From base fees and priority tips to blob transactions and Layer-2 economics—everything you actually need to know about how Ethereum charges for computation.
• Gas is the unit that measures the computational work an Ethereum transaction requires. Gas fees are what you pay for that work, denominated in gwei (1 gwei = 0.000000001 ETH).
• Since the London upgrade (August 2021), every transaction fee is calculated as gas used × (base fee + priority tip). The base fee is burned; the tip goes to the validator — not a miner. Ethereum became a proof-of-stake network at The Merge in September 2022.
• Since the Dencun upgrade (March 2024), Layer-2 rollups have their own cheap blob data lane on Ethereum. That single change cut L2 fees by roughly 10–100×.
• Two follow-up upgrades—Pectra (May 2025) and Fusaka (December 2025)—further expanded blob capacity and brought programmable smart-account features to regular wallets.
• As of May 2026, Ethereum mainnet gas typically sits well belowwellbelow 1 gwei. A simple ETH transfer costs around $0.10–$0.20 on L1 and fractions of a cent on most Layer 2s.
This is a complete rebuild of how Ethereum prices transactions, written for the network as it actually exists today—not as it did in 2021.
Every action on Ethereum—sending ETH, swapping a token, minting an NFT, calling a smart contract—uses computing power supplied by thousands of nodes around the world. Each of those nodes runs the same code, verifies the same state, and stores the same results. That coordination has a cost, and gas is how Ethereum measures and prices it.
Two reasons gas exists at all:
1. Incentives. Incentives. Validators commit hardware and staked capital to run the network. Fees pay them for the service. Without compensation, no rational party would lend infrastructure to the chain.
2. Spam prevention. If transactions were free, anyone could flood the network with junk to slow it down or crash nodes. Attaching a cost to every operation—even a small one—makes denial-of-service attacks prohibitively expensive.
Gas is paid in ether (ETH), the native currency of the Ethereum network, and priced in a smaller unit called gwei (short for “giga-wei”). One gwei is one billionth of an ETH (0.000000001 ETH). The unit exists for the same reason we quote gasoline incentives per gallon instead of dollars per microgallon: it’s easier to say "0.5 gwei" than “0.0000000005 ETH.”
The smallest denomination of ETH is the wei (1 ETH = 10¹⁸ wei), but in practice nobody quotes anything in wei. Gwei is the working unit for end users and developers alike.
A useful mental model: gas is thegas isthe work; gwei is the price per unit of work; ETH is the currency you settle in.
Here is the formula every Ethereum user—from a once-a-month DeFi participant to a smart contract developer—should commit to memory:
Transaction fee = Gas used × (Base fee + Priority tip)
Three variables, each playing a distinct role:
You also set two ceilings:
• Gas limit — the maximum amount of gas you authorize the EVM to consume for your transaction. Any unused gas is refunded.
• Max fee (per gas) — the maximum gwei price you are willing to pay per unit. If the base fee plus your tip exceeds this, the transaction won’t be included until conditions change.
This entire structure was introduced by EIP-1559 as part of the London upgrade in August 2021. Before then, Ethereum used a simple first-price auction: you bid whatever you thought would get included, and you paid that full amount whether or not it was actually necessary. Fees were unpredictable, frequently overpaid, and miserable to forecast.
The base-fee burn is the most consequential—and most often misunderstood—piece of Ethereum’s fee model.
Burning means destroying. When you pay the base fee, that ETH is not sent to anyone. It is removed from the total ETH supply forever. This accomplishes two things:
1. It anchors fees in real demand. The base feed readjusts smoothly between blocks—up to 12.5% per block—based on whether the previous block was over- or under-filled. This produces a much more predictable fee curve than the old auction model.
2. It ties network usage to monetary policy. Whenactivity is heavy enough that the daily ETH burn exceeds new ETH issued to validators, Ethereum becomes net deflationary. This is why peoplesometimes describe ETH as “ultrasound money” — its supply curve flexes withusage rather than mechanically inflating like Bitcoin’s mining schedule.
The tip, by contrast, is a normal payment. It goes to the validator, who includes your transaction in their block. During congestion, raising your tip is how you bid for priority.
Let's say Alice sends Bob 1 ETH on the Ethereum mainnet in May 2026.
• Gas used: 21,000 (the standard for a simple ETHtransfer — this is fixed)
• Base fee: 0.4 gwei (typical mainnet base fee inlate May 2026)
• Priority tip: 0.1 gwei (a reasonable default fornon-urgent transactions)
• Max fee: 1 gwei (Alice’s ceiling—generous enough that small spikes won’t block the transaction)
Calculation:
Effective gas price = 0.4 + 0.1 = 0.5 gwei.
