.png){kind=logo}
Every Ethereum transaction pays a fee, but very few users can say where the number on their wallet screen actually comes from. The answer is not an auction run by miners and it has not been since August 2021. It is a formula written into the protocol itself, with three moving parts: a base fee the network sets algorithmically and destroys, a priority fee you choose to pay the validator, and, for rollups posting data, a third fee market for blobs that runs on its own exponential curve. This article walks through each mechanism as specified in EIP-1559 and EIP-4844, the two protocol documents that define them, and shows the exact arithmetic. The mechanism itself has been stable for years and is likely to stay recognisable for years more, even as the parameters around it keep moving. Nothing here constitutes investment or financial advice, it is a technical reference on how the fee calculation works.
What actually determines the fee you pay?
The fee is the amount of gas your transaction consumes multiplied by the sum of two per-unit prices, the base fee and the priority fee. Gas is the unit of computational work. A plain ETH transfer always consumes 21,000 gas, while contract interactions consume more depending on what the code does. The per-unit price is quoted in gwei, which is one billionth of an ETH.
The formula, as documented by ethereum.org's developer documentation, is: total fee equals units of gas used multiplied by the sum of base fee and priority fee. Their worked example: a 21,000 gas transfer at a 10 gwei base fee with a 2 gwei tip costs 21,000 times 12, which is 252,000 gwei or 0.000252 ETH.
| Component | Who sets it | Where it goes | Defined in |
|---|---|---|---|
| Base fee per gas | The protocol, per block, by formula | Burned, permanently destroyed | EIP-1559 |
| Priority fee per gas | The sender, as a voluntary tip | The validator who includes the block | EIP-1559 |
| Blob base fee (rollup data only) | The protocol, per block, exponential formula | Burned, even if the transaction fails | EIP-4844 |
When you sign a transaction you do not set the base fee. You set two ceilings instead. The first is max fee per gas, the absolute maximum per-unit price you are willing to pay in total. The second is max priority fee per gas, the maximum tip. The protocol charges you the current base fee plus whichever is smaller, your stated tip or the headroom left between the base fee and your overall ceiling. Anything unused under your max fee is simply not charged. This is why the fee your wallet estimates and the fee you actually pay are often different: the estimate is a ceiling, the charge is a formula evaluated at inclusion time.
How is the base fee calculated?
The base fee moves block by block according to how full the previous block was, and the maximum move in either direction is one eighth, or 12.5 percent. This mechanism arrived with EIP-1559 in the London upgrade, which went live on 5 August 2021 at block 12,965,000, replacing the old first-price auction where users blindly bid against each other.
The specification defines two constants that drive everything. The elasticity multiplier is 2, meaning each block has a gas target equal to half its gas limit. The base fee max change denominator is 8. When a block lands exactly on target, the base fee does not move. When a block is completely full, the next base fee rises by the full 12.5 percent. When a block is empty, it falls by 12.5 percent. Partial deviations scale proportionally: a block 50 percent over target pushes the base fee up by 6.25 percent.
Two consequences follow directly from the arithmetic. First, sustained congestion compounds fast. At the maximum 12.5 percent rise per block, the base fee roughly doubles every six blocks, which is about 72 seconds at Ethereum's 12 second block time. This is why fees spike so sharply during NFT mints or liquidation cascades. Second, the burn removes the incentive problem of the old auction. Because validators never receive the base fee, they cannot profit by stuffing blocks to inflate it. The EIP-1559 text is explicit that the base fee is always burned, destroyed by the protocol.
The practical takeaway is that the base fee is not negotiable and not manipulable by any single actor, it is a deterministic function of recent block fullness. Anyone quoting you a way to pay below the current base fee is describing something other than an Ethereum layer 1 transaction.
What are the current block size parameters?
As of 2026 the mainnet gas limit sits at 60 million, giving a target block of 30 million gas. According to ethereum.org, the validator community raised the limit from 30 million to 60 million during 2025, and the Fusaka upgrade's EIP-7935 standardised the higher figure as the client default. Reporting by The Defiant on the Glamsterdam devnets indicates core developers are targeting a further rise towards 200 million in the fork planned for the second half of 2026. That last figure is a plan, not a shipped change, and plans on Ethereum's roadmap have slipped before.
What does the priority fee do?
The priority fee is the only part of the fee that is actually a market, and it buys ordering, not inclusion in the abstract. Validators choose which pending transactions to include and in what order, and since they keep the tip, they naturally prefer higher tips. Under normal load a small tip suffices because blocks have spare capacity. During congestion, when demand exceeds even the elastic 2x block size, the tip becomes the tiebreaker among transactions all willing to pay the base fee.
ethereum.org describes the tip plainly as an incentive for validators to select your transaction for the next block. There is no protocol-defined minimum. A transaction with a zero tip is valid, it just relies on a validator including it out of spare capacity.
