Gwei is a denomination of ether used to quote gas prices. One gwei equals one-billionth of an ETH (0.000000001 ETH). A base fee of 10 gwei means each unit of gas costs 0.00000001 ETH.

Why do gas fees fluctuate so much?

Block space is limited and demand is not. When many users compete to transact at once — during volatile markets, popular mints, or large liquidation events — the base fee ratchets up with every over-target block, and users add larger priority tips on top. When demand fades, the base fee decays just as quickly.

What happens to the base fee after I pay it?

It is burned — permanently destroyed by the protocol rather than paid to anyone. Only the priority tip goes to the validator. The burn, introduced by EIP-1559 in August 2021, removes ETH from circulation with every transaction.

Do I get back unused gas?

Yes. If your transaction uses fewer gas units than your gas limit, the unused units are not charged. You are also refunded the difference between your max fee and the actual base-fee-plus-tip per gas. You only overpay if you set an unnecessarily high priority tip.

Why did my failed transaction still cost money?

Because validators performed the computation before it failed. A revert or out-of-gas error still consumed real network resources up to the point of failure, and gas pays for computation attempted, not outcomes. The transaction's effects are rolled back, but the gas spent is not returned.

Do Layer 2 transactions use gas too?

Yes. L2s use the same gas model — units of computation priced in gwei and paid in ETH on most networks — but execution happens off Ethereum's mainnet and data is posted in cheap blobs, so the same swap that costs dollars on L1 typically costs a few cents or less on an L2.

Can I pay Ethereum gas fees with tokens other than ETH?

Not at the protocol level on mainnet: the base fee and tip are denominated and paid in ETH. Some wallets and smart-account 'paymaster' services let you fund fees with other tokens by converting behind the scenes. Among L2s, Starknet natively accepts STRK for fees alongside ETH.

Why are Ethereum gas fees so low in 2026?

Three forces stacked: the block gas limit doubled from 30 million to 60 million during 2025, expanding supply; most routine activity migrated to Layer 2s, reducing L1 demand; and rollup data moved into the separate blob fee market, freeing regular blockspace. The result is average gas prices around half a gwei through much of 2026 — with simple transfers costing on the order of a cent — versus tens or hundreds of gwei at the 2021 peak.

What is the blob fee market?

A second, independent fee market introduced by EIP-4844 for the ~128 KB data packages (blobs) that rollups post to Ethereum. Blob gas has its own base fee that adjusts exponentially around a per-block target and burns like the regular base fee, but the two markets move independently: regular gas can spike while blob gas stays at its minimum, and vice versa. After the Fusaka upgrade and its parameter forks, blocks target 14 blobs with a maximum of 21 as of January 2026.

What is EIP-7702, and can an app pay gas for me?

EIP-7702, shipped in the May 2025 Pectra upgrade, lets an ordinary wallet address temporarily act as a smart-contract account. That unlocks batching (approve and swap in one transaction instead of two) and gas sponsorship: a paymaster service can pay your fee in ETH while charging you in another token like USDC, or an app can cover it entirely. The protocol still gets paid in ETH — sponsorship changes who pays, not what the network accepts.

Will Ethereum's gas limit keep increasing?

That is the stated plan. Validators raised the limit from 30 million to 60 million during 2025, and the Ethereum Foundation's 2026 priorities target pushing it toward and beyond 100 million, with researchers eyeing roughly 200 million once the Glamsterdam upgrade ships enshrined proposer-builder separation and block-level access lists. Increases happen gradually via validator signaling so node operators can be sure the network handles the larger blocks safely.

Ethereum Gas Fees Explained

Gas is the metering system of Ethereum: every computation the network performs is measured in gas units and paid for in ETH. Understanding how those units are counted, how the EIP-1559 fee market prices them, and why fees spike turns gas from an annoying surprise into something you can reason about — and mostly avoid overpaying for.

