What Replaces the Flat 21,000 Gas Fee?
For a decade, "a simple ETH send costs 21,000 gas" has been the closest thing Ethereum had to a law of physics. Ethereum's Glamsterdam upgrade repeals it, and the tools most retail traders touch every day are the ones written against the old constant.
What replaces the flat 21,000 gas fee is a matrix of six distinct minimums defined by EIP-2780, "Resource-based Intrinsic Transaction Gas." The EIP decomposes the old constant into named components — TX_BASE_COST of 12,000 gas for signature recovery, sender account access and block inclusion; COLD_ACCOUNT_ACCESS of 3,000 for touching the recipient; and TX_VALUE_COST of 6,000 for the recipient balance write and transfer log . Those three still sum to 21,000, but only for a value transfer to an existing, distinct account.
Quick Answer: Ethereum's Glamsterdam upgrade, planned for mainnet in Q4 2026, replaces the flat 21,000-gas transfer cost with six resource-based minimums under EIP-2780. A transfer to a brand-new address jumps to 204,600 gas — 21,000 execution plus 183,600 state gas — nearly 10× today's charge .
The Ethereum Foundation's Protocol DevOps team put the warning in writing on August 17, 2026, in the blog post announcing the Platåberget public testnet:
"Any tool that relies on a hardcapped maximum gas limit (think wallets, indexers and gas estimators) will break and needs to be updated… a plain ETH transfer is no longer always flat 21,000 gas," — Protocol DevOps team, Ethereum Foundation (source: blog.ethereum.org, 2026-08)
Here is the replacement schedule as written in the current EIP text :
| Transaction type | Minimum gas after Glamsterdam | Change vs. today |
|---|---|---|
| ETH self-transfer | 12,000 | Cheaper |
| Zero-value transaction to an EOA or empty account | 15,000 | Cheaper |
| Value transfer to an existing account | 21,000 | Unchanged |
| Transfer to an EIP-7702 delegated account | 24,000 + execution | More expensive |
| Zero-value contract creation | 24,000 intrinsic + state gas | More expensive |
| Transfer creating a new account | 21,000 execution + 183,600 state = 204,600 | ~10× more expensive |
The structural change matters more than any single number. EIP-2780 splits gas into intrinsic costs — state-independent, and used to determine whether a transaction is even valid — and runtime costs that depend on chain state and are charged only when incurred . The EIP is explicitly not backward compatible, and it is tracked in the gas-repricing meta-EIP EIP-8007 inside the wider Glamsterdam bundle, EIP-7773, which lists 19 scheduled EIPs plus 5 networking and 2 informational entries . All of these documents remain in Review or Draft status, so implementers should track canonical EIP text and client test fixtures rather than secondhand summaries .
Why 'Send Max' and Empty-Wallet Flows Break First
The first failures will land in "send max" and "empty wallet" flows, because those code paths compute a transferable remainder by subtracting an assumed fee from the balance — and the assumed fee is almost always built on 21,000 gas. If the recipient address has never been used, the real charge under EIP-2780 and EIP-8037 is 21,000 execution gas plus 183,600 state gas, or 204,600 total — close to 10× the assumed figure. The transaction underfunds and fails. The Ethereum Foundation's Protocol DevOps team put it directly in its August 17, 2026 Platåberget announcement: "any tool that relies on a hardcapped maximum gas limit (think wallets, indexers and gas estimators) will break and needs to be updated" (source: Ethereum Foundation, 2026-08).
The affected surface is wider than the wallet UI. Anywhere 21,000 is used as a validation floor, a fallback gasLimit, or a human-facing "simple transfer" estimate needs review:
- Wallets — send-max and empty-wallet remainder math, plus fee previews shown before signing
- Gas-estimation libraries and RPC providers — single-dimension estimators cannot model the split between intrinsic and state-dependent charges
- Transaction builders and bundlers — hardcoded limits on user operations and batched sends
- Exchanges and custodians — withdrawal fee schedules and sweep logic that assume a flat per-transfer cost
- Indexers and explorers — cost labelling and reconstruction of "standard transfer" categories
- dApps with custom send flows — any bespoke path that skips the wallet's own estimator
Contract-level senders face the same arithmetic. DeFi systems that programmatically push ETH to fresh addresses — airdrop distributors, payroll and streaming contracts, onboarding flows that fund new users — are precisely the workloads that create accounts, so they absorb the full state-gas charge on most of their outbound transfers and must re-budget gas limits accordingly (source: crypto.news, CoinDesk).
