What is a gas fee — and why is the one you actually pay set by your exchange?
Two things about gas fees are widely believed and both are out of date. The first is that Ethereum is expensive: today a token transfer on it costs about five cents. The second is that the fee you see on your withdrawal screen is the gas fee. It is not. It is a price your exchange chose, and picking the wrong line of that dropdown can cost 150 times the right one. This page covers what gas actually is, what a transfer really costs right now, why the exchange’s number is different, and the one comparison that tells you whether a transfer is worth making at all.
KEY TAKEAWAYS
- Gas is a unit of work. Your fee is gas used × gas price, and the amount you are sending is not in that formula anywhere.
- At the 0.374 gwei quoted on 17 Sep 2026 with ether near $1,883, a 65,000-gas token transfer costs about $0.05 — and about $12.24 if gas spikes to 100 gwei.
- A withdrawal fee is a flat price the exchange sets. On Binance’s own table on 11 Sep 2026, USDT cost 0.01 over BNB Smart Chain and 1.5 over Tron — 150 times, one dropdown apart.
- That price has come unglued from the chain. Solana, one of the cheapest chains alive, was charged at roughly 500× its own cost; BNB Smart Chain at less than cost.
- Compare the fee to your risk budget, not your position. On a $500 position risking 1%, a $3.00 Tron round trip is 60% of everything you were willing to lose — and 150% on a $200 position.
What is a gas fee, in one paragraph?
Every blockchain has a hard limit on how much work fits in each block. Because the room is scarce and the demand is not, the room is auctioned. Gas is the unit the work is measured in: a plain coin transfer is 21,000 gas on Ethereum, a token transfer roughly 65,000, a decentralised-exchange swap several times that. Your fee is simply the gas your transaction consumes multiplied by the gas price bid at that moment, quoted in gwei — a billionth of an ether. Other chains use other words. Tron charges energy and bandwidth, Solana charges a tiny base fee plus an optional priority tip. The vocabulary changes; the economics do not. You are renting a slot in the next block, and the rent floats with how many other people want one.
Since Ethereum’s 2021 fee reform, that rent splits in two. A base fee rises and falls automatically with how full recent blocks were, and it is destroyed rather than paid to anyone. On top of it sits a priority fee, a tip you choose, which is what actually persuades a validator to put you near the front of the queue. When people talk about gas “going up”, they almost always mean the base fee — the crowd, not the tip.
The important consequence is buried in that description: nothing in it refers to the amount being moved. The chain does not know or care whether the transfer field says 50 or 500,000. It is selling space.
Why doesn’t the fee depend on how much you send?
Because both transactions are the same job. Sending 50 USDT and sending 500,000 USDT call the identical contract function, change the identical two balance entries, and occupy the identical room in the block. Around 65,000 gas each. The number in the amount field is just a number in a field.
This is the single most useful thing to understand about gas, and it has an uncomfortable implication. A fee that does not scale with size is a fixed cost, and fixed costs are regressive. The same five cents is 0.1% of a $50 transfer and 0.00001% of a $500,000 one. Nobody notices the fee on the large transfer because it is genuinely irrelevant there. The small account, which can least afford it, pays the highest effective rate — and that pattern repeats at every layer of this business, from withdrawal fees to minimum order sizes.
It also explains a habit worth building: move money in fewer, larger transfers. Four transfers of $250 cost four fees. One transfer of $1,000 costs one. Nothing else about the two routes differs.
What does a transfer actually cost right now?
Here is the arithmetic in full, so you can redo it with tomorrow’s numbers instead of trusting today’s. Cost in ether = gas used × gas price in gwei × 10−9; cost in dollars = that, times the ether price. Take a token transfer at 65,000 gas, and the gas price quoted on Etherscan on 17 September 2026, 0.374 gwei, with ether around $1,883:
| Gas price | Working | In ether | In dollars |
|---|---|---|---|
| 0.374 gwei (quoted 17 Sep 2026) | 65,000 × 0.374 × 10−9 | 0.0000243 | $0.05 |
| 2 gwei (an ordinary calm day) | 65,000 × 2 × 10−9 | 0.000130 | $0.25 |
| 10 gwei (busy) | 65,000 × 10 × 10−9 | 0.000650 | $1.22 |
| 30 gwei (crowded) | 65,000 × 30 × 10−9 | 0.00195 | $3.67 |
| 100 gwei (a genuine spike) | 65,000 × 100 × 10−9 | 0.00650 | $12.24 |
Two readings come out of that table. The first is that Ethereum’s reputation for being expensive is several years stale. The fee structure changed, then the data-availability upgrade of 2024 cut costs again by roughly 95%, and mainnet has spent long stretches of 2026 quoting a fraction of a gwei. A five-cent transfer is not a rounding error in anyone’s favour — it is a different network from the one people remember.
