A minimum withdrawal exists because every transfer has a floor cost that does not shrink with the amount. Below some size, the fixed charge — the network fee, the venue’s handling, the accounting — becomes a large share of what is being moved, and the venue stops offering the service rather than lose money on it. The number you see is that floor, priced in.

The consequence is easy to miss: on a small withdrawal you are not paying a fee, you are paying a percentage. At $20 the same flat charge that is invisible on a $2,000 transfer can take a tenth of the principal. What follows covers where the floor comes from, what the effective rate looks like at each size, and how to decide between saving up and moving now — including the part of the calculation most people leave out, which is what the balance costs you while it waits.
- A minimum is a cost floor, not a courtesy. It is set where the venue’s fixed cost per payout becomes too large a share of the amount.
- The fee is flat; the percentage is not. The smaller the withdrawal, the more the fixed part dominates, and the effect is steep below a few hundred dollars.
- Dust is a separate concept. On account-based chains it means a balance too small to be worth moving; on Bitcoin it means an output too small to be worth spending.
- Waiting has a price too. Every month you leave a balance on a venue to clear a minimum is a month of counterparty exposure, which is a cost even when nothing goes wrong.
Why a minimum exists at all
A withdrawal is not a database update. It is a transaction that must be built, signed, broadcast, confirmed and reconciled, and on most networks the venue pays for it. The payment is roughly constant per transaction regardless of value, so the cost per dollar moved rises sharply as the amount falls. At some point the transfer costs more to process than it delivers, and the venue declines to process it.
That point is a business decision, not a protocol rule. Different venues set different minimums for the same asset on the same network, and the same venue revises them when network conditions or internal costs change. The figure on the withdrawal page is a current price, not a constant.
The floor cost has several parts
The obvious part is the network fee. The less obvious parts are the hot-wallet operations, the review that a withdrawal may trigger, the reconciliation between the venue’s ledger and the chain, and the fraud cost of small withdrawals, which are disproportionately used to test stolen accounts. A minimum spreads those costs over an amount large enough to absorb them.
On Bitcoin the floor is partly structural. Each unspent output is a discrete thing that must be consumed whole and recreated as change, so a venue paying out many withdrawals builds a transaction with many outputs. That works while the outputs are worth more than the cost of spending them later, and stops working when they are not.
Dust is the physical version of the same problem
Bitcoin nodes apply a relay threshold below which an output is treated as dust and refused by default, because it would cost more to spend than it is worth. The commonly cited figures are 546 satoshis for a legacy P2PKH output and 294 for a native SegWit output, derived from the default dust relay fee rather than fixed in the protocol. The numbers move if the relay policy does.
The arithmetic behind the threshold is the cost of creating the output and later spending it. A P2PKH output is 34 bytes and the input that spends it is about 148 virtual bytes; at the default dust relay rate of 3 satoshis per vbyte, (148 + 34) × 3 comes to roughly 546. A native SegWit output is cheaper on both sides, which is why its threshold is closer to 294.
At 10 satoshis per vbyte, spending a 546-satoshi output costs about 1,480 satoshis — nearly three times what it holds. An output like that is not money you have failed to spend. It is money that cannot be spent at that fee level, which is why venues will not create one for you.
The percentage nobody quotes
Exchanges publish withdrawal fees in absolute terms, which makes them easy to compare across venues and almost impossible to compare against the size of your transfer. The table below converts two hypothetical flat charges into their real cost at five sizes. The charges are illustrative; substitute the numbers on your venue’s page.
| Withdrawal | Cost at a $1.50 fee | As a share | Cost at a $6.00 fee | As a share |
|---|---|---|---|---|
| $20 | $1.50 | 7.50% | $6.00 | 30.00% |
| $100 | $1.50 | 1.50% | $6.00 | 6.00% |
| $500 | $1.50 | 0.30% | $6.00 | 1.20% |
| $2,000 | $1.50 | 0.075% | $6.00 | 0.30% |
| $10,000 | $1.50 | 0.015% | $6.00 | 0.06% |
The shape is the point. Doubling the amount halves the percentage; there is no size at which the absolute cost falls. A $20 withdrawal on a busy L1 at a $6 fee loses thirty cents on the dollar before the money has moved anywhere, and that is before the spread on converting it back if you ever do.
Why the fee looks small until you divide
Six dollars is a coffee. Six dollars on a twenty-dollar withdrawal is a different thing, and the reason it does not feel like one is that the comparison is made in absolute terms against income rather than in relative terms against the transfer. The same instinct that makes a 1% trading fee feel cheap makes a $6 withdrawal fee feel cheap, even when one is 0.3% of the amount and the other is 30%.
The cliff at the minimum
Minimums create a specific trap: a balance just below the threshold. You have $18 and the minimum is $20, so the only way out is to buy more or to leave it. Buying more to clear a minimum is how a $6 fee becomes a $6 fee plus a card fee plus a spread, and the deposit is usually charged at a worse rate than the withdrawal saves.
The alternative — leaving it — is often correct. An $18 balance is not stranded in the sense that it cannot be recovered; it is stranded in the sense that recovering it costs more than it is worth today. Writing it down as lost is more honest than pretending it is still liquid.