Total fee =21,000 × 0.5 gwei = 10,500 gwei = 0.0000105 ETH
At an ETH price of roughly $2,500, that comes to about $0.026—a couple of cents. Alice’s wallet is debited 1.0000105 ETH. Bob receives exactly 1 ETH. The validator earns 2,100 gwei in tips. The remaining 8,400 gwei (the base fee) is burned.
Now compare this to October 2021, when the original version of this article was written. Back then, average gas was around 131 gwei. The same simple transfer would have cost roughly 131 × 21,000 = 2,751,000 gwei, or about 0.00275 ETH. With ETH at $4,000 then, that’s around $11 — for a single transfer.
The drop from $11 to 3 cents is not a market accident. It is the cumulative result of EIP-1559, The Merge, Dencun, Pectra, Fusaka, and the migration of activity to Layer 2 rollups. Each of these matters, and we'll unpack them below.
Not every transaction uses the same amount of gas. The 21,000-unit baseline applies only to a plain ETH transfer between two externally owned accounts. Anything that touches a smart contract—and most things you do on Ethereum today touch a smart contract somewhere—costs more.
A smart contract interaction can cost 5–10× more than a simple transfer because the EVM has to execute every instruction, write to storage (which is the single most expensive operation), and emit logs. Storage writes are where most gas goes in non-trivial contracts—writing a new value to storage costs 22,100 gas; updating an existing value costs 5,000.
This is why gas-efficient contract design is a real discipline. Saving 5,000 gas per call on a function that gets executed millions of times produces meaningful aggregate savings.
The gas limit is the most misunderstood field in any Ethereum wallet. Most users never touch it; most wallets estimate it automatically. But understanding what it does protects you from a class of avoidable losses.
The gas limit is the maximum amount of gas you authorize the EVM to spend trying to execute your transaction. Two outcomes are possible:
• You set it too low. The EVM runs your transaction, hits the limit before finishing, and reverts. The transaction fails—but the EVM still consumed gas trying. You lose the gas you spent, and the operation does not complete.
• You set it too high. No penalty whatsoever. The EVM only charges for the gas it actually uses and refunds the difference.
This is asymmetric in a useful way: setting your limit a bit high costs nothing, while setting it too low costs everything. Wallets default to estimating gas via a dry-run simulation, then adding a buffer. Most of the time this is fine. The places it can fail:
• Recently changed contract state. A simulation that succeeded a minute ago might fail now if conditions shifted (price changed, liquidity moved, or an approval expired).
• Highly variable execution paths. Contracts with conditional logic—loops, switch statements, dynamic-size operations—can consume wildly different amounts of gas depending on inputs.
• Maliciously crafted contracts. Some scam contracts deliberately consume your full gas limit during a "transfer" to extract value through fees.
When in doubt, set the limit a bit higher than the wallet's estimate. The unused portion comes back to you.
Ethereum's fee history has four turning points. Understanding them is the difference between a 2021 mental model and a 2026 one.
Before London, gas was a first-price auction. You guessed a price, paid whatever you bid if you got included, and frequently overpaid by enormous margins during congestion. Fleeing wildly from block to block.
EIP-1559 replaced this with the base-fee-plus-tip model described above. Three immediate consequences:
• Fees became predictable. The base fee moves at most 12.5% between blocks. You can forecast the next block’s fee with near-perfect accuracy.
• Refunds became automatic. The difference between your max fee and actual fee is refunded. You no longer accidentally pay 10× the going rate.
• ETH supply became reactive. With the base fee burned, ETH issuance and ETH destruction now balance against each other based on network usage.
London did not, by itself, reduce average fees. It made them better-behaved.
In September 2022, Ethereum executed The Merge—switching its consensus mechanism from proof-of-work (mining) to proof-of-stake (validators). This was a fundamental change with several knock-on effects:
• No more miners. Anywhere you read about Ethereum mining as a current activity, the source is outdated. Validators replaced miners. To run one, you stake 32 ETH (now up to 2,048 ETH after Pectra) instead of buying GPUs.
• Energy use dropped ~99.95%. Ethereum went from a small-country level of electricity consumption to a rounding error.
• Issuance dropped sharply. New ETH issued per day fell by roughly 90%, since validators are paid less than miners were. Combined with the base-fee burn, this is why ETH can be net deflationary during busy periods.
The Merge did not directly affect gas fees—block space and the EIP-1559 model stayed the same—but it changed who gets paid and how the network's economics work.
