EIP-2780: Reduce intrinsic cost of transactions

Here is what you need to know before reading the full breakdown:

• Flat intrinsic gas: Every Ethereum transaction currently pays 21,000 gas before any EVM code runs. This constant dates from Ethereum's 2014 Yellow Paper and has never been updated.

• The problem: A zero-value call and a value transfer that creates a brand-new account both pay 21,000 gas, even though the latter permanently expands chain state.

• EIP-2780's fix: Decompose 21,000 into four named, benchmark-derived primitives: TX_BASE_COST (12,000), COLD_ACCOUNT_ACCESS (3,000), TX_VALUE_COST (4,244), and TRANSFER_LOG_COST (1,756). For an existing-account ETH transfer these sum to exactly 21,000, same as today.

• What actually gets cheaper: Zero-value calls drop to 15,000 gas. Self-transfers drop to 12,000 gas. Account-creating transfers become more expensive via a separate state-gas charge of ~183,600.

• The "71% cheaper" headline is wrong for the current version: That figure came from an intermediate draft where TX_BASE_COST fell to 4,500. The current spec keeps existing-account transfers at 21,000.

• Status (June 2026): Draft, Considered for Inclusion (CFI) in Glamsterdam. Not Scheduled for Inclusion (SFI). Inclusion depends on testing progress and the headliner EIPs (ePBS, BALs) landing on schedule.

Every Ethereum transaction has two cost layers. The intrinsic cost is deducted upfront, before any EVM code executes, and is non-refundable even if the transaction reverts. The execution cost covers the actual opcode-by-opcode work of a smart contract.

The intrinsic cost today is:

g_0 = 21,000 (base)

    + calldata cost
    + 32,000 (if contract creation)
    + access-list costs (EIP-2930)
    + EIP-7702 authorization costs

The 21,000 base was assigned in the original Yellow Paper as G_transaction, intended to cover ECDSA signature recovery, reading the sender's nonce and balance, and touching the recipient. It is a single constant applied identically to every transaction regardless of what that transaction does to chain state.

The 21,000 constant charges every transaction the same regardless of the work performed. A zero-value call to an existing contract and a value transfer that creates a brand-new account both pay 21,000 gas. The first transaction is overpriced. The second is underpriced: it writes a new state trie leaf, emits a transfer log under EIP-7708, and permanently grows the chain's storage footprint.

This also creates an asymmetry between top-level transfers and internal ones. Creating an account via a smart contract's CALL already charges for the new state it creates. Sending ETH directly to a fresh address does not. EIP-2780 closes that gap by internalizing state-creation costs at the intrinsic level, and ties each primitive to empirical client benchmarks rather than a 2014 guess.

EIP-2780 has gone through four distinct versions since 2020. Most secondary sources describe an earlier draft, not the current one. 

Version 3 is the one that generated the "up to 71% cheaper" headline because a plain ETH transfer would have cost 6,000 gas. The current canonical Version 4 keeps the existing-account ETH transfer total at 21,000 because the benchmark-fitted primitives happen to sum to that figure. The spec is now called "Resource-based intrinsic transaction gas."

The spec decomposes 21,000 into four named parameters, each derived from NNLS regression of measured client runtimes across all five major execution clients (Besu, Erigon, Geth, Nethermind, Reth), anchored at 100,000 gas per millisecond of CPU time: 

TX_BASE_COST (12,000) + COLD_ACCOUNT_ACCESS (3,000) + TX_VALUE_COST (4,244) + TRANSFER_LOG_COST (1,756) = 21,000. Same headline, different anatomy. 

The spec also inherits a two-dimensional gas model from EIP-8037: regular gas for computation, and state gas for permanent storage growth. Account-creating transfers pay 21,000 regular gas plus ~183,600 state gas from a separate reservoir. 

Reference costs by transaction type:

EIP-2780 imports COLD_ACCOUNT_ACCESS from EIP-8038 and the state-byte cost from EIP-8037, so future repricings of those parameters automatically flow through to intrinsic gas. The spec also has explicit rules for EIP-7702 delegated accounts: a value transfer to a 7702-delegated address costs 24,000 regular gas.

Ordinary users: Plain ETH transfers between existing accounts are unchanged at 21,000 gas. Zero-value calls get cheaper (15,000). Sending to a fresh address gets more expensive because of the state-gas charge. If you regularly send to new counterparties or run airdrops, expect higher costs.

Wallet developers and gas estimators: This is the most operationally significant change. Any tool that hard-codes a flat 21,000 intrinsic base must be updated. More importantly, the two-dimensional gas model means a transaction's state-gas cost can dwarf its regular-gas cost: ~183,600 state gas versus 21,000 regular gas for a new-account transfer. Begin modeling this now. The trigger to act is EIP-2780 moving from CFI to SFI on EIP-7773.

L2s and rollups: The direct effect is modest since the dominant levers remain blob and calldata pricing. The indirect benefit is that cheaper zero-value transactions reduce fixed overhead on batched operations, and accurate account-creation pricing prevents mispriced state-growth transactions from being undercharged in L2 batches. For Rango, we are tracking EIP-7773 closely and will update our gas estimation engine ahead of any Glamsterdam devnet activation. You can swap and bridge across 80+ chains today on Rango Exchange.

EIP-2780 was moved to Considered for Inclusion at ACDE #226 on December 18, 2025. As of June 2026, it is CFI but not SFI. The Reth team rates it "Tier 1: High Priority." Erigon, Reth, and Besu developers have publicly supported evaluating it, and the EF research team backs the full repricing bundle. The key technical gating condition is all five major execution clients meeting the 100 Mgas/s benchmark on the heaviest transfer path (unique 24-KiB-code receiver).

The main open risks are: spec ambiguity on edge cases flagged by Jochem Brouwer on Etheruem Magicians; the 2300-stipend interaction in corner cases; and a process concern that the broader repricing bundle is "too monolithic." Watch the bi-weekly Glamsterdam Repricings call series and the EIP-7773 hardfork meta for updates.

EIP-2780 replaces Ethereum's 2014 flat intrinsic gas constant with four empirically grounded primitives and internalizes the real cost of state creation. The headline number for a standard ETH transfer stays at 21,000, but the components behind it finally have a measurable justification. For developers and infrastructure teams, the practical takeaway is straightforward: stop assuming a flat 21,000 base, start modeling two gas dimensions, and watch for EIP-2780's graduation from CFI to SFI.

• Ethereum/EIPs: EIP-2780 specification

• Ethereum Magicians: EIP-2780 discussion thread

• Ethereum/pm: Glamsterdam Repricings call notes

• Reth team: Glamsterdam priorities list

• Christine D. Kim: ACDE #226 notes