Key Takeaways

Gas measures computational work; your fee is gas units used multiplied by the per-unit price: Total fee = gas used × (base fee + priority tip).
The base fee is set by the protocol, not by validators, and it is burned — permanently destroyed — with every block.
Blocks target 50% fullness; the base fee moves at most 12.5% per block, up when blocks run over target and down when they run under.
A simple ETH transfer always costs exactly 21,000 gas; DEX swaps and other contract interactions cost several times more.
Fee spikes are demand events: volatile markets, mints, and liquidations make users bid up priority tips for scarce block space.
Layer 2s run the same gas model off-mainnet and post data via cheap blobs, which is why the same action costs cents there.
2026 is the cheapest fee era in Ethereum's history: average gas prices hover around half a gwei, the block gas limit doubled to 60 million in 2025 with 100M+ targeted, and smart-account paymasters can sponsor fees entirely.

What Gas Actually Measures

Per ethereum.org, gas is "the unit that measures the amount of computational effort required to execute specific operations on the Ethereum network." Every low-level operation the Ethereum Virtual Machine can perform — adding numbers, reading storage, writing storage, calling another contract — has a fixed gas cost defined by the protocol. A transaction's gas usage is simply the sum of the operations it triggered.

Two things follow from this design. First, gas is a unit of work, not a price — the price attaches when you multiply gas by an ETH-denominated rate. Second, fees exist for security, not revenue: because every node re-executes every transaction, unpriced computation would invite spam and accidental (or hostile) infinite loops. Making each unit of work cost something is what keeps a global shared computer usable.

The Units: Wei, Gwei, ETH

Gas prices are quoted in gwei, a small denomination of ether: one gwei is one-billionth of an ETH (0.000000001 ETH). The name is short for "giga-wei" — a wei being the smallest unit of ETH, as a cent is to a dollar, but divided much further. Quoting in gwei keeps the numbers readable: saying the base fee is "12 gwei" is easier than "0.000000012 ETH per gas unit."

Gas Limit vs Gas Price

Two separate dials control what a transaction can spend, and confusing them is the most common gas mistake:

The gas limit is the maximum amount of work you authorize. A plain ETH transfer needs exactly 21,000 gas, so wallets set that automatically. Contract interactions are less predictable, so wallets estimate and add a buffer. If execution hits the limit before finishing, the transaction fails with an "out of gas" error — its effects are reverted, but the gas consumed is still paid, because validators did the work. Unused gas below the limit is simply not charged.
The gas price is what you pay per unit, set by the fee market described below. Since EIP-1559, you do not pick a single price; you set a maximum fee and a tip, and the protocol computes the actual rate.

Blocks have a gas limit too: the total work one block may contain. Validators adjust it gradually by signaling — they raised it to 60 million gas in late 2025 — and under EIP-1559 each block targets half its limit, leaving headroom for demand surges.

EIP-1559: Base Fee, Burn, and Tip

Before August 2021, Ethereum priced gas with a blind first-price auction: users guessed what to bid, chronically overpaying. The EIP-1559 upgrade (London hard fork) replaced that with a protocol-computed price and a small auction on top:

Base fee. The protocol sets a per-gas base fee algorithmically from how full recent blocks were. Blocks target 50% of the gas limit (an "elasticity multiplier" of 2 in the spec); when a block runs over target the base fee rises for the next block, and when it runs under, it falls — by at most 12.5% per block in either direction. Sustained congestion therefore compounds the base fee exponentially until demand backs off, which is exactly the mechanism that makes fee spikes sharp but self-limiting.
The burn. The base fee is not paid to validators — it is destroyed. This removes the incentive for validators to manipulate fees by stuffing blocks, and it ties ETH's supply to network usage: more than six million ETH had been burned by late 2025, per burn-tracker estimates. The burn is a key part of the ETH value-accrual argument.
Priority fee (tip). The only part validators keep. It is your bid for placement: in quiet conditions a minimal tip suffices, while during congestion users raise tips to jump the queue.
Max fee. You also set a ceiling (max fee per gas). If the base fee plus your tip comes in under the ceiling, you pay only the actual amount and the difference is refunded — so a generous max fee protects against base-fee movement without forcing you to overpay.

Putting it together with a concrete example: a 21,000-gas transfer with a 10 gwei base fee and a 2 gwei tip costs 21,000 × 12 gwei = 252,000 gwei, or about 0.00025 ETH. The 10-gwei share is burned; the 2-gwei share goes to the block's validator.

