When a gas fee is telling you something is wrong

You are about to confirm a swap and the network fee is 40 dollars on a transaction that cost two dollars last week. Nothing on the screen explains it. A gas fee is the price of the computation your transaction asks the network to perform, priced by auction, and reading it is one of the few checks available before you sign.

Most of the time a high fee means the network is busy and nothing more. Sometimes it means the transaction is doing more than you think it is. Telling those apart takes about thirty seconds and is worth doing on anything expensive.

What the number is actually made of

Two things multiply together, and only one of them is about you.

The gas limit is how much computation the transaction needs. A plain transfer is 21,000 units and always has been. A token transfer is nearer 65,000. A swap through a router that touches three contracts can be 200,000 or more.

The gas price is what the network is charging per unit right now. It moves with demand, minute to minute, and it is the same for everyone competing for the same block.

Your gas fee is those two multiplied. That matters because the two have completely different meanings:

• A high gas price with a normal limit is congestion. Wait, and it falls.
• A normal gas price with a high limit is a complicated transaction. Ask why.
• Both high is a complicated transaction at a bad moment.

The fee is not a fraud signal on its own. It is a signal about how much work you are authorising, and that is a question worth asking.

When is a gas fee actually telling you something?

Three situations where the number is the clue rather than the cost.

The estimate is far above what the action implies. Sending one token should not need 400,000 units. When a wallet estimates that much for something described as a transfer, the contract is doing more than transferring — looping over a list, writing storage for every holder, or calling out to another contract. That is not automatically malicious. It is automatically worth reading the transaction before signing.

The wallet cannot estimate it at all. When a wallet shows a warning that the transaction may fail, it usually means the simulation reverted. People raise the limit manually to push it through. That is backwards: a reverting simulation is the network telling you the transaction does not work, and forcing it through burns the fee and changes nothing.

The gas fee is a large fraction of the amount. Paying 30 dollars of gas to move 40 dollars of tokens is a bad trade you can simply decline. This sounds obvious and is the most commonly ignored one, because the fee appears at the end, after the decision feels made.

The fee you pay when nothing happens

A failed transaction still costs a gas fee. The network performed the computation up to the point it reverted, and that work is paid for.

This is the mechanism behind a specific kind of loss. A contract that is designed to fail after consuming most of the gas limit collects nothing itself — but if you retry three times at a raised limit, you have paid three times for nothing. Anyone who has tried to force a stubborn transaction through has done a smaller version of this.

The rule that avoids it: when a transaction reverts twice, stop and find out why. The reason is on the block explorer, in plain words, next to the failed transaction. Common causes are slippage, an expired deadline, or an allowance that was never set — and all three are fixed by changing something, not by paying more.

Does a low fee mean it is safe?

No, and this is the more dangerous direction. The most expensive signatures in Web3 cost almost nothing in gas.

Approving a token costs about 45,000 units — a few cents on most networks. That single cheap transaction can grant a contract permission to move your entire balance of that token, for ever, which is the mechanism behind most drainer losses. The relationship between what a transaction costs and what it can do is close to nonexistent.

Signing a message costs zero gas, because nothing goes on chain. An off-chain signature can still authorise a transfer through a permit, and it appears in no fee estimate at all because there is no transaction to estimate. Our note on transaction safety covers what those messages actually say.

So the honest framing is this: a gas fee tells you how much computation you are buying. It tells you nothing about what permission you are granting. Those are separate questions and the second one matters more — see token approvals for how to read it.

Reading the fee before you sign

A short routine, in the order that catches the most.

1. Look at the limit, not just the total. The total moves with the market; the limit is about your transaction.
2. Compare it to the action. Transfer, approve, swap and mint have rough shapes — 21k, 45k, 150k+, varies. A number far outside that range is a question.
3. Check whether the wallet simulated it. A simulation that succeeded is real evidence. One that failed is a stop sign.
4. Check the fraction. Fee against value moved. Decline anything absurd.
5. Then read the permission, which is the part gas cannot tell you about.

Steps 1 and 2 take seconds once you know the shapes. Step 5 is the one that actually protects you, and it is the one people skip because the fee felt like the check they had already done.

What the fee looks like on a busy day

Worth having a sense of scale, because "high" is relative and the relative part is what makes people accept a bad number.

On a quiet Ethereum block, a simple transfer might cost a fraction of a dollar. During a heavily contested mint it can be 50 times that, for the identical 21,000 units of work. Nothing about your transaction changed — you are bidding against more people for the same space.

Layer 2 networks change the arithmetic rather than the principle. Fees are much smaller because the computation is settled in batches, but the same relationship holds: limit times price, and the limit still tells you how much work you authorised. A small gas fee deserves the same reading, and gets it less often precisely because the number is small enough to ignore.

The practical consequence is that waiting is usually free. Unless you are competing for something with a deadline, a transaction that costs too much right now costs less in an hour, and nothing about it expires meanwhile. Our risk analysis guides cover the cases where waiting is the wrong call.

For the underlying mechanics, the fee market is specified in EIP-1559, which is worth reading once if you want to understand why the number moves the way it does.