EIP-8007 Glamsterdam gas repricing collection (Glamsterdam Gas Repricings)
No call decision · Fees & gas · EIP status Draft · created 2025-08-21
A meta EIP that serves as a guide, collecting together the gas cost repricing proposals under discussion for the Glamsterdam fork. It functions only as a list that gives an overview of each proposal's content and status.
Why it is needed
On Ethereum, every operation is priced in a fee unit called gas. However, the current gas model has many mismatches between the price and the actual computing resources consumed. The source text gives as an example that even among pure computation operations, there is a large variance in execution time per unit of gas. This means some operations take much longer despite yielding the same gas value. This kind of price distortion causes certain operations to become bottlenecks, making it hard to raise the throughput of the whole network. The core goal of the Glamsterdam fork is improving L1 scalability, and the precondition for that is aligning gas costs with actual resource usage. This EIP gathers together the proposals belonging to that effort, organizing which proposals are under consideration for inclusion and which have been declined, so the community can track progress. The source text notes that this list is a first step and that further adjustments are expected to be needed in later hard forks.
What changes
Today
- Every transaction pays a single fixed base cost of 21,000 gas
- Actual execution time per unit of gas varies greatly from operation to operation
- The cost of creating new storage (state) is cheaper than its real burden
- Data attached to transactions (calldata) is cheap, which can let blocks grow excessively large
After EIP-8007
- The transaction base cost is broken down into items by actual resource used
- Creating new state incurs a separate fixed per-byte cost (CPSB 1,530) in its own gas dimension
- The cost of reading and writing state rises to match the larger state size
- Calldata and access list data costs rise, shrinking the maximum block size
Who is affected
- Users: The cost of a simple transfer stays at 21,000, but transactions with no value become cheaper, and transactions that create new accounts or storage become more expensive.
- Wallet & app developers: Gas estimation logic needs to be updated to match the rise in calldata and access list costs and the breakdown of transaction costs.
- Client teams: Proposals that change the gas accounting structure itself, such as a separate state gas dimension and excluding refunds from gas accounting, need to be implemented.
- L2s and rollups: As the calldata floor cost rises from 10/40 to 64/64 per byte, the cost structure for posting data to L1 changes.
In depth
The nature of this meta EIP and the three proposal types
This EIP is of type Meta, meaning it does not itself change the protocol. It states explicitly that it is purely informational and plays no active role in Glamsterdam governance. The EIPs in the list fall into three types. Broad harmonization proposals rebalance an entire category of operations to remove bottlenecks. Pricing extension proposals target a specific opcode or gas model element and usually come paired with a new mechanism. Supporting proposals do not perform repricing directly but underpin other repricing efforts or increase scalability potential.
7 proposals Considered for Inclusion
The proposals considered for inclusion are EIP-2780 (breaking the fixed 21,000 intrinsic cost down into per-resource primitive items), EIP-7778 (excluding refunds from block gas accounting to prevent bypassing the block gas limit), EIP-7904 (an empirical analysis of compute operations, concluding no change needed), EIP-7976 (raising the calldata floor cost to 64/64 per byte), EIP-7981 (charging 64 gas per byte for access lists), EIP-8037 (unifying and raising state creation costs based on CPSB and separating them into their own state gas dimension), and EIP-8038 (raising state access costs to reflect the larger state size). By resource category, this breaks down as multi-resource, general accounting, compute, data (2), and state (2).
Compute operations: conclusion is no change
EIP-7904 originally proposed raising the cost of some compute opcodes and precompiles, but it has now been converted to Informational, concluding that no repricing is needed. After rerunning benchmarks and the estimation pipeline on a client reflecting the parallel execution optimizations enabled by EIP-7928 (Block-Level Access Lists), every compute opcode and precompile reviewed met the 100 Mgas/s throughput target at its current gas cost. Therefore the entire compute category remains unchanged.