This is the single most important fee event in Ethereum's history. Before Dencun, Layer-2 rollups (Arbitrum, Optimism, Base, zkSync, and others) had been promising cheaper transactions for years—but their cost floor was tied to mainnet. Rollups work by batching thousands of L2 transactions and posting a compressed summary back to Ethereum. That summary used to be posted as ordinary on-chain calldata, which lives in Ethereum’s state forever and competes with other transactions for block space.
EIP-4844 introduced a new transaction type: blob-carrying transactions. Instead of calldata, rollups can now post their batched data as “blobs”—temporary datapackets that live on Ethereum’s consensus layer for about 18 days, then get pruned. Blobs have their own separate fee market, decoupled from regular transaction demand.
The numbers, within a week of Dencun's activation on March 13, 2024:
• Median L2 fees on Arbitrum, Optimism, Base, and Lineadropped from the $0.20–$1.00 range to $0.01–$0.10.
• L2 transaction volume doubled, then tripled over the following months.
• Many L2-native consumer apps that had been economically marginal—micropayments, social posts, in-game actions—suddenly became viable.
By late 2025, L2s were handling roughly 95% of Ethereum's total transaction volume. The rollup-centric vision Vitalik Buterin sketched in 2020 is now Ethereum's lived reality.
Pectra activated on Ethereum mainnet on May 7, 2025. It bundled 11 EIPs—the largest single upgrade in Ethereum's history at that point—but two changes matter most for users:
• EIP-7702: programmable EOAs. Regular wallets (externally owned accounts) can now temporarily behave like smart contracts during a transaction. This unlocks transaction batching, gas sponsorship (someone else can pay your fees), session keys, and social-recovery flows—features that previously required full smart-contract wallets like Safe or Argent.
• EIP-7691: doubled blob throughput. The blob target per block doubled from 3 to 6, with a maximum of 9. More blob capacity means more headroom for L2s before blob fees start rising.
Pectra also raised the maximum validator stake from 32 ETH to 2,048 ETH, which simplifies operations for large stakers but doesn't directly affect end users.
Fusaka activated on December 3, 2025. Its headline feature is PeerDAS (Peer Data Availability Sampling), introduced by EIP-7594. PeerDAS lets validators verify blob data by sampling small portions of it across the network rather than downloading every blob in full. This dramatically reduces the bandwidth required to run a validator and removes the bottleneck that prevented further blob expansion.
In the days following Fusaka, two Blob Parameter Only (BPO) forks pushed the per-block blob target from 6 to 14 and the maximum from 9 to 21. The long-term roadmap targets 128 blobs per block under full Danksharding.
In practical terms: more blobspace, lower L2 fees, and fewer congestion spikes on rollups during major events (airdrops, big NFT mints, viral on-chain moments).
The single most useful fact about Ethereum fees in 2026 is that the answer to "What does a transaction cost?" depends entirely on which layer you’re using.
Approximate median fees for atypical token swaps (May 2026):
These are ballpark medians, not guarantees—fees rise during congestion and fall during quiet periods on every chain. But the rough pattern is durable: mainnet is roughly 10–100× more expensive than rollups, and the gap is sticky.
When does Mainnet still sense?
• Large transactions where fees are a rounding error. A $10MA$10M settlement doesn’t care about a $2 fee.
• Bridging and final settlement. Moving value between L2s often routes through L1.
• Contract deployments meant for permanent reference. Deploying onceDeployingonce on L1 is more durable than deploying on a single L2.
• Composability with L1-native protocols. Some lendingSomelending markets, staking flows, and governance systems live primarily on L1.
For nearly everything else—daily trading, NFT minting, gaming, social apps, payments—an L2 is the right default.
Most “save on gas” advice from 2021—transact on weekends, transact between midnight and 4 AM, watch the gas tracker like a hawk—is now mostly obsolete. Fees are low almost all the time, and the techniques that matter have shifted accordingly.
What still works in 2026:
3. Default to a layer 2. Unless you have a specific reason to use mainnet, you should be on a rollup. The user experience is nearly identical, and the cost difference is 10–100×.
4. Use a fee tracker before unusual transactions. Etherscan's gas tracker, Blocknative, or your wallet’s built-in estimator will show you the current base fee and recommended tips. If the mainnet is unusually busy, a one-hour wait can save real money on large transactions.
5. Batch operations when possible. EIP-7702 (Pectra) makes this much easier—wallets can now batch an approval and a swap into a single transaction, saving the gas overhead of two separate ones. Many modern wallets do this automatically.