How do blob fees work?
Blob fees are a second, fully independent fee market introduced by EIP-4844 in the Dencun upgrade of March 2024, and they are priced by an exponential function rather than the base fee's linear percentage steps. Blobs are large data packets, 128 kilobytes each, that rollups such as Arbitrum, Base and Optimism use to post their transaction data to Ethereum cheaply. Blob data is not accessible to smart contracts and is pruned after a few weeks, which is why it can be priced far below regular calldata.
The EIP-4844 specification defines each blob as consuming 2 to the power 17, that is 131,072, units of blob gas. The protocol tracks a running counter called excess blob gas: each block adds the blob gas it used and subtracts the target, with the counter floored at zero. The blob base fee is then computed as a minimum price of 1 wei multiplied by an exponential of that excess, using the spec's fake_exponential function. The update fraction in the original specification, 3,338,477, was chosen so that a sequence of maximally full blob blocks raises the price by a factor of about 1.125 per block, deliberately mirroring EIP-1559's 12.5 percent cadence.
The exponential form matters at the low end. When blob demand sits below target for a long stretch, the excess counter drains to zero and the blob base fee falls all the way to its 1 wei floor, effectively free. When demand crosses the target, the price climbs from that floor exponentially. This is why rollup posting costs can sit near zero for months and then jump several orders of magnitude within hours, the curve is exponential by design. Two further spec details are worth knowing: the blob fee is deducted from the sender's balance before execution and burned even if the transaction fails, and there are no refunds on it. Fusaka's EIP-7918 later added a lower bound tying the blob base fee to execution costs, so the near-free regime is now bounded rather than open ended.
How many blobs fit in a block today?
As of January 2026 the network targets 14 blobs per block with a maximum of 21. The capacity has been raised in steps, each step changing the target and maximum while leaving the pricing mechanism untouched.
| Change | Date | Target blobs | Max blobs | Source |
|---|---|---|---|---|
| Dencun (EIP-4844 launch) | March 2024 | 3 | 6 | EIP-4844 specification |
| Pectra (EIP-7691) | May 2025 | 6 | 9 | ethereum.org |
| Fusaka activation | 3 December 2025 | 6 | 9 | Ethereum Foundation blog |
| BPO1 fork | 9 December 2025 | 10 | 15 | Ethereum Foundation blog |
| BPO2 fork | 7 January 2026 | 14 | 21 | Ethereum Foundation blog, The Block |
The Fusaka upgrade activated on mainnet on 3 December 2025 at 21:49:11 UTC, at slot 13,164,544, per the Ethereum Foundation's announcement. Its headline change, PeerDAS, lets validators verify blob availability by sampling rather than downloading every blob in full, which is what makes the higher blob counts safe to run. Further blob parameter forks towards much higher counts have been discussed by core developers but are not activated, so treat any figure beyond 14 target and 21 max as speculative.
What has this meant for real fee levels?
Fee levels in 2026 are low by any historical standard, with ethereum.org reporting standard gas around 0.15 gwei and daily averages near 0.5 gwei through April 2026. At those prices a simple transfer costs a fraction of a cent at prevailing ETH prices. Note the condition attached to that figure: it is a network-wide average reported by the Ethereum project's own site, not an independent measurement, and gas prices are volatile enough that any snapshot ages quickly. The structural reasons are the ones described above, a doubled gas limit, blob capacity up more than fourfold since launch, and most transaction demand migrating to rollups whose costs ride the cheap blob market.
Low fees are not guaranteed to persist. The same formulas that produced sub-gwei base fees will reprice within minutes if demand returns faster than capacity grows. The mechanism is stable, the price level is not, and readers should not extrapolate either direction. Fee dynamics also feed into adjacent debates: institutional custody of assets that live on these networks, covered in our piece on the FDIC's approval for banks to hold bitcoin, ultimately depends on operational costs like these, and the capital treatment questions raised in the Basel III crypto capital requirements fight assume banks can transact on-chain at predictable cost.
Where does this model break down or mislead?
The formula is exact, but three things still make fees feel unpredictable, and it is worth being honest about each. First, gas consumption is not always knowable in advance. A swap's gas cost depends on the contract path executed at inclusion time, so wallets simulate and pad the estimate. You are refunded for unused gas, but a transaction that runs out of gas fails while still paying for everything it consumed. Second, base fee movement between signing and inclusion means the effective price can rise up to 12.5 percent per block while you wait. Wallets compensate by setting max fee well above the current base fee, which is safe, since the excess is never charged, but it makes the displayed maximum look scarier than the likely cost. Third, ordering during contention is an opaque market. Tips interact with block building infrastructure and private order flow in ways the EIPs do not specify, and no formula predicts the winning tip in a gas war.
