Quick answer: Ethereum gas fees are the price you pay for computational work. Every operation, from a simple transfer to a complex smart contract call, consumes “gas,” and you pay for it in ETH based on current network demand. Fees spike when lots of people compete for limited block space and fall when the network is quiet. Since the London upgrade (EIP-1559), each transaction fee splits into a base fee, which is burned, and an optional priority tip that goes to the validator.
Few things frustrate new crypto users faster than opening a wallet, trying to send a small amount of ETH, and finding that the fee costs more than the transfer. It feels arbitrary from the outside. It isn’t. There’s a clear, if unforgiving, economic logic behind it, and in the last couple of years several upgrades have changed which fees you actually need to worry about.
What “Gas” Actually Measures
Every action on Ethereum (sending tokens, swapping on a DEX, minting an NFT, calling a smart contract) needs computational work from the network. Gas is the unit that measures how much work a given operation requires. According to ethereum.org, a standard ETH transfer requires a gas limit of 21,000 units. A complex DeFi transaction that touches several contracts can use many times that.
The fee you pay is gas used multiplied by the price per unit of gas. That price is quoted in gwei, a billionth of one ETH. So the cost of a given transaction depends on two things you don’t control much (how busy the network is, and what ETH is worth when you convert the fee into dollars) and one thing you partly do (how much computation your action needs).
Why Fees Spike: Supply and Demand for Block Space
Ethereum produces a block roughly every 12 seconds, and each block can hold only so much gas. That’s a hard capacity limit. When more people want to transact than a block can fit, they effectively bid against each other for inclusion, and fees rise.
This is why gas jumps during popular NFT mints, token launches, or sharp market moves when everyone wants to trade at once. It’s a plain supply-and-demand story, just applied to computation instead of a physical good.
EIP-1559: The Base Fee, the Tip and the Burn
Before the London upgrade in 2021, Ethereum used a first-price auction. Everyone guessed a gas price, the highest bidders went first, and plenty of people overpaid just to be safe. EIP-1559 replaced that with a structure that has three moving parts. Here are the mechanics as ethereum.org’s gas documentation describes them (page last updated June 24, 2026):
- Base fee. Set by the protocol, not by users, and calculated from the previous block. It must be paid for a transaction to be valid, and it is burned, meaning permanently removed from circulation.
- Target and adjustment. The target block size is half the block gas limit. The base fee rises or falls by at most 12.5% per block depending on whether the previous block was above or below target. A completely full block pushes it up 12.5%; an empty block pulls it down 12.5%.
- Priority fee (tip). Paid to the validator as an incentive to include your transaction. It rises when demand is high and many people want to jump the queue.
- Max fee. An optional cap you set. It must exceed the base fee plus the tip, and anything above what’s actually charged is refunded to you.
A Worked Example in Gwei
Here’s the arithmetic, using ethereum.org’s own numbers. Say the base fee is 10 gwei and you add a 2 gwei tip to send a simple ETH transfer of 21,000 gas.
The total is 21,000 x (10 + 2) = 252,000 gwei, which is 0.000252 ETH. Of that, 210,000 gwei (the base fee portion) is burned and 42,000 gwei goes to the validator.
Now suppose your wallet set a max fee of 30 gwei per gas to be safe. You are still charged only for what was needed. The difference between 30 gwei and 12 gwei, across 21,000 units, is 378,000 gwei, which comes back to you. That’s why a high max fee mostly protects you against a sudden base fee jump rather than costing you more.
How fast can the base fee climb? If blocks stay completely full, ethereum.org’s example shows it rising from 100 gwei to about 202.7 gwei in eight blocks, because each full block adds 12.5% to the last. That compounding is what turns a busy five minutes into a painful fee spike.
Why Layer-2s Exist, and What Dencun Changed
Layer-2 rollups such as Arbitrum, Optimism and Base exist largely because mainnet gas is expensive and volatile. They process transactions off the main chain and post compressed data back to Ethereum, spreading that posting cost across many users. We compare the big three in our layer-2 comparison.
The Dencun upgrade activated on 13 March 2024 and introduced EIP-4844, sometimes called proto-danksharding. It created a new type of data attachment called a blob, which rollups use to post their data more cheaply. Blobs are kept for roughly 18 days and then pruned. ethereum.org says blob storage significantly reduces rollup costs. The main effect is on layer-2 fees, not on ordinary mainnet transactions, which still compete for regular block space.
Pectra and Fusaka: What Changed Since Then
Two later upgrades kept building on this, and they’re worth knowing because older articles (including earlier versions of this one) can leave a wrong impression.
Pectra (7 May 2025). Per ethereum.org, EIP-7691 raised the blob target from 3 to 6 per block and the maximum from 6 to 9, adding capacity for rollup data. Pectra also changed calldata pricing (EIP-7623) so that data-heavy transactions pay more, which nudges layer-2s toward blobs, while leaving over 99% of transactions unaffected.