The second is that the range is 267 times wide, and the top of it is not hypothetical. Gas spikes exactly when everyone wants to move at once, which is to say during the market events that made you want to move in the first place. Any plan that depends on a cheap transfer is a plan that fails on the day it matters.
One more thing this arithmetic hides: a failed transaction still costs gas. If your swap reverts because the price moved past your slippage limit, the work was still done and the fee is still taken. You paid for the room, and the room does not care what happened in it.
Then why is your exchange charging so much more?
Because a withdrawal fee is not a pass-through of what the chain charged. It is a flat price the exchange sets, per token and per network, and it moves on the exchange’s schedule rather than the network’s. We read Binance’s own USDT withdrawal table on 11 September 2026 and published all nineteen networks. Set four of those lines beside what the underlying chains actually cost and the connection between the two numbers falls apart:
| Network | Roughly what the chain charges | Binance fee, 11 Sep 2026 | Ratio |
|---|---|---|---|
| Solana | ~$0.0006 | 0.3 USDT | ~500× the chain cost |
| Ethereum (ERC-20) | ~$0.05 at 0.374 gwei | 0.3 USDT | ~6.6× |
| Tron (TRC-20) | ~$1.60–2.20 burned without staked energy | 1.5 USDT | roughly at cost |
| BNB Smart Chain (BEP-20) | a few cents | 0.01 USDT | below cost |
Read the first and last rows together. Solana is one of the cheapest chains in existence and is charged at roughly five hundred times what it costs. BNB Smart Chain is charged at a fraction of a cent, which cannot be covering the gas. Neither number is dishonest — a flat fee is a product decision, some networks subsidise others, and exchanges stake energy in bulk on chains like Tron to buy at a price you cannot get. But it does mean one thing plainly: the chain’s cost and your cost have come unglued from each other.
Which produces the spread that matters. In Binance’s own table on that date, the same USDT cost 0.01 over BNB Smart Chain and 1.5 over Tron — 150 times more, for an identical balance arriving at the other end. That 150× is not a difference between exchanges, or between a good decision and a bad one. It is one dropdown on one screen.
Should you compare the fee to your position, or to your risk?
Almost everyone compares it to the position. “1.5 USDT on a $500 transfer is 0.3% — who cares.” That denominator is wrong, and it is wrong in a way that quietly ruins small accounts.
You never risk the position. You risk the distance to your stop. If you follow the ordinary 1–2% rule, a $500 position at 1% risk puts $5 at stake — that is the whole budget, the entire amount the trade is permitted to lose. Now put a Tron round trip beside it: 1.5 USDT in, 1.5 USDT out, $3.00. That is 60% of the budget, spent before the trade has an opinion about anything.
| Position | Risk at 1% | Tron round trip $3.00 | Same trip over BEP-20, $0.02 |
|---|---|---|---|
| $200 | $2.00 | 150% of risk | 1.0% of risk |
| $500 | $5.00 | 60% of risk | 0.4% of risk |
| $1,000 | $10.00 | 30% of risk | 0.2% of risk |
| $5,000 | $50.00 | 6% of risk | 0.04% of risk |
| $20,000 | $200.00 | 1.5% of risk | 0.01% of risk |
Read the first row again. On a $200 position at 1% risk, the Tron round trip costs 150% of the risk budget — more than the trade was ever allowed to lose. There is no entry, no stop placement and no market view that repairs that; the trade was decided before it was taken. And the right-hand column is the same money on the same day through a different line of the same dropdown.
If you want a threshold rather than a feeling, here is ours and the working is one line. Call a transfer negligible when the round trip is under 5% of the risk budget. At $3.00 that needs a risk budget of $60, which at 1% risk means a position of about $6,000. Below that, on that network, transferring is a real cost that belongs in the decision. Note what the rule does not say: it does not say trade bigger. It says transfer less and choose the line deliberately — move once, trade many times from the same place, and stop shuttling money between venues for small differences.
Which network should you actually pick?
In this order, and the first question outranks every other consideration:
- Does the destination support that exact network? Not the coin — the network. Same ticker, different chain, different contract, and a mismatch usually means the money is gone. This question is not negotiable and is not a cost question.