6. Set a sensible max fee, not a tight one. The difference between your max fee and the actual fee is refunded. Setting it generously high doesn’t cost you anything; setting it too tightly during a spike just means your transaction gets stuck pending.
7. Use gas sponsorship where available. Many consumerManyconsumer dApps (especially on L2s) now pay users’ gas via paymasters. If a dApp offers gasless transactions, take advantage.
8. Be deliberate during major events. Big airdrops, viral NFT mints, and high-profile launches still produce visible fee spikes—both on L1 and on the affected L2. Either move quickly or wait it out.
9. For contract developers: optimize storage. Storage operations dominate gas costs in real contracts. Packed structs, immutable variables, off-chain storage with on-chain hashes, and careful use of mappings versus arrays all matter.
Are gas fees still paid to miners?
No. Ethereum transitioned from proof-of-work to proof-of-stake in September 2022—an event called "The Merge." Transaction tips now go to validators, who stake ETH instead of running mining hardware. The term “miner” no longer applies to Ethereum, though you’ll stillsee it in older articles.
Why is part of every fee burned?
The base fee is permanently destroyed. This makes fees more predictable than the old auction model and gradually reduces ETH supply during periods of high network usage. When dailyburn exceeds new issuance, ETH becomes net deflationary.
What's the difference between gas, Gwei, Ether, and Wei?
Gas is the unit of computational work. Gwei is the price per unit of gas. Ether (ETH) is the currency the fee is paid in. Wei is the smallest possible denomination of ETH (1 ETH = 10¹⁸ wei). In practice, you’ll only ever see prices quoted in gwei and ETH.
Can a transaction failand still cost me money?
Yes. If your gas limit is too low, if a smart contract reverts due to a changed state, or if you tried to interact with a malicious contract, the EVM still consumes the gas it used trying to execute. Most wallets warn you when their pre-flight simulation predicts a revert.
Do I need ETH to use a layer 2?
You need a small amount of ETH on the L2 itself to pay gas. Most bridges and onboarding tools handle this in one step. Some L2s let you pay gas in stablecoins or other tokens viapaymasters — this is becoming more common.
Why are some transactions much more expensive than others on the same chain?
Because they consume more gas. A simple transfer is 21,000 units; a complex DeFi interaction can be 300,000+. The fee scales with the work, not the value being transferred. Sending $100,000 worth of ETH costs the same as sending $10.
Why do the networks sometimes have spikes even though average fees are low?
Two reasons. First, demand isn't constant—major NFT mints, airdrops, or arbitrage opportunities can fillblocks for short periods. Second, the base fee can move 12.5% per block, so a sustained burst of activity can quickly raise the base fee severalfold before usage drops back to normal.
What is "Danksharding," and when will it ship?
Full Danksharding is the long-term vision: dramatically expanded data availability via sharded blob storage and PeerDAS-style sampling. Fusaka (December 2025) shipped the first major component (PeerDAS). The remaining pieces—primarily expanding to 128 blobs per block—will roll out incrementally through future upgrades and BlobParameter Only forks.
Is ETH inflationary or deflationary?
It depends on usage. Validators are paid newly issued ETH, but the base fee is burned. When transaction activity is high (and base fees are non-trivial), the burn outpaces issuance and ETH supply shrinks. When activity is low, supply grows slowly. Over the network spent more time in deflationary periods than inflationary ones.
Gas isn't a tax. It's the price signal that keeps Ethereum running honestly. Validators get paid for real work, spam gets priced out, and the base-fee burn ties network usage directly to ETH's monetary policy. The combination of EIP-1559, The Merge, EIP-4844, and the L2 explosion has transformed what was once Ethereum's biggest UX problem into something most users barely notice.
If you’re a user: default to a Layer 2 for everyday activity, watch the gas tracker only for unusually large mainnet transactions, and let your wallet handle gas limits unless you have reason to override them.
If you’re a developer: design contracts to be gas-efficient regardless of which layer they live on. Storage operations are where most gas goes. Take advantage of EIP-7702 to give your users batched, sponsored, and recoverable transaction flows.
If you’re a business evaluating Ethereum infrastructure, the economics now reward both ends of the stack. L1 gives you maximum security and the deepest liquidity. L2s give you near-free transactions with the same settlement guarantees. The right architecture combines both.
Building on Ethereum or evaluating blockchain infrastructure for your business? Spydra provides enterprise-grade tokenization, smart contracts, and decentralization-as-a-service across both public and private networks. Talk to our team at spydra.app/contact-us or explore the platform at spydra.app.