Typical Gas Costs of Common Operations

Because gas counts work, cost scales with complexity. A transfer touches two balances; a swap reads pool state, computes prices, moves two tokens, and updates reserves. Approximate figures, per ethereum.org and wallet-industry estimates:

Operation	Approximate gas units	Why
Simple ETH transfer	21,000 (fixed)	The protocol minimum; no contract code runs
ERC-20 token transfer	~45,000–65,000	Executes contract code and updates storage slots
ERC-20 approval	~45,000	One storage write authorizing a spender
DEX swap	~100,000–200,000	Multiple contract calls, price math, two token moves
Complex DeFi interaction / contract deployment	Hundreds of thousands to millions	Many storage writes; deployment scales with code size

Storage is the expensive part: writing a new storage slot costs far more than arithmetic, which is why "simple-looking" actions that create new state (a first-time token balance, a new position) cost more than repeat interactions.

A Brief History of Ethereum Fees

Fee levels have always tracked whatever Ethereum's dominant use case was that year, and the history explains why each piece of today's machinery exists. The first real congestion event was CryptoKitties in December 2017, when a single collectibles game backed up the whole network — the earliest demonstration that one popular contract could price out everyone else. DeFi summer 2020 made congestion chronic rather than episodic: suddenly there were financial reasons to pay almost any fee, and gas regularly ran into the hundreds of gwei. The 2021 NFT boom stacked a second demand wave on top, producing the $50-swap era that still defines Ethereum's reputation for many people.

Each pain point produced a protocol response, traceable through ethereum.org's upgrade history: the London upgrade (August 2021) shipped EIP-1559 to end blind fee guessing; the rollup-centric roadmap moved execution demand to L2s; Dencun (March 2024) gave those L2s their own cheap data lane; and the Pectra/Fusaka sequence of 2025 doubled the gas limit's ceiling and multiplied blob capacity. Notably, the Merge (September 2022) — often assumed to be a fee upgrade — changed consensus, not capacity, and had no direct effect on gas prices. The cumulative result is the environment described next: the busiest settlement layer in crypto, running cheaper than at any point since its early years.

How Cheap Is Gas in 2026? (And Why)

If your mental model of Ethereum fees was formed in 2021 — $50 swaps, $200 mints — it is badly out of date. As of 2026, mainnet gas spends most of its time below one gwei: Etherscan's gas tracker routinely shows sub-gwei prices, daily averages ran around 0.5 gwei in early August 2026, and April 2026 averaged roughly 0.47 gwei — down from about 1.67 gwei a year earlier, per 2026 gas-fee statistics compilations. In dollar terms, a simple transfer costs on the order of a cent, and even swaps frequently land under a dollar on L1 itself.

Run the arithmetic from the earlier example at today's prices and the change is stark: 21,000 gas × 0.5 gwei is about 0.0000105 ETH — a hundredth of what the same transfer cost at 50 gwei. Three structural forces, not a lull in interest, drove this:

Supply doubled. Validators raised the block gas limit from 30 million to 60 million during 2025 (next section), doubling the space every block offers.
Demand migrated. The routine activity that used to congest L1 — retail swaps, transfers, minting — now mostly happens on Layer 2s, which post compressed data back to Ethereum instead of competing for ordinary blockspace.
Rollup data moved lanes. Blobs took rollup data out of the regular fee market entirely, so L2 growth no longer bids up L1 gas.

The flip side: cheap gas means less base fee is burned per block, which has revived debate about ETH's fee-driven economics (covered in ETH vs L2 tokens). And "usually cheap" is not "always cheap" — the ratchet mechanics below still produce sharp spikes when a major event hits; they just start from a far lower floor.

The Gas Limit: From 30 Million Toward 200 Million

The block gas limit — total work per block — is not fixed in the protocol; validators adjust it by signaling, with each block allowed to nudge the limit slightly, so sustained majority support moves it over days. For years it sat at 30 million while client teams prioritized safety. That conservatism ended in 2025: the limit stepped up through 36 and 45 million before validator signaling carried it to 60 million by late 2025, per The Block's coverage of the signaling campaigns.

The roadmap goes further. The Ethereum Foundation's published 2026 priorities aim to push the limit toward and beyond 100 million, and researchers see roughly 200 million as reachable after the Glamsterdam upgrade — in final testing as of August 2026 — ships enshrined proposer-builder separation (EIP-7732) and block-level access lists (EIP-7928), which widen the block propagation window and let nodes validate transactions in parallel, per The Defiant. For fee-payers the implication is one-directional: each increase adds supply, and each supply increase lowers the congestion level at which fees start to climb.

