A transaction on Base or on Arbitrum is priced as two separate charges, not one. There is a fee for running the transaction on the layer-2 network itself, and a fee for getting a record of that transaction posted back to Ethereum. Per each chain’s own documentation, those two charges are calculated by different formulas, and on Base, the Ethereum-posting charge is typically the larger of the two.
The two-part structure
Per Arbitrum’s documentation, a user submitting a transaction pays two parties. The ‘poster’ is reimbursed for the cost of the parent chain (Ethereum) resources needed to post the transaction, mainly calldata. The ‘network fee account’ covers the child chain’s own resources: computation, storage, and the cost of Arbitrum’s chain-specific precompiles, referred to in the docs as ArbOS.
Per Base’s documentation, the split is described in similar terms but different language: every Base transaction has an L2 execution fee, which is the cost of running the transaction on Base, and an L1 security fee, the estimated cost of publishing that transaction back to Ethereum. Base’s documentation states plainly that the L1 fee is typically higher than the L2 fee.
How Arbitrum prices the Ethereum-posting piece
Arbitrum’s documentation describes an estimation problem: the exact cost of posting a batch of transactions to Ethereum is not known until the batch is actually posted, but the fee has to be set at the moment a transaction is sequenced, before that. Arbitrum’s answer, per its documentation, is to compress each transaction on its own using the Brotli algorithm, at its fastest and least aggressive setting, then multiply the resulting compressed size by 16 – the same per-non-zero-byte gas charge Ethereum itself uses. That calculation produces an estimate of the transaction’s ‘data units.’ The estimate is then priced against a rate, expressed in wei per data unit, that adjusts over time so that the total collected from users tracks what the batch poster actually pays Ethereum. If the batch poster’s real costs run ahead of what fees have collected, Arbitrum’s documentation says the per-unit price rises; if costs run behind collections, it falls.
The result reported to the user is called the ‘poster fee.’ Arbitrum’s documentation is explicit that this is an approximation – it does not know in advance how well a given transaction will compress once combined with others in the same batch, so it estimates each transaction’s contribution individually and adjusts the market-wide rate over time rather than pricing any single transaction with certainty.
How Base prices the Ethereum-posting piece
Base’s documentation points to a smart contract, the GasPriceOracle, deployed at a fixed address on Base, which lets a wallet or app query the current L1 fee components before a transaction is even signed: the current Ethereum L1 base fee as Base sees it, the current EIP-4844 blob base fee, and the scalars applied to each. A function called getL1Fee returns an exact figure for a fully built transaction; getL1FeeUpperBound gives a faster approximate figure from an estimated byte length.
Base also documents a floor under its own execution fee. As part of what Base’s documentation calls the Jovian upgrade, Base set a minimum base fee of 5,000,000 wei, or 0.005 gwei, on Base Mainnet, so the L2 fee cannot fall to near-zero during quiet periods. Base’s documentation works through what that floor means in dollar terms: at 0.005 gwei and an assumed ETH price of $2,000, a typical 200,000-gas transaction costs approximately $0.002. That is a stated illustrative calculation from Base’s own docs, not a running average of what users have actually paid on any particular day.
Base’s fee-adjustment mechanism, per its documentation, is its own implementation of EIP-1559 with two set parameters: an elasticity multiplier of 6, meaning a block can absorb up to six times its target gas before the fee reacts sharply, and a base fee change denominator of 125, which controls how gradually the fee moves. Together, per Base’s documentation, those parameters cap the maximum base fee increase at 4% per block, and Base’s own arithmetic in that documentation puts the minimum time for the base fee to double at 18 blocks, or 36 seconds, given a roughly two-second block time.
What actually moves the number
Because the largest single component of a Base transaction fee, per Base’s own documentation, is normally the L1 fee, and because that fee tracks Ethereum’s base fee and blob fee in real time via the GasPriceOracle, a spike in the price of a Base transaction usually reflects congestion on Ethereum layer 1, not a change in how busy Base itself is. Base’s documentation makes the same point directly: users can lower costs by timing transactions for periods of lower L1 gas, giving weekends as an example.
Arbitrum’s fee, per its documentation, moves for a related reason but a different mechanical path. The poster fee tracks the price per data unit that ArbOS is currently charging, which itself adjusts based on the gap between what batch posters have actually paid Ethereum and what fees have been collected from users. A rise in Ethereum’s own base fee raises the cost of posting a batch, which – with a lag, since batch posting reports travel through a delayed inbox before ArbOS processes them – feeds through into a higher per-data-unit price on Arbitrum.
The common misreading
It is tempting to read a Base or Arbitrum wallet quote as one number reflecting how busy that layer-2 is. Per both chains’ documentation, that is usually the wrong read. The larger, more volatile piece of the fee on both networks is tied to conditions on Ethereum’s base layer, not to traffic on the layer-2 itself. A quiet day on Arbitrum or Base can still produce an expensive transaction if Ethereum’s base fee or blob fee has spiked at the same time.
About the ‘$15 to $0.0001’ claim
A figure that circulates widely – that Arbitrum reduces Ethereum gas fees from $15 to $0.0001 – appears in this evidence pack only in a Gate Wiki crypto-encyclopedia article published 15 January 2026, which describes itself as covering Arbitrum’s whitepaper, roadmap and team. It does not appear in Arbitrum’s own gas-and-fees documentation reviewed for this page. The same wiki article, single-sourced in this evidence pack, also carries claims about Arbitrum capturing more than 60% of layer-2 total value locked, a treasury it describes as “$2+ billion,” and leadership described as Princeton academics and former White House officials. None of those figures are corroborated by Arbitrum’s own technical documentation or by any other source reviewed here, and this page treats them as unverified marketing claims rather than facts about how fees are calculated.
What this page does not tell you
This page does not tell you which chain is cheaper right now. Neither Arbitrum’s documentation nor Base’s documentation publishes a synchronized dollar figure that lets a reader compare the two chains at the same instant, and both chains’ Ethereum-posting fees move block to block with Ethereum’s own base fee and blob fee. Base’s $0.002 worked example is tied to a stated assumption of $2,000 per ETH and to Base’s minimum base fee floor; it is an illustrative calculation from Base’s documentation, not a historical average of what users paid on any given date. Arbitrum’s documentation does not provide an equivalent dollar worked example at all, so no side-by-side arithmetic is possible from the evidence available here. Both sets of technical documentation are undated live reference pages that the respective teams can and do update, so figures such as Base’s 5,000,000 wei minimum base fee or its 4% maximum base-fee change per block reflect the configuration described in the version of the docs reviewed for this page, not necessarily a permanent setting – Base’s own engineering blog lists ‘op-code repricing’ and unspecified ‘economic changes’ to its base fee mechanics as planned features of a future hard fork called Base V3. Finally, the frequently cited claim that Arbitrum reduced gas costs from $15 to $0.0001 comes from a single third-party wiki source rather than from Arbitrum’s own documentation, and this page does not treat it as verified.
Sources
Every fact above is attributed to one of these reports. Where they disagree, the article says so.
