The price of a stablecoin payout is the sum of three separate costs that most quotes present as one number: the spread between the rate you get and the mid-market rate, the fee for the local rail that delivers the money, and the cost of moving the stablecoin on the blockchain it lives on. A provider's page can be honest about all three and still be hard to compare with another provider's page, because each one chooses which of the three to make visible and which to fold into the rate.
This guide takes the payout apart so a finance team can compare offers line by line, and so the reconciliation entry at month-end explains itself. What the API exposes to make that possible is the subject of what a stablecoin API is. The mechanics of the payout itself, from a stablecoin balance to a bank account, are in what a stablecoin off-ramp is; this post is only about what it costs and why.
The three costs, and which one is usually the largest
Start with the shape of a payout. A business holds USDC, wants a supplier in Brazil paid in reais, and sends the instruction. Three things happen and each has a price.
The conversion has a spread. Somewhere a market maker exchanges USDC for reais, and the rate at which that happens sits some distance from the mid-market reference. That distance, expressed in basis points, is the spread, and on most corridors it is the largest of the three costs by a wide margin. It is also the one most often hidden, because a provider can quote "no fee" and earn entirely on the rate.
The rail has a fee. Pix (Brazil's instant payment system, run by the Central Bank) costs the sending institution something per transfer, SPEI (Mexico's interbank transfer system, run by Banxico) costs something else, and a SWIFT wire costs far more than either and may cost again at an intermediary bank the sender never chose. Rail fees are usually flat per payment, which means they matter on small payouts and disappear into the spread on large ones.
The blockchain has a cost. Moving USDC on Ethereum is paid in the network's own unit, and the price moves with congestion; moving the same USDC on Polygon, Base or Tron costs a small fraction of that. The public reference for Ethereum is a live tracker such as Etherscan's gas tracker, which is enough to show that the number is a variable and not a constant. A provider that absorbs this cost into a flat fee is taking a small risk on your behalf; one that passes it through is giving you a number that changes hour to hour.
Where the spread hides in a quote
A quoted rate has two components a buyer should learn to separate: the reference and the markup.
The reference is what the provider calls mid-market, and providers do not all use the same one. Some take a published interbank fix, some take the mid of their own liquidity, some take a stablecoin exchange price that already includes the market's own premium or discount against the dollar. Two providers quoting "20 basis points over mid" can therefore land on different rates. Ask which reference a provider uses and at what time it is sampled.
The markup is the provider's margin on the conversion, and it is where volume tiers live. Most pricing models step the markup down as monthly transacted volume rises, with the tier reassessed on trailing volume. What varies is whether the step-down is automatic or negotiated, whether it applies from the first dollar of the new month or only after the threshold is crossed, and whether tiers are per currency or aggregated. A model that tiers per currency rewards concentration in one corridor; a model that aggregates rewards breadth.
The cleanest test is to ask for both numbers on the same payout: the reference rate at the moment of the quote and the rate you will actually receive. The difference is the all-in spread for that transaction, and it is the figure to put in the comparison sheet. A provider that cannot produce the two numbers separately has told you something about how its pricing is built.
Floating and locked rates are different products with different prices
A floating rate is a reference at the moment of the quote that may move by the time the payout executes. It is the right instrument for a flow that executes immediately and where a small variance is acceptable, and it normally carries no fee beyond the spread.
A locked rate holds a quoted rate for a fixed window and is priced accordingly. The window matters: a lock of thirty seconds is enough for an automated flow to receive a quote and submit the payout, and can be offered without a fee; a lock of one minute or five minutes gives a human time to review and confirm, and the provider carries the volatility for that time, so it charges for it. The fee is small in absolute terms, and it is a real cost that a comparison should count.
The choice is not a pricing choice only. A locked rate is what lets a business show a supplier the exact amount that will arrive, and what lets a marketplace promise a seller a fixed payout. The premium buys a number that can be written on an invoice. Whether it is worth paying depends on whether your customer needs that number before the payout runs, and the answer differs across the flows described in what a stablecoin corridor is.