Why Fees Spike

Fee spikes are always the same story: fixed short-term supply of block space meeting a burst of inelastic demand. The recurring triggers:

Volatile markets. Sharp price moves trigger urgent trading, arbitrage, and — in DeFi — waves of liquidations whose executors will pay almost any tip, because the liquidation bonus dwarfs the fee.
Launches and mints. Popular NFT mints and token launches concentrate thousands of users on one contract in minutes.
Airdrop claims and migrations. Deadline-driven events synchronize demand that would otherwise spread out.
MEV competition. Bots bidding against each other for profitable ordering (next section) add high-tip transactions at exactly the moments markets are most active.

The base-fee algorithm turns these bursts into a visible ratchet: each over-target block compounds the base fee by up to 12.5%, so fees can multiply within minutes — and decay just as fast once the rush ends. Watching activity concentrate in one corner of the market on our live bubble map is often watching tomorrow's congestion form; current network conditions are visible on our ETH gas page.

MEV and Priority Fees

Some participants earn money from where their transactions land in a block, which makes them price-insensitive tip bidders. This is MEV — maximal extractable value — defined on ethereum.org as the value extractable from block production "by including, excluding, and changing the order of transactions," beyond standard rewards.

The common forms:

Arbitrage. Buying a token cheaper on one DEX and selling it dearer on another in a single atomic transaction — riskless if it lands, worthless if it does not, so searchers bid tips aggressively.
Liquidations. Racing to be the one who liquidates an undercollateralized loan and collects the fee.
Sandwich attacks. Spotting a large pending swap, buying just before it and selling just after, profiting from the price impact — which shows up for the victim as worse execution. This is why slippage limits matter when trading, especially in thin markets (see our guide on token liquidity).

For ordinary users, the practical takeaways are modest: MEV competition is part of why tips spike during volatility, most MEV flow now travels through private channels to specialized block builders rather than the public mempool, and tight slippage settings are your main personal defense. Large trades moving through the market are also visible on-chain — our guide to tracking Ethereum whales covers how.

How Layer 2s Change the Math

Layer 2 rollups do not abolish gas; they relocate it. Execution happens on the L2 (same gas-unit model, usually paid in ETH), and only compressed transaction data returns to Ethereum. Since the Dencun upgrade in March 2024, that data travels in "blobs" — roughly 128 KB data packages, defined by EIP-4844, priced in their own fee market that is separate from (and usually far cheaper than) regular gas. The Fusaka upgrade of December 2025 and its follow-up parameter forks raised blob capacity to a target of 14 and maximum of 21 per block by January 2026, per the Ethereum Foundation.

One L1 blob purchase is shared by thousands of L2 transactions, so per-user costs collapse: a mid-2026 academic study measured median fees of roughly $0.02 per transaction on Base and $0.04 on Arbitrum One. For most everyday activity, the practical answer to "how do I pay less gas" is simply "do it on an L2" — you can compare the networks on our L2 dashboard.

The Blob Fee Market: Ethereum's Second Fee Lane

Blobs are worth understanding as a fee market in their own right, because they mirror the EIP-1559 design while staying fully independent of it. Each block has a blob target and maximum; blob gas has its own base fee, which rises exponentially while blocks carry more blobs than target and decays when they carry fewer; and like the regular base fee, the blob base fee is burned. Because the two markets are separate, an NFT-mint frenzy can send regular gas to 50 gwei while blob gas sits at its floor of a single wei — and a rollup data surge can, in principle, price blobs up while ordinary transfers stay cheap.

Capacity has been scaling on an aggressive schedule. Fusaka's PeerDAS lets nodes verify blob availability by sampling instead of downloading everything, which allowed the Ethereum Foundation to ship "blob parameter only" forks between hard forks: BPO1 (December 2025) raised the target/max from 6/9 to 10/15, and BPO2 (January 7, 2026) raised it again to 14/21, per the Foundation's January 2026 checkpoint. Further steps toward roughly 48 blobs are planned as the mechanism proves out, per the ethereum.org roadmap. For most of blob history, supply has outrun demand — which is exactly why blob fees, and therefore L2 fees, have stayed near their floors.