Switching networks is not a free fix
Most assets exist on several networks, and the same withdrawal can cost very different amounts depending on which one you choose. That is the obvious lever, and it works — but only if the receiving side supports the network and you intend to hold what arrives.
A cheaper network usually delivers a different representation of the asset: a wrapped or bridged version that the destination chain recognises as its own. If you are moving to another exchange and it supports that network, the saving is real. If you are moving to a wallet and will eventually need the canonical asset, the bridge or swap back is a cost you have moved rather than removed.
What the minimum is really protecting
When a venue raises a minimum during congestion, it is not being punitive. The chain fee has risen, the flat charge no longer covers it, and the alternative to a higher minimum is a withdrawal that loses the venue money on every transaction. Reading the minimum as a signal about network conditions is more useful than reading it as a fee increase.
The internal-move shortcut
Transfers between accounts at the same venue commonly cost nothing, because nothing touches a chain. That does not help you leave, but it does help you consolidate: gathering scattered balances into one account before a single withdrawal avoids paying the fixed charge several times. The mechanics of why that matters are covered in Why Withdrawal Fees Cost More Than Trading Fees, which is about the flat-versus-percentage structure rather than the minimum itself.
Minimums are a venue decision, so they move independently of the network fee you just worked through. If the cliff is close enough to matter, the platform you already hold the balance on is the only one whose schedule counts — the current OKX terms are below for comparison.
Welcome rewards
Up to $200
-
Spot, derivatives and a bundled self-custody wallet -
Clean order entry for limit orders -
Earn products in the same app
Open an OKX account
Ends Nov 30, 2026 · Global (see terms)
Save up or withdraw now
The decision is a comparison between two costs, and only one of them is on the fee page. Withdrawing now pays the fixed charge once and ends your exposure to the venue. Waiting amortises the charge over a larger amount and extends the exposure. Neither is free.
The arithmetic of batching
Batch four $250 withdrawals into one $1,000 withdrawal and you pay one $6 fee instead of four, saving $18. That is the whole case for waiting, and it is a strong one at small sizes. The saving scales with the number of events avoided, not with the amount, which is why the habit of withdrawing every time a trade closes is the expensive one.
The arithmetic of exposure
While the balance sits on the venue, it is an unsecured claim on a business. The expected cost is the balance multiplied by your estimate of the annual probability of a loss event, multiplied by the share you would not recover. None of those inputs can be measured from outside; they are judgements, and they belong in the calculation as judgements rather than as zeros.
A hypothetical: $1,000 held for a year, a 0.5% annual probability of a loss event, and 50% recovery. The expected cost is $1,000 × 0.005 × 0.5 = $2.50. Compare that with the $18 saved by batching and the batching still wins — but the margin is now a number rather than an assumption, and it flips if you hold a larger balance or judge the risk higher.
The order to decide in
- Price the withdrawal: the venue’s flat fee for your asset and network, in dollars.
- Divide by the amount. If the result is above roughly 1%, the fee is the dominant cost of the transfer.
- Check whether a cheaper network is supported at both ends, and whether it delivers an asset you want to hold.
- Estimate the exposure cost of waiting: balance × annual loss probability × unrecovered share.
- Compare the fee saved by waiting with the exposure incurred by waiting. If the saving is larger, wait; if not, move.
- Set a threshold and hold to it, rather than re-deciding at every price move.
A minimum withdrawal does not protect you from a fee. It protects the venue from a transfer that costs more to process than it moves — and it tells you, in a single number, where that line currently sits.
Dust that arrives and cannot leave
Not all dust is created by rounding. Some is sent to you deliberately. Spam tokens are airdropped to wallet addresses in the hope that the holder will try to sell them, interact with a malicious contract, or visit a site to “claim” them. The token has no market, so its balance is a number rather than a value, and moving it can cost more than it will ever be worth.
The mechanics of why these appear, and what to do about them, are covered in Spam Tokens: Why They Appear and What to Do. The relevant point here is economic: a balance below the cost of moving it is not an asset with a small value, it is a liability with a small value, and treating it as part of your portfolio overstates what you hold.
Consolidation is a cost you choose
On Bitcoin, a wallet holding many small outputs will pay more to spend them than one holding a single output of the same total, because each input adds bytes. Consolidating during a quiet period — when fees are low — converts many small outputs into one and reduces the cost of every future spend. Doing it during congestion is the opposite trade.
Rounding is not the main cost
The visible rounding on a withdrawal is usually a few cents of precision. The real rounding cost is structural: the fixed charge that makes small transfers expensive, the minimum that forces a top-up, and the dust that cannot be moved at all. Those three are the same problem seen from different angles, and they all scale against the small holder.
What to check before you hit withdraw
The fee page is the starting point, not the answer. Three checks turn it into a real number: the fee for your specific asset and network, the minimum for that same pair, and whether the destination supports the network you chose. A cheap route the receiver cannot process is a support ticket, and the first transfer is worth treating as a test — a practice set out in The Test Transfer: Why Small First.
Then do the division. Almost every bad withdrawal decision comes from comparing an absolute fee to a balance without converting it into a share. The number that matters is not what the withdrawal costs; it is what fraction of the money it removes.
0 comments
No comments yet — ask a question and we will answer it here.