What is gas?

Gas is the fee you pay to do anything on Ethereum. Sending ETH, swapping a token, minting an NFT, voting in a DAO — all of them consume a metered amount of computation on the network, and that computation has to be paid for in ETH. Gas is how Ethereum turns a shared computer into a priced resource.

Why computation needs a price

Ethereum is a single global state machine that every node has to keep in sync. Every transaction runs on every node that validates blocks. If the network were free, anyone could publish an infinite loop and halt the chain. Gas solves that by charging for each low-level operation (opcode) the virtual machine executes. A simple transfer costs 21,000 gas. A token swap might cost 150,000 to 200,000. A complex DeFi transaction touching half a dozen contracts can hit a million.

The unit of gas is arbitrary — it measures work, not money. What converts gas into a fee is the gas price, denominated in gwei (one billionth of an ETH). Your total fee is roughly gas used times gas price. If your swap uses 150,000 gas at 30 gwei, you pay 0.0045 ETH.

Gas is the reason a complicated contract call costs more than a vanilla send, and it is also why poorly optimized contracts are measurably worse to use than well-optimized ones. Smart-contract engineers spend a lot of time shaving gas.

The EIP-1559 fee market

Since August 2021, Ethereum has used a two-part fee structure introduced in EIP-1559. Each transaction includes a base fee, which is set by the protocol and burned, and a priority fee (a tip), which goes to the validator as incentive to include the transaction.

The base fee adjusts automatically block by block. If the last block was more than half full, the base fee rises; if less than half full, it falls. The system targets 50-percent block utilization and responds quickly to demand shocks. This means fees are more predictable than they used to be: during calm periods the base fee drifts low, and during NFT mints or market panics it climbs within a few blocks.

The burn matters economically. When network usage is high enough, more ETH gets burned per day than is issued to validators, making ETH slightly deflationary. That relationship — activity translating into supply pressure — is part of why post-Merge ETH is described as "ultra sound money" by its fans.

Why fees sometimes spike

Fees do not correlate with transaction size in a straight line. They correlate with block space demand. A highly anticipated NFT mint can drive gas prices into the hundreds of gwei; a quiet weekend can leave them in the single digits. The Ethereum mempool is a live market for block space, and wallets estimate fees by watching it.

When the user signs a transaction, they authorize a max fee per gas — the ceiling they are willing to pay. If network conditions are below that, they pay only what the current base fee plus their tip works out to. If the max is too low, the transaction sits in the mempool until conditions improve, or drops out.

This is why "stuck" transactions happen. The sender set a low max fee, the network got busy, and the transaction did not get included. Most wallets let the user resubmit with a higher fee to replace it (a "speed up") or cancel it.

Rollups and the cheap seats

Layer-2 rollups (Arbitrum, Base, Optimism, zkSync, and others — see the rollup explainer) exist in large part to make gas affordable. They batch thousands of transactions into one cheap post to Ethereum, paying mainnet fees collectively. A swap on Base or Arbitrum costs a few cents, versus a few dollars on Ethereum itself.

Since the Dencun upgrade introduced blob data in March 2024, rollup fees have fallen by another order of magnitude. For most users, the practical answer to "Ethereum is too expensive" is now "you are on the wrong layer."

Why it matters

Gas is the thermostat of the Ethereum ecosystem. When it is cheap, activity expands; when it is expensive, users migrate to rollups or wait. Understanding that the fee is priced computation — not a flat transaction charge — is what makes the rest of on-chain behavior legible. It explains why apps batch operations, why sophisticated traders use flashbots to avoid the public mempool, and why a failed transaction still costs money: the network did the work whether or not the outcome was what you wanted.