Two details matter for scoping the work. First, EIP-2780 is explicitly not backward compatible, so there is no graceful-degradation path — a stale estimator produces failed or underfunded transactions rather than merely inaccurate previews (source: EIP-2780). Second, EIP-8037 notes that deterministic deployment factories with hardcoded gas limits will fail on networks that activate the change, while factories already deployed on mainnet and existing L2s are unaffected; EIP-7997 adds a Deterministic Factory Predeploy to address cross-chain address consistency (source: EIP-8037). Adjacent repricings compound the audit: EIP-7976 moves the calldata floor to a uniform 64/64 gas per byte and requires wallets to update eth_estimateGas for the new parameter or risk underestimating gas, even though analysis over blocks 20644414–23272414 found only 1.5% of transactions would have been affected (source: EIP-7976).
The Full Repricing Bundle: EIP-8037 and EIP-8038
EIP-8037, "State Creation Gas Cost Increase," is the proposal that adds a second gas dimension to Ethereum: state gas, metered at runtime and priced at a cost-per-state-byte (CPSB) of 1,530 gas . Applied to the byte sizes the spec assigns to each object, that produces 183,600 gas for a new account modeled at 120 bytes, 97,920 gas for a new storage slot at 64 bytes, and 35,190 gas for an EIP-7702 delegation indicator at 23 bytes . The new-account figure is roughly a 7× increase over today's approximately 25,000 gas (source: EIP-8037).
The mechanism matters as much as the numbers. EIP-8037 uses a reservoir model: a transaction still carries a single gas limit field, but at transaction start that limit is split into execution gas and a state_gas_reservoir, with state-growth charges drawing from the reservoir first while execution charges draw only from gas_left . For wallet and estimator teams that rules out the tempting shortcut of replacing one constant with a larger one — fee estimation and display have to model state-independent intrinsic gas and state-dependent runtime charges as two separate quantities (source: EIP-8038).
EIP-8038, "State-Access Gas Cost Update," supplies the component prices that feed EIP-2780 and the opcode layer. Its authors are Maria Silva, Wei Han Ng and Ansgar Dietrichs, and it requires EIP-8037 .
| Parameter | Current | Post-Glamsterdam | Change |
|---|---|---|---|
| COLD_ACCOUNT_ACCESS | 2,600 | 3,000 | +15% |
| ACCOUNT_WRITE | 6,700 | 9,000 | +34% |
| STORAGE_WRITE | 2,800 | 10,000 | +257% |
| CREATE_ACCESS | 7,000 | 12,000 | +71% |
| STORAGE_CLEAR_REFUND | 4,800 | 11,616 | +142% |
| Access-list address | 2,400 | 2,900 | +21% |
| Access-list storage key | 1,900 | 2,000 | +5% |
| COLD_STORAGE_ACCESS | 2,100 | 2,100 | unchanged |
| WARM_ACCESS | 100 | 100 | unchanged |
The stated motivation is twofold. First, harmonization: contract deployment currently runs at roughly 200 gas per byte while storage slots run at roughly 313 gas per byte, an inconsistency the CPSB constant removes . Second, state growth: after the gas limit rose from 30M to 60M, average daily new state more than tripled, from about 105 MiB to about 326 MiB . EIP-8037 was created on 2025-10-01 by Maria Silva, Carlos Perez, Jochem Brouwer, Ansgar Dietrichs, Łukasz Rozmej, Anders Elowsson, Francesco D'Amato and Dragan Rakita, and both proposals remain in Review status, with the gas-repricing bundle tracked under meta-EIP EIP-8007 and the wider fork scope under EIP-7773 . The practical read for traders and builders: contracts that mainly read or modify existing state are largely unaffected, while anything that writes new state pays materially more.