Smart Accounts and Sponsored Gas (EIP-7702)

The May 2025 Pectra upgrade quietly changed who has to pay gas. EIP-7702 lets an ordinary wallet address (an EOA) temporarily designate smart-contract code for itself, giving normal wallets the powers that previously required a dedicated smart-account contract:

Batching. The classic two-step of "approve, then swap" becomes one atomic transaction — one base-fee payment, one signature, and no dangling approvals. For interactions that used to require several transactions, this is a direct gas saving.
Gas sponsorship. A paymaster — a third-party service or the app itself — can pay the ETH fee on your behalf and charge you in another token, or nothing at all. Circle's paymaster, for example, lets users transact holding only USDC, with the ETH gas handled invisibly, per developer documentation of the 7702 ecosystem.
Session keys and spending limits. Wallets can authorize limited-scope keys so an app can act for you without a signature per action — fewer transactions, less gas, less friction.

Adoption moved fast: major wallets shipped 7702 support through 2025, and by 2026 sponsored transactions number in the millions daily across Ethereum and its rollups, per 2026 account-abstraction surveys. None of this changes the protocol economics — the network is still paid in ETH, the base fee still burns — but it increasingly means the person transacting is not the one watching the gas price. One caution: signing a 7702 authorization delegates real control of your account to the designated code, so only authorize implementations from wallets and apps you trust.

Practical Fee Estimation and Timing

A few habits cover most of what fee optimization can honestly deliver:

Trust the structure, tune the tip. Your wallet reads the current base fee from the chain; the only judgment call is the tip. Urgency is the question — settlement-time-sensitive actions deserve a higher tip, and anything else can ride the default.
Set the max fee generously, not the tip. Because the unused portion of the max fee is refunded, a high ceiling costs nothing extra and prevents your transaction from stalling if the base fee climbs while it is pending.
Time flexibly, in UTC. Demand follows human activity: weekends and hours outside overlapping US/European trading tend to run quieter, though a single market event can override any pattern. Timing helps at the margin; it is not a strategy.
Do not starve the gas limit. Trimming a wallet's estimated limit to "save" fees is false economy — you do not pay for unused gas, but an out-of-gas failure costs real money and achieves nothing.
Batch and prune. Fewer approvals, consolidated transfers, and doing repetitive activity on an L2 beat any amount of tip micro-optimization.

Stuck, Pending, and Replaced Transactions

The fee market also explains the most common transaction pathologies:

Stuck pending. If the base fee rises above your max fee after you submit, your transaction cannot be included until conditions ease — it waits in the mempool, valid but unaffordable at current prices. This is the failure mode that a generous max fee prevents.
Speed-ups and cancellations. Every transaction from an account carries a sequential nonce, and validators will accept a replacement for a pending nonce if it pays a sufficiently higher fee. A wallet's "speed up" button rebroadcasts the same transaction with better pricing; "cancel" replaces it with a do-nothing transfer to yourself. Both are new transactions — a cancellation still pays gas.
Ordering constraints. Because nonces are sequential, one underpriced transaction blocks everything queued behind it from the same account. If a batch of transactions is stuck, fixing the oldest one releases the rest.

None of this requires manual intervention in normal conditions; modern wallets estimate the base fee from recent blocks and pick sensible defaults. Understanding the mechanics matters when conditions are not normal — precisely the moments when mistakes are most expensive.

Frequently Asked Questions

Related Guides

Ethereum Layer 2s Explained — the scaling architecture that moved most everyday transactions off expensive L1 blockspace.
ETH vs L2 Tokens — how the gas burn feeds into ETH's economics compared with L2 governance tokens.
Ethereum DeFi Explained — the applications whose liquidations and trading drive the sharpest fee spikes.
What Is the Ethereum Ecosystem? — the broader map of networks and assets this fee machinery supports.

Sources & Further Reading

Disclaimer: Gas mechanics described here reflect the protocol as of August 2026; parameters such as block gas limits and blob counts change through network upgrades and validator signaling. Always review your wallet's estimated fees and gas limit before confirming a transaction. This guide is educational only and is not financial advice.