Changes to state access costs (EIP-8038)
The first-write surcharge STORAGE_WRITE rises from 2,800 to 10,000, COLD_ACCOUNT_ACCESS rises from 2,600 to 3,000, the account-leaf first-write surcharge ACCOUNT_WRITE rises from 6,700 to 9,000, and CREATE_ACCESS rises from 7,000 to 12,000. For access lists, ACCESS_LIST_STORAGE_KEY_COST rises from 1,900 to 2,000, and ACCESS_LIST_ADDRESS_COST rises from 2,400 to 2,900. EXTCODESIZE and EXTCODECOPY gain a new additional read cost equal to WARM_ACCESS, reflecting a second DB lookup. STORAGE_CLEAR_REFUND rises from 4,800 to 11,616. Meanwhile COLD_STORAGE_ACCESS (2,100) and WARM_ACCESS (100) remain unchanged.
State creation costs and a separate state gas dimension (EIP-8037)
This introduces a fixed cost per byte of state, CPSB = 1,530 gas. This value is derived from a 150M block gas limit baseline. The number of new state bytes is 120 per account, 64 per storage slot, and 23 per EIP-7702 authorization marker. Accordingly, GAS_CREATE, GAS_NEW_ACCOUNT, and PER_EMPTY_ACCOUNT_COST each become 120 x CPSB = 183,600 state gas, GAS_STORAGE_SET becomes 64 x CPSB = 97,920 state gas, PER_AUTH_BASE_COST becomes 23 x CPSB = 35,190 state gas, and GAS_CODE_DEPOSIT changes from 200 per byte to CPSB (1,530) per byte. State gas is metered in a dimension separate from execution gas, though some items are also accompanied by execution gas such as CREATE_ACCESS or ACCOUNT_WRITE.
Changes to transaction-level costs
EIP-2780 decomposes the 21,000 gas intrinsic cost of a simple transfer to an existing account into TX_BASE_COST (12,000) + COLD_ACCOUNT_ACCESS (3,000) + TX_VALUE_COST (6,000). Transactions with a zero value drop to 15,000 (12,000 if sending to self), and the intrinsic cost of a contract creation transaction drops from 53,000 to 24,000 execution gas. Instead, a transfer or creation that creates a new account pays 183,600 state gas at runtime. EIP-7702 authorization changes to a structure where runtime state-dependent costs are added to an intrinsic cost of 7,816. Also, EIP-7825's TX_MAX_GAS_LIMIT applies only to execution gas, and state gas is bounded only by tx.gas, so large contract deployments are no longer blocked by the per-transaction gas limit. The system call gas limit rises from 30,000,000 to 30,000,000 + STATE_BYTES_PER_STORAGE_SET x CPSB x SYSTEM_MAX_SSTORES_PER_CALL.
11 proposals declined for inclusion
The proposals declined for inclusion are EIP-2926 (code chunking in the MPT context), EIP-7686 (linear caps on memory and subcall gas limits), EIP-7923 (replacing quadratic memory costs with a page-based linear model), EIP-7971 (lowering TLOAD/TSTORE costs and a transaction-wide limit), EIP-7973 (introducing warm account writes), EIP-8011 (per-EVM-resource gas accounting), EIP-8032 (making SSTORE cost proportional to contract storage size), EIP-8053 (introducing an internal accounting unit of milli-gas), EIP-8057 (a temporal-locality discount spanning multiple blocks), EIP-8058 (a discount for duplicate bytecode deployment), and EIP-8059 (scaling up gas parameters via a REBASE_FACTOR multiplier). Many of the memory and state alternatives were left out this time.
Status is Draft and the figures are unconfirmed
This EIP is in Draft status, so its contents may still change. The source itself states that the gas cost figures in the Rationale are preliminary values, not yet finalized, and subject to change. The source also notes that this list is only meant to be a meaningful first step in aligning gas costs with actual resource usage, and that further iterations will likely be needed in future hard forks. The provided call records contain no fork-inclusion decision for this EIP.