Fusaka (3 December 2025). The headline feature, PeerDAS, lets each node hold only a portion of blob data instead of all of it, which allows far more layer-2 data to be supported. Fusaka also introduced “blob-parameter-only” (BPO) forks, which raise blob counts between major upgrades. EIP-7918 sets a reserve price under blobs so the blob fee can’t collapse to the minimum when execution gas is expensive. The default block gas limit rose to 60 million from 45 million, and a single transaction is now capped at 2^24 (16,777,216) gas.
The key point for everyday users: ethereum.org is explicit that Fusaka doesn’t directly lower layer-1 gas fees. Its main effect is more blob space, which is meant to make layer-2 fees cheaper. A bigger gas limit can add capacity on mainnet, but whether fees actually fall depends on how much demand shows up to fill it. Looking ahead, our coverage of Glamsterdam’s testnet status and the 200M gas limit discussion explains why that proposal isn’t live on mainnet.
Blob fees also run on their own market, separate from regular gas. When you use a layer-2, the fee you see is partly the layer-2’s own execution cost and partly the cost of posting data to Ethereum, which today usually goes through blobs.
Practical Ways to Manage Gas Costs
- Watch timing. Activity follows daily and weekly rhythms tied to when major markets are active. Transacting during quieter hours, such as weekends, can mean meaningfully lower fees, though there’s no guarantee.
- Use a layer-2 for everyday activity. For anything that doesn’t need mainnet settlement right away, a layer-2 is usually the biggest cost lever. Just remember that bridging between chains has its own risks, which we cover in our bridge explainer.
- Batch actions where the app supports it. Bundling several actions into one transaction can save the fixed overhead you’d otherwise pay repeatedly. Smart accounts make this easier, as we explain in our account abstraction guide.
- Check the estimate before you confirm. Good wallets show base fee and tip separately. If the tip looks absurd for a non-urgent transaction, lower it.
- Beware of bots during busy moments. On-chain trades can be reordered by searchers chasing profit, which we cover in our MEV guide.
A Practical Scenario
Imagine you’re trying to claim an NFT during a hyped mint. Thousands of people are transacting in the same few minutes. Blocks fill up, the base fee ratchets higher with each full block, and everyone attaches aggressive tips to get in before the mint sells out. You could pay many times a normal fee, not because Ethereum “changed its pricing” but because you bid against a crowd for the same scarce space at the same moment.
If you can wait a few hours, fees often normalise once the rush fades. And if the project also lives on a layer-2, minting there might cost a fraction as much, though you’d want to check how that specific network prices its fees.
Common Mistakes
- Confusing the tip with the whole fee. The base fee is usually the bigger part, and you can’t haggle with it.
- Setting a huge max fee on an unfamiliar app, then discovering the transaction is a poorly built contract call that burns gas anyway.
- Assuming a failed transaction costs nothing. The computation up to the failure still uses gas.
- Believing every upgrade cuts mainnet fees. Dencun, Pectra and Fusaka mainly expanded capacity for layer-2 data.
- Ignoring gas on small transfers. Moving a tiny amount on mainnet during a busy period can cost more than it’s worth.
Key Takeaways
- Gas measures computation. A simple ETH transfer uses 21,000 units; complex contract calls use far more.
- Under EIP-1559, the base fee is set by the protocol, moves by at most 12.5% per block, and is burned. The tip goes to the validator.
- Dencun (13 March 2024), Pectra (7 May 2025) and Fusaka (3 December 2025) increased room for layer-2 data through blobs, which mainly lowers layer-2 fees.
- Fusaka doesn’t directly lower layer-1 gas fees, according to ethereum.org.
- Timing, layer-2s and sensible wallet settings are the most practical levers.
FAQ
Why did my Ethereum transaction fail even though I paid gas?
Failed transactions still consume gas for the computation done before the failure. A swap that hit its slippage limit is a common example, which is why sensible slippage and gas limits matter.
What is gwei?
Gwei is one billionth of an ETH, used to quote gas prices in a readable unit.
Does burning ETH reduce total supply?
The burn removes ETH permanently, but net supply change also depends on how much new ETH is issued to validators at the time, so supply can still grow or shrink.
Are layer-2 transactions free?
No, but they are generally much cheaper than mainnet because data-posting costs are shared across many users, and blobs have lowered that cost since Dencun.
Do the Fusaka and Pectra upgrades make mainnet transactions cheaper?
Not directly. Their main effect is more data capacity for layer-2s. Mainnet fees still depend on demand for block space.
Can I predict gas fees in advance?
Reasonably well with gas trackers, which estimate current congestion. A sudden demand spike can still move fees faster than any estimate.
Sources and Further Reading
- ethereum.org: Gas and fees (base fee, tip, max fee, 21,000 gas, 12.5% adjustment)
- ethereum.org: Dencun upgrade (activation 13 March 2024, EIP-4844 blobs)
- ethereum.org: Pectra upgrade (activation 7 May 2025, EIP-7691, EIP-7623)
- ethereum.org: Fusaka upgrade (activation 3 December 2025, PeerDAS, BPO forks, EIP-7918)
- Related on Fiscal Frontier: Layer-2 comparison, MEV explained, Uniswap and AMMs
This article is for educational purposes only and is not financial advice. Protocol rules change with upgrades; the details above reflect ethereum.org documentation as of October 2026. Always do your own research (DYOR) before making investment decisions.