Timeline: Platåberget Testnet to Q4 2026 Mainnet
Glamsterdam's rollout runs through a long-lived public testnet before it touches mainnet. Platåberget forked to Glamsterdam rules on August 20, 2026, and the Ethereum Foundation's Protocol DevOps team intends it to run for several months — deliberately longer than the short-lived devnets that preceded it — as a stable surface for wallet, dApp and infrastructure teams to test against before Sepolia and Hoodi . Mainnet activation is targeted for Q4 2026 . That gives teams roughly a quarter of live testing runway, not a scramble in the final weeks.
Platåberget was built to behave like a real network rather than a lab fixture. Genesis validators were bootstrapped from public deposits rather than pre-allocated, targeting roughly 50,000 validators across about 50 nodes running mixed client combinations, with a faucet live at faucet.plataberget.ethpandaops.io . For a wallet team, that means send-max flows, fresh-recipient transfers and estimator logic can be exercised against realistic client diversity now, instead of discovering divergence on Sepolia.
Key dates and requirements as they currently stand:
- August 17, 2026 — Ethereum Foundation announces Platåberget and warns that wallets, indexers and gas estimators relying on a hardcapped maximum gas limit will break .
- August 20, 2026 — Platåberget forks to Glamsterdam rules and begins its multi-month run .
- Sepolia and Hoodi — activation epochs still marked TBD by client teams under meta-EIP EIP-7773, which remains in Draft .
- Q4 2026 — targeted mainnet activation, requiring updated execution and consensus client releases from every node operator .
One deadline is easy to underweight because the backtest looks benign. EIP-7976 moves the calldata floor to a uniform 64/64 gas per byte with TOTAL_COST_FLOOR_PER_TOKEN set to 16, cutting worst-case block size by roughly 37% — from about 1.07 MB to about 0.67 MB. Analysis across blocks 20644414 to 23272414 found only 1.5% of transactions would have been affected . That small share is not permission to skip the work: the EIP explicitly requires wallets to update eth_estimateGas for the new parameter, and teams that do not risk underestimating gas and shipping failed transactions to users . EIP-7976, like the rest of the repricing bundle, is still in Review, so implementers should track canonical EIP text and client test fixtures rather than freezing against today's numbers.
Bull Case: Repricing Clears the Runway to 200M Gas
The bull case for Glamsterdam is that repricing state growth is the precondition for raising Ethereum's gas limit from roughly 60M toward the 200M target core developers aligned on at the Soldøgn interop event . Under today's schedule, a new account costs about 25,000 gas — roughly one-seventh of the 183,600 gas EIP-8037 assigns it once state is metered at 1,530 gas per byte across a 120-byte account record . Raising the block ceiling without fixing that gap would multiply the cheapest way to bloat the state. The evidence is already on record: after the limit moved from 30M to 60M, average daily new state more than tripled, from about 105 MiB to about 326 MiB .
The throughput unlock does not come from repricing alone. Glamsterdam's two headline EIPs do the structural work:
- EIP-7732 (Enshrined Proposer-Builder Separation) — pulls the MEV-Boost relay dependency, used in 80–90% of blocks today, into the protocol itself and widens the data propagation window from roughly 2 seconds to about 9 seconds .
- EIP-7928 (Block-Level Access Lists) — makes access lists mandatory at the block level, enabling parallel execution and faster sync .
- EIP-7778 — stops SSTORE refunds from reducing gas counted against the block limit, closing what researchers describe as "block gas smuggling," while still crediting the refund to the user .
Execution risk also looks lower than a typical single-team upgrade. The Ethereum Foundation's Soldøgn interop recap reports that more than 100 core contributors worked on Glamsterdam in late April and early May 2026, where the final EIP-8037 repricing numbers were locked . That breadth of participation is consistent with the Platåberget testnet design, which bootstrapped roughly 50,000 validators across about 50 nodes on mixed client combinations from public deposits rather than pre-allocations .
The Ethereum Foundation's Protocol DevOps team framed the tooling requirement bluntly in the Platåberget announcement: "any tool that relies on a hardcapped maximum gas limit (think wallets, indexers and gas estimators) will break and needs to be updated" (source: Ethereum Foundation, 2026-08). Read constructively, that is the price of admission for a higher ceiling.