- What does your exchange charge for it? Read the posted fee on the withdrawal screen for two or three supported networks. The spread between them is typically larger than anything else on this page.
- Is the destination congested or paused? Exchanges suspend networks during upgrades and incidents. A suspended network is a stuck withdrawal, not a lost one, but it is a wait you did not plan.
- Do you actually know this chain? A network you have used before, at a fee two cents higher, beats an unfamiliar one every time. The downside distribution is not symmetric: the saving is cents, the mistake is the whole amount.
And the step most people skip: send a small test transfer first when the route is new. Yes, you pay the fee twice. Given section five, you now know exactly what that insurance costs you, which is the point of knowing the number at all.
SEE IT YOURSELF
The whole of section four is checkable in two minutes and costs nothing. Open the withdrawal page for a stablecoin you hold, change the network selector two or three times, and watch the posted fee move while the chain you are comparing against does not. Then open any public gas tracker in another tab and compare. The gap you are looking at is the exchange’s price, written in its own words:
Referral links — they never change our assessment. Education only; most retail traders lose money.
What a gas fee is NOT
Not a percentage. There is no rate. There is a quantity of work and a price per unit of work, and your transfer size appears in neither.
Not money paid to your exchange. Gas goes to the network — partly destroyed, partly to whoever produces the block. The exchange’s withdrawal fee is a separate charge that merely happens to be denominated in the same units of pain, and section four is the evidence that the two have stopped tracking each other.
Not a purchase of certainty. A priority fee buys you position in the queue, not a guaranteed outcome. Pay a large tip into a volatile block and you can still be reverted, front-run, or simply beaten to the price.
Not refundable when things go wrong. Failed and reverted transactions still burn the gas they consumed. You bought the room; whether the meeting went well is not the landlord’s problem.
Not charged at all on a centralised exchange trade. Buying and selling inside an exchange never touches a chain — it edits rows in that company’s database. Gas appears twice only: on the way in, and on the way out. This follows directly from what a blockchain actually is.
Where this reasoning breaks down
Where should a beginner go from here?
If the withdrawal-fee side interested you most, the full nineteen-network table sits on Binance withdrawal fees by network. Otherwise the natural next step is the layer underneath this one: what a blockchain actually is, which explains why blocks are scarce in the first place and why your deposit waits for confirmations. From there, withdrawing crypto safely turns the network-selection question in section six into a checklist you can follow while tired.
On the money side, the comparison in section five only works if you already know what your risk budget is, which is what the position size calculator and the 1–2% rule exist to settle. And if the fee-versus-cost habit interests you, maker and taker fees and slippage are the two costs that are quietly larger than gas for anyone trading rather than transferring.
FAQ
What is a gas fee in simple terms?
A gas fee is the price of space in the next block. You are renting room for a piece of computation, so the fee is set by how much work your transaction takes and how many other people want room at the same moment. It has nothing to do with how much money you are moving.
Why is the gas fee the same whether I send $50 or $50,000?
Because both transactions run the same code and take the same space. A token transfer costs roughly 65,000 gas whether the amount field says 50 or 50,000. The chain charges for work, not value, which makes the fee a fixed cost — painless on large transfers and punishing on small ones.
How much does an Ethereum transfer cost in 2026?
Far less than its reputation. At the 0.374 gwei gas price quoted on 17 September 2026 and ether near $1,883, a 65,000-gas token transfer works out at about $0.05. The same transfer costs about $1.22 at 10 gwei and about $12.24 if gas spikes to 100 gwei, so the number is a snapshot rather than a constant.
Why is my exchange withdrawal fee different from the gas fee?
Because a withdrawal fee is a flat price the exchange sets per token and per network, not a pass-through of what the chain charged. On Binance’s own USDT table on 11 September 2026, Solana cost 0.3 USDT against a chain cost near $0.0006 — roughly 500 times — while BNB Smart Chain cost 0.01 USDT, less than the gas it consumes. The two numbers are no longer connected.
When is a transfer fee too big to be worth paying?
Compare it to your risk budget, not to your position. On a $500 position risking 1%, the money at stake is $5, so a 1.5 USDT Tron fee each way eats 60% of it. On a $200 position the round trip costs 150% of the risk budget — more than the whole amount you were willing to lose. Below roughly a $6,000 position at 1% risk that round trip is not a rounding error, though the same trip over a 0.01 USDT network costs 1% of the same budget.