Rail fees: flat, per rail, and where the exceptions are
Rail fees are the easiest cost to read because they are usually published as a flat amount per payment and per rail. The comparison is straightforward on the happy path and misleading on the exceptions.
The exceptions are returns and investigations. An ACH payout that is returned by the receiving bank, a wire that triggers an investigation at an intermediary, a SEPA transfer that is recalled: each of these can carry a handling fee that does not appear in the per-payment price. On a program with a low return rate they are a rounding error; on a program paying out to a long tail of new destinations, where some fraction of bank details will be wrong, they are a line item. Ask for the exception schedule, and ask what the return rate looks like for flows like yours.
Cut-off times are a cost too, though not one that shows on a price list. A rail that settles same-day only until a cut-off pushes a late payout to the next business day. Pix and SPEI run around the clock, so a payout over them at eleven at night lands at eleven at night.
Network cost, and why the chain is a pricing decision
The blockchain leg is the cost most often ignored by finance teams and most often obsessed over by engineers, and both are wrong in the same way: it is small on most payouts and it is a variable.
Two decisions set it. The first is which network the stablecoin moves on. USDC and USDT exist on several, and the cost per transfer differs by orders of magnitude between Ethereum and the alternatives. A provider that lets the integrator choose the network per transaction, or set a default, is handing over a lever; one that fixes the network is making the decision for you and pricing it in.
The second is whether the provider charges the network cost as incurred or as a flat allowance. A flat allowance is simpler to reconcile and slightly more expensive on average, because the provider prices in the variance; a pass-through is cheaper on average and produces a different number on every payout. For a treasury flow of a few large payouts a month, either is fine. For a payout program of thousands of small transfers, the flat model is easier to explain to the finance team and the pass-through is easier to explain to the engineering team, and the finance team is the one that has to close the books.
How partner fees change the number your customers see
A platform that builds a product on top of a provider's API does not pay the provider's price; it pays the provider's price and then sets its own. Partner fees are the mechanism: the platform adds a rate in basis points, a flat amount per transaction, or both, on top of the provider's pricing, and the provider collects the fee from the end customer and settles it to the platform.
This changes the reading of every number above. The spread your customer experiences is the provider's spread plus your markup, and the choice of where to put your margin, in the rate or in a visible fee, is the same choice the provider faced one layer down. Putting it in the rate is invisible and harder to explain when a customer compares against a reference; putting it in a fee is visible and easier to justify. Many platforms use both.
The practical detail is when the partner fee attaches. On a locked rate it has to be part of the quote, because the quote is the promise; on a floating rate it can attach at execution. A platform that quotes a customer a locked rate and adds its fee afterwards has quoted the wrong number, and the customer will notice.
Reading a pricing page, and the questions that expose the difference
Pricing pages for stablecoin infrastructure fall into two families. One publishes a table: a rate per currency pair, a fee per rail, a fee per account or wallet, and lets a buyer compute. The other publishes a model and asks for a conversation: usage-based, tiered to volume, with a setup fee and a monthly minimum, and the actual rates come in a proposal.
Neither family is more honest than the other. The published table is easier to compare and tends to describe a self-serve product with the pricing of a self-serve product; the model-and-proposal form describes a contract with commitments on both sides, and the monthly minimum in it is the provider's way of saying that it prices the relationship and not the transaction. A buyer with real volume usually pays less under the second form; a buyer testing an idea usually pays less under the first.
Five questions expose the difference between two offers faster than a spreadsheet does. Which reference rate, sampled when. Is the spread tiered, on what volume, reassessed how often, per currency or aggregated. What is the fee schedule for returns and investigations. Is the network cost passed through or flat, and who chooses the network. And what is in the monthly minimum, meaning whether it is a commitment against which fees count or a floor charged regardless.
When stablecoin payout pricing is the wrong thing to optimize
There are flows where the per-payout price is not the number that matters and optimizing it is a distraction.
If the alternative is a correspondent-banking wire, the comparison is not between two spreads but between a spread and an unknown. A SWIFT payment into Latin America carries the sending bank's fee, an intermediary's fee that is deducted on the way and not disclosed in advance, the receiving bank's fee, and a conversion at the receiving bank's rate, which the sender never sees quoted. Against that, any stablecoin payout with a published spread is already a different category of cost, and shaving basis points off it matters less than the fact that the number is known before the money leaves.
