COST: 6,000
- TRANSFER_LOG_COST: 1,756
These add up to the familiar 21,000 for an ordinary transfer path. But there’s a second piece: a separate state-gas charge for new accounts based on bytes written. The spec references a multiplier CPSB; for example, with CPSB equal to 1,530, the new-account state charge is roughly 183,600 gas. That’s on top of the base primitives (EIP‑8007 / EIP‑2780 (EIPs repository)).
What does that mean in practice? If you send ETH to a truly new externally owned account that doesn’t exist yet, your transaction could incur a hefty one-time state charge. If you send to an existing, already-in-state account, you avoid that state growth cost.
Contrast at a glance
| Scenario | Pre-Glamsterdam mental model | Glamsterdam repriced model |
|---|---|---|
| Transfer to existing EOA | ~21,000 intrinsic | Primitives sum to ~21,000; no new-account state charge |
| Transfer to new EOA | ~21,000 intrinsic | Primitives (~21,000) + new-account state charge (example magnitude ≈ 183,600 if CPSB=1,530) |
| Transfer to contract | Varies by code path, often estimated | Primitives + any touched cold slots + execution; state growth priced more directly |
There’s nuance on the contract side. Many dapps already estimate gas properly because execution paths were never 21k. But wallets and services that special-case value transfers need to revisit assumptions. The repricing also creates clearer incentives around state hygiene: reusing accounts or paying via smart accounts that aggregate operations may be more gas-efficient than spraying fresh addresses for every payout.
Block limits: heading to the 200M era
Repricing is one half of the story. The other half is capacity. Ethereum’s L1 is moving toward larger blocks, with a reference around 150M gas and a target floor around 200M as part of the Glamsterdam plan. The point isn’t just bigger blocks for the sake of it. It’s to create headroom while repricing state correctly so throughput doesn’t come at the cost of a swollen, underpriced state.
The signposts are clear:
| Date | Milestone | Why it matters |
|---|---|---|
| Aug 5, 2026 | ACDE #242 names “Platåberget” public testnet; permissionless launch, ~50,000 validators on ~50 nodes | Operational path for broad testing without gatekeepers (Etherspot) |
| Aug 6, 2026 | Roadmap page updates with 150M reference block gas, target path to 200M floor | Explicit capacity targets published for Glamsterdam (ethereum.org) |
| Mid Aug 2026 | ACDE #243 confirms final gas-repricing figures and 200M target path | Spec stability for client teams and infra (KuCoin) |
Throughput vs state growth
More block gas capacity could reduce inclusion delays for bursts of simple activity, but it won’t magically make state growth cheap. That’s the whole point of splitting the costs. Expect a future where basic transfers to known accounts are easier to include when blocks are roomy, while minting lots of new state stays properly pricey.
What it means for rollups and MEV
Rollups live downstream of L1 pricing. Bigger L1 capacity, paired with realistic state costs, might change the timing of batch posting and how sequencers think about cross-rollup settlement. On MEV, more capacity can compress some priority gas auctions, but new arbitrage edges often appear when fee markets change. Builders and searchers will adapt quickly; that is the only constant.
What this means for wallets, dapps, and exchanges
Three buckets of impact: estimation, UX, and treasury ops.
Estimation: 21k is no longer safe
If your code path special-cases a plain value transfer at 21,000 gas, treat that as a bug. You need a real estimator for whether the destination exists and what state is touched. Many libraries can already tell you if an address is empty; now that check may directly change the gas you allocate.
For account abstraction flows, bundlers and paymasters need to model the state charge. The nightmare scenario is a paymaster fronting fees for users who unintentionally create a wave of new accounts and blow up the budget. Put guards in place.
UX: messaging and fees
Users will ask why sending to a fresh address costs more. Explain it in plain terms: new addresses add weight to the network, and the network is pricing that explicitly. Good wallets can soften the blow with smart defaults: suggest reusing known addresses when safe, or batch small payouts to reduce creation churn.
Treasury ops: exchanges and services
Exchanges that fan out withdrawals to never-before-used addresses for every customer might see higher costs at the margin. Payment processors that rotate fresh deposit addresses constantly should run the numbers again. Some policies were designed in a world where the extra cost was negligible. That world is changing.
Preparing for the break: a practical checklist
Here’s a simple sequence that teams can run through before Glamsterdam hits mainnet.
- Audit all 21k assumptions. Search your codebase for hardcoded 21000 or “intrinsic” shortcuts and remove them.
- Detect address existence. Before estimating, check if the destination address is empty to model potential state charge.
- Update gas estimators. Use node-level simulation or reputable libraries that reflect the repriced primitives and state costs.
- Test on Platåberget. Spin up infra against the permissionless public testnet once live to validate end-to-end flows (Etherspot).
- Stress smart-account flows. Model paymaster exposure and add caps. Make sure wallet UIs communicate when a fresh-address charge applies.
- Revisit payout policies. If you rotate deposit addresses or “spray and pray” withdrawals, compare costs for reuse vs creation under repricing.
- Monitor client releases. Track the finalized parameters and any last-mile tweaks from client teams as testnet feedback rolls in (ACDE coverage; ethereum.org).
Risks & What Could Go Wrong
- Wallet breakage. Any product that assumes 21k for transfers could mis-price fees and strand users in pending limbo.
- Paymaster drain. Poorly capped account abstraction setups might subsidize a wave of new-account creations.
- Mempool churn. If many txs underbid due to stale estimators, you get more replacements and congestion churn.
- State-growth whiplash. If parameters are off, state growth could still outpace pricing and stress node storage.
- Propagation pressure. Higher block capacity combined with busy periods can stress p2p propagation for some nodes.
- UX confusion. Users will see inconsistent fees between sending to known vs new addresses and interpret it as a bug.
The repricing fixes incentives, but only if the ecosystem upgrades its habits. Treat 21k as a memory, not a guarantee.
Frequently Asked Questions
Does 21,000 gas go away entirely?
No. The sum of primitives for a basic transfer still adds to roughly 21,000. What changes is that new-account creation incurs a separate state charge. So 21k is no longer a safe one-size-fits-all number.
Will a 200M gas target make fees cheaper?
It could reduce inclusion delays and help during spikes, but base fee still responds to demand. More capacity doesn’t guarantee cheaper fees. The repricing mostly ensures state growth is not underpriced.
How do I know if a transfer will trigger the new-account state charge?
Check if the destination address is empty on L1. If it has no code and no nonce and no balance, creation costs can apply. Estimators or a quick on-chain existence check should inform the gas budget.
What if my wallet only supports a fixed gas limit for transfers?
Update it. Hardcoding 21k will fail in edge cases. At minimum, integrate a node simulation step or an existence probe and adjust the limit dynamically.
What’s the status of Glamsterdam’s parameters?
Developers confirmed the latest repricing figures as final and adopted a path toward a 200M gas target during ACDE #243, and the roadmap lists a 150M reference with a configurable path to 200M. Follow client releases for any polishing on testnets (KuCoin; ethereum.org).
Where can I test my flows before mainnet?
Platåberget, the first public Glamsterdam testnet, is planned as a permissionless network that starts pre-fork and activates the upgrade about a week after genesis. Expect on-ramps for around 50,000 validators across roughly 50 nodes (Etherspot).
Does this impact smart accounts and EIP-4337 users?
Yes. Bundlers and paymasters must account for the state charge when users create or touch fresh addresses. Build in caps, clear messaging, and smarter estimation to avoid fronting unexpected costs.