For ETH holders, the distinction that matters is structural versus cyclical. This is not a fee event that resolves in a week. Accurate state-growth pricing changes what the network can safely carry — and the direction is mixed rather than uniformly inflationary, since base costs fall for self-sends (12,000 gas) and zero-value transactions (15,000 gas) while state creation gets materially more expensive . One caveat belongs in any bull case built on the 200M figure: the gas limit is set by validator gas-vote signaling, not enforced by the fork itself . Glamsterdam removes the technical objection to raising it; validators still have to vote it up.
Bear Case: Wallet Breakage Is a Real Near-Term Risk
The bear case is an integration risk, not a protocol risk: EIP-2780 is explicitly not backward compatible, so any team that misses the update window ships failed or underfunded transactions to real users on day one . The Ethereum Foundation's Protocol DevOps team put it directly in its August 17, 2026 Platåberget announcement — "any tool that relies on a hardcapped maximum gas limit (think wallets, indexers and gas estimators) will break and needs to be updated" . That is a broad blast radius: RPC providers, gas-estimation libraries, transaction builders, bundlers, exchanges, explorers, custodians and dApps with custom send flows all sit inside it .
The economic edge of the bear case is the new-account charge. A transfer that creates a fresh account costs 21,000 execution gas plus 183,600 state gas under EIP-8037's 1,530-gas cost-per-state-byte model — roughly a 7× increase over today's ~25,000 gas for the same operation . Products whose core loop is paying strangers absorb that first: onboarding flows that fund new wallets, airdrop distributors, and payroll contracts that batch ETH to previously unused addresses all need re-budgeting rather than a config tweak .
Contract-level breakage is narrower but sharper. EIP-8037 notes that deterministic deployment factories carrying hardcoded gas limits will fail on networks that activate the EIP, though factories already deployed on mainnet and existing L2s are unaffected until activation; EIP-7997 adds a Deterministic Factory Predeploy to address cross-chain address consistency .
The subtler risk is partial compliance. Glamsterdam's repricing arrives as a bundle — EIP-2780, EIP-8037, EIP-8038, EIP-7976, EIP-7778 and EIP-7708 — and each touches a different layer of the stack. EIP-7976 alone requires wallets to update eth_estimateGas for the new 64/64-per-byte calldata floor or risk underestimating gas and producing failed transactions . An updated wallet paired with a stale indexer or an unpatched estimator can still surface a confidently wrong number, and the failure mode is silent underestimation rather than a loud error. Compounding it, EIP-2780, EIP-8037, EIP-8038 and EIP-7976 all remain in Review status, and EIP-2780's own rationale calls TX_BASE_COST and TX_VALUE_COST provisional and benchmark-derived . Teams that hardcode against today's draft numbers may need a second update pass before mainnet.
Base Case and Portfolio Implication: What to Do Before Q4 2026
The base case is a staggered, mostly uneventful transition: large wallets, exchanges and RPC providers ship EIP-2780 and EIP-8037 support during the Platåberget window, while a tail of smaller custom send-flow tools lags and produces isolated user-facing failures around mainnet activation in Q4 2026 . Platåberget forked to Glamsterdam rules on August 20, 2026 and is designed to run for several months — longer than the short devnets before it — precisely so teams have a stable staging surface before Sepolia and Hoodi .
For traders, the practical move is due diligence rather than positioning. Treat Glamsterdam readiness as a checkmark on any wallet, bridge, custodian or exchange holding your ETH: has the provider publicly stated EIP-2780 and EIP-8037 support, and has it tested against the Platåberget faucet environment at faucet.plataberget.ethpandaops.io ? Silence from a venue you use to move ETH is the signal worth acting on, especially for flows that sweep a full balance to a fresh address.
For builders and DeFi teams, the audit list is narrow and concrete:
- Any hardcoded
21000used as a fallbackgasLimit, validation floor, or displayed "simple transfer" estimate . - "Send max" subtraction logic, where a transfer creating a new account costs 21,000 execution gas plus 183,600 state gas .
eth_estimateGaspaths that ignore the EIP-7976 calldata floor parameter and can underestimate gas .- Deterministic deployment factories with fixed gas limits, which fail on networks activating EIP-8037 .
- Airdrop distributors, payroll and onboarding contracts that programmatically pay fresh addresses.