There is also an honest limitation in this article's own sourcing. Protocol mechanics have no government source and little peer-reviewed literature, the canonical documents are the EIP specifications and Ethereum Foundation publications, which are project documentation rather than independent primary sources in the usual hierarchy. For the mechanism itself that is acceptable, the spec is the ground truth by definition, code implements the spec. For claims about fee levels and future upgrades it is weaker, and the table below grades each claim accordingly.
| Claim | Type of evidence | Strength |
|---|---|---|
| Base fee formula, 12.5 percent max step, elasticity of 2, base fee burned | EIP-1559 final specification, the normative protocol document | Strong, this is the definition itself |
| London activation on 5 August 2021 at block 12,965,000 | Contemporary reporting (CoinDesk) of an on-chain event | Strong, independently checkable on any block explorer |
| Blob gas constants, exponential pricing, blob fee burned without refund | EIP-4844 final specification | Strong for the original constants, note that per-fork parameters have since changed |
| Blob capacity schedule, Fusaka and BPO dates | Ethereum Foundation blog, corroborated by The Block and CoinDesk | Strong for dates, moderate overall, the EF is the project's own publisher |
| Gas limit raised to 60 million during 2025 | ethereum.org project documentation | Moderate, project documentation, not an independent source |
| Standard gas around 0.15 gwei, daily averages near 0.5 gwei through April 2026 | ethereum.org, a single non-independent snapshot with unstated methodology | Weak, treat as indicative only |
| Glamsterdam raising the gas limit towards 200 million in H2 2026 | Trade press reporting on devnet plans (The Defiant) | Weak, unshipped roadmap, dates and figures may change |
FAQ
Why was my transaction fee lower than my wallet's estimate?
Because the estimate is a ceiling, not a quote. Your wallet sets a max fee with headroom for the base fee to rise before inclusion. You are charged the actual base fee at inclusion plus your tip, and any unused headroom under your max fee is never taken.
Does the priority fee guarantee my transaction goes through?
No. It raises your priority in validator ordering because validators keep the tip. Under normal load even a minimal tip gets included promptly. In extreme congestion, transactions whose max fee falls below the rising base fee wait in the mempool regardless of tip.
Who receives the gas fees I pay?
The base fee is destroyed by the protocol, no one receives it. Only the priority fee goes to the validator who includes your transaction. Blob fees paid by rollups are likewise burned. This burn design, specified in EIP-1559, exists to stop block producers profiting from inflating the fee they control.
Do blob fees affect the fee I pay for a normal transaction?
Not directly. Blob gas is an independent market with its own base fee, and only blob-carrying transactions, in practice rollup batchers, pay it. Indirectly it matters a great deal, since cheap blobs keep rollup fees low and pull demand away from layer 1, which eases the regular base fee. Stablecoin transfers, one of the biggest sources of on-chain volume discussed in our coverage of US Treasury stablecoin guidelines, increasingly settle on rollups for exactly this reason.
Can gas fees go to zero?
The execution base fee cannot be negative and can drift very low under sustained low demand, but every transaction still pays for its gas at whatever the base fee is, and validators still need tips to prioritise. The blob base fee has an explicit floor of 1 wei in the specification, and since Fusaka's EIP-7918 it is additionally bounded by execution costs, so truly free blob space no longer occurs.
Will this article's numbers still be right next year?
The mechanisms almost certainly, the parameters probably not. The formulas of EIP-1559 and EIP-4844 have survived every upgrade since their introduction. The gas limit, blob targets and typical fee levels have all changed within the last twelve months and are scheduled to change again. Projects whose economics lean on cheap block space, from the ESG crypto fund niche to the remnants of the fractional NFT experiment, should re-check current parameters rather than cite this page's snapshot.
Sources
Protocol mechanics have no state or peer-reviewed source of record. The EIP specifications are the canonical normative documents and are treated here as ground truth for the mechanism. Ethereum Foundation and ethereum.org material is project documentation, reliable for protocol facts, weaker for network statistics, and graded accordingly in the evidence table above.
- EIP-1559: Fee market change for ETH 1.0 chain, final specification
- EIP-4844: Shard Blob Transactions, final specification
- Ethereum Foundation: Fusaka Mainnet Announcement, 6 November 2025
- ethereum.org developer documentation: Gas and fees
- ethereum.org: Building on Ethereum in 2026
- CoinDesk: Ethereum's London Hard Fork Is Now Live, 5 August 2021
- The Block: Ethereum rolls out final planned Blob Parameters Only fork
- The Defiant: Glamsterdam Upgrade Enters Final Devnet Phase With 200M Gas-Limit Target
This article is a technical reference on protocol mechanics. It does not constitute investment or financial advice, and nothing in it should be read as a recommendation to buy, sell or hold any asset.