If the volume is small and irregular, the setup fee and the monthly minimum dominate everything else, and the right comparison is between total monthly cost at your actual volume, not between rates. And if the payout is a treasury movement of a large amount a few times a year, a locked rate and a slower rail are worth more than a lower spread, because certainty on a large number is worth more than a few basis points on it.
How the costs show up on a Lumx payout
Lumx is stablecoin payments infrastructure for businesses that move money between Latin America and the rest of the world: one API to collect, hold, convert, and pay out in BRL, MXN, COP, USD, EUR, and GBP or in USDC and USDT, over local rails such as PIX, SPEI, PSE, ACH, FEDWIRE, SEPA, and Faster Payments, with SWIFT and on-behalf-of payments and collections (POBO and COBO) in USD, EUR, and GBP, plus named virtual accounts, custodial wallets, and KYB/KYC built in.
The pricing model is usage-based and tiered to volume, and the components map onto the three costs above. Conversion is priced per currency as a spread over the mid-market reference, with the tier set by monthly transacted volume and reassessed quarterly on trailing volume, upgrades applied automatically. Rails carry a flat fee per payment, per rail, with returns and wire investigations on their own exception schedule. The blockchain leg is chosen per transaction or by project default across Ethereum, Polygon, Base, Tron and, for on-ramps, Stellar, so an integrator paying suppliers in Brazil over USDC to BRL can keep the network cost small by not defaulting to Ethereum. Named accounts, custody and verification are priced as their own lines, and a standard agreement carries a setup fee and a monthly minimum that a ramp-up period waives for the first months. The rates themselves come in a proposal, because the tier depends on volume; the structure is published so the proposal can be read.
Two mechanics matter for reading the numbers on a payout. Every exchange rate request returns both the base rate and the final rate on the same response, so the all-in spread for that transaction is one subtraction, and the locked variant holds the quote for thirty seconds without a fee or for one or five minutes with one, per the exchange rates documentation. And a partner fee, in basis points or flat or both, is configured once, applied by reference on each transaction, included in the locked quote when the quote is requested, and settled to the platform's own wallet. The receipt on every completed payout carries the amounts, the rate and the fees, which is what makes the reconciliation entry in global payments write itself rather than require a spreadsheet.
The position I hold on pricing, and the one I argue for in every proposal we send, is that a spread the customer can verify beats a lower spread the customer has to take on faith. We return the base rate next to the final rate on purpose, knowing it invites comparison, because a finance team that can compute our margin from our own response is a finance team that stops asking whether there is a hidden one. The number of prospects who have used that transparency to negotiate is smaller than the number who have used it to sign.
Verified on September 25, 2026. Operational context, not legal, tax, or investment advice.
Cover photo: Frantisek Duris on Unsplash.
What is the biggest cost in a stablecoin payout?
On most corridors it is the FX spread, the distance between the rate you receive and the mid-market reference. Rail fees are flat per payment and matter mainly on small payouts, and the blockchain cost is small on every network except Ethereum during congestion. Compare offers on the all-in spread first.
Is a locked exchange rate worth the extra fee?
It is when your customer needs to know the exact amount before the payout runs, for example to put it on an invoice or to promise a seller a fixed payout. Short locks of around thirty seconds are often free and enough for automated flows. Longer locks carry a fee because the provider holds the volatility for that window.
How do volume tiers work?
The spread steps down as monthly transacted volume rises, and the tier is usually reassessed on trailing volume at a fixed cadence. The details that change the number are whether the step-down is automatic, whether it counts volume per currency or in aggregate, and when in the month a new tier starts to apply.
Can I add my own fee on top of the provider's pricing?
Yes, through a partner fee. You set a rate in basis points, a flat amount per transaction, or both, the provider collects it from your customer and settles it to you. On a locked rate the partner fee has to be included when the quote is requested so that the number the customer sees is the number they pay.