The net thesis: this is a fee-mechanics and infrastructure-risk event, not a valuation catalyst on its own. Repricing does not change ETH issuance or demand directly; its investable content is whether it credibly de-risks the path from today's roughly 60M gas limit toward the 200M target core developers aligned on, alongside EIP-7732 and EIP-7928 . Concrete takeaway before Q4 2026: verify your custody stack's Glamsterdam statement, avoid full-balance sweeps to never-used addresses in the days around activation, and treat draft numbers as provisional until client test fixtures are final.
Frequently asked questions
What is EIP-2780?
EIP-2780, "Resource-based Intrinsic Transaction Gas," replaces Ethereum's flat 21,000-gas constant with named components that sum differently depending on what a transaction actually does: TX_BASE_COST of 12,000 gas for signature recovery, sender account access and block inclusion; COLD_ACCOUNT_ACCESS of 3,000 gas for touching the recipient; and TX_VALUE_COST of 6,000 gas for the recipient balance write and transfer logging . Its structural change matters more than the arithmetic: intrinsic costs, which are state-independent and determine transaction validity, are split from runtime costs, which depend on state and are charged only when incurred. Authors include Matt Garnett, Uri Klarman, Ben Adams, Maria Inês Silva, Anders Elowsson, Anthony Sassano and Dragan Rakita. The EIP is explicitly not backward compatible and is still in Review status, tracked under the gas-repricing meta-EIP EIP-8007.
Will my ETH transfers cost more after Glamsterdam?
It depends entirely on the recipient. A value transfer to an existing, distinct account still costs 21,000 gas — unchanged. An ETH self-transfer drops to 12,000 gas, and a zero-value transaction to an EOA or empty account costs 15,000. But a transfer that creates a new account costs 21,000 execution gas plus 183,600 state gas, and a transfer to an EIP-7702 delegated account costs 24,000 plus execution . The net direction is mixed rather than uniformly inflationary: routine sends between wallets you already use get cheaper or stay flat, while operations that grow Ethereum's state get materially more expensive. Contracts that primarily read or modify existing state are largely unaffected .
When does Glamsterdam launch on Ethereum mainnet?
Glamsterdam is targeted for Ethereum mainnet in Q4 2026, with no activation epoch confirmed yet . The Platåberget public testnet forked to Glamsterdam rules on August 20, 2026 and is intended to run for several months — deliberately longer than the short-lived devnets before it — as a stable surface for wallet, dApp and infrastructure teams, ahead of the Sepolia and Hoodi stages . Activation epochs for Sepolia, Holešky/Hoodi and mainnet remain TBD by client teams, and the meta-EIP EIP-7773 that lists the 19 scheduled EIPs is still in Draft status. Treat Q4 2026 as a target, not a commitment.
Do I need to update my wallet before Glamsterdam?
Yes, if it hardcodes 21,000 gas anywhere. Ethereum Foundation's Protocol DevOps team stated that "any tool that relies on a hardcapped maximum gas limit (think wallets, indexers and gas estimators) will break and needs to be updated" . The specific places to check are send-max and empty-wallet flows, fallback gasLimit values, and validation floors in gas estimators. Most retail traders do not patch their own wallet code — the practical step is to look for an official EIP-2780/EIP-8037 compatibility statement from your wallet provider, exchange or custodian before Q4 2026. The same requirement extends to RPC providers, bundlers, explorers and any dApp with a custom send flow .
Why does sending ETH to a new address cost so much more?
Because Glamsterdam starts charging for permanent state growth. EIP-8037 sets a cost-per-state-byte of 1,530 gas and models a new account record at 120 bytes — hash key, nonce, balance, code hash and storage root — producing 183,600 gas on top of the 21,000 execution cost, for 204,600 total . That is roughly a 7× increase over today's approximately 25,000 gas for account creation, and nearly 10× the flat 21,000 that wallets assume. The stated motivation is state growth: after the gas limit rose from 30M to 60M, average daily new state more than tripled, from about 105 MiB to about 326 MiB. The same model charges 97,920 gas per new storage slot and 35,190 per new EIP-7702 delegation indicator.
Enjoyed this article? Subscribe to get new stories by email whenever they're published.
