A cross-border payment in crypto means moving a stablecoin like USDT (Tether) across a border instead of routing a bank wire through a chain of correspondent banks. This breakdown is for the finance person paying suppliers or contractors abroad. You will see, line by line, where a $10,000 payment actually goes — and where cashing out to local currency quietly eats the saving that every headline promises.
Most articles on this topic put a single number in the title, “$10–30 per transfer,” and stop there. That number is real for one leg of the payment and misleading for the whole. The on-chain transfer is nearly free. The cost that decides whether crypto is cheaper sits somewhere else, and the field tends to leave it out.
Why a cross-border bank transfer costs more than the fee on your statement
A $10,000 international wire rarely costs the $45 your bank prints on the confirmation. Three charges stack, and the largest is never itemized: the foreign exchange spread. Published bank schedules and World Bank data both point to the same pattern, a small visible fee sitting on top of a large hidden one.
The World Bank’s Remittance Prices Worldwide (Issue 54, September 2025) recorded a global average cost of 6.36% for sending $200, rising to 14.99% through banks, the most expensive channel. Those are retail remittance figures, not a B2B wire rate, so treat them as an anchor rather than a quote: traditional rails are expensive, banks are the worst of them, and the spread is the part you cannot see.
For a business wire the structure breaks into three pieces. The outbound fee runs $35–50 (NerdWallet, 2026). Correspondent banks along the SWIFT route each deduct a “lifting fee” of roughly $15–50 from the amount in transit (Airwallex, 2026). Then the FX markup: banks commonly quote a rate 1–3% worse than the mid-market rate and keep the difference (Wise, 2026 guidance). On $10,000, a 2% markup alone is $200. The fee is the rounding error.
The reason the spread stays hidden is structural, not a trick by any one bank. In the correspondent model, each bank in the chain is a separate business pricing its own leg, and no single party sees the total cost end to end. The sending bank does not know what the correspondent will charge, and the correspondent does not know the receiving bank’s fee (Sphere, 2026). The Bank of England has made the same observation from the other direction: the more correspondent banks in a chain, the slower and more expensive the payment becomes, with a fee deducted at each stage (via Demivolt, 2026). A worked example from Corpay (2026) makes it concrete: you send $10,000, your supplier sees $9,930, and the missing $70 went to banks you never chose and cannot find on your statement.
One more figure reframes the whole comparison. Retail and small-business cross-border payments run 1.5–6%, while high-volume wholesale B2B flows average around 0.1%, because scale and negotiating power buy a better rate (IMF 2025 data, via Demivolt). A mid-market company paying a $10,000 invoice sits in the first band, not the second, so the fee-plus-spread math below is the one that applies to it.
The three legs of a cross-border payment
Every cross-border payment has three legs: send, move value, cash out. The cost hides in a different leg depending on the rail, so naming them separately is the only honest way to compare a bank wire against a stablecoin transfer.
- Send. Initiating the payment. A bank charges an outbound wire fee here; a crypto payment gateway charges nothing to start.
- Move value. The actual crossing of the border. For a bank this is the correspondent chain plus the FX conversion. For crypto this is the on-chain transfer, which costs cents.
- Cash out. Converting to the currency the recipient will spend. A bank bakes this into the FX spread. A stablecoin recipient faces it as a separate, optional step, the off-ramp to local fiat.
Here is the sleight of hand in most comparisons: they price only the “move value” leg for crypto, the near-free on-chain transfer, and set it against the full cost of a bank. An honest comparison prices all three legs on both sides. Do that, and the picture changes.
One $10,000 payment, line by line
On a $10,000 cross-currency invoice, a bank wire costs roughly $135–420, almost all of it the invisible FX spread. The same payment in USDT costs cents on-chain plus a gateway fee of 0.5% — but only if the recipient keeps the stablecoin. Convert to local fiat, and a variable off-ramp cost narrows the gap.
| Leg | Bank wire (SWIFT) | USDT via Speend |
|---|---|---|
| Send | $35–50 outbound fee | no connection fee |
| Move value | correspondent deductions $15–70 + FX spread 1–3% ($100–300) | on-chain fee: cents, network-dependent |
| Gateway | — | from 0.5% = $50 on $10,000 |
| Cash out to local fiat | included in the FX spread above | variable, provider-dependent (off-ramp) |
| What you think you paid | ~$45 | “$10–30” (the field’s headline) |
| What it really costs | ~$135–420 | on-chain cents + 0.5% + off-ramp |
Read the crypto column carefully, because it splits in two.
Keep the balance in stablecoin, and the cost stops at the on-chain fee plus the gateway’s 0.5%, about $50 on $10,000, against $135–420 by wire. The saving is real and large. It is also not “free”: the on-chain leg is nearly free, but 0.5% is the floor a gateway charges, and $50 is not zero.
Convert and cash out to local currency, and two more costs appear. Speend’s auto-conversion to a different payout currency adds 1% where it applies, another $100 on this payment. On top of that sits the off-ramp to fiat, priced per jurisdiction and not published. That last line is the one the “$10–30” headline never shows.
The symmetry is the point. A bank buries its biggest cost in the FX spread; a crypto comparison buries its biggest variable in the off-ramp. Both rails quote you the small visible number and stay quiet about the large invisible one.
Reproduce this for your own corridor
The math above is not proprietary. To run it for the payment in front of you, fill four lines and compare the totals:
- Visible bank fee. Read it off your bank’s fee schedule for an outbound international wire, usually $35–50.
- FX spread. Compare the rate your bank quotes against the mid-market rate on the day, then multiply the gap by the amount. At 1–3% on $10,000 that is $100–300, and it is the number that decides the comparison.
- On-chain fee. Check a live network fee for the stablecoin and network you would use (see the next section). On most networks it is cents.
- Off-ramp. Ask your gateway what converting to your payout currency and withdrawing to a local account costs. If the answer is “keep it in stablecoin,” this line is zero.
If line 4 is smaller than line 2, crypto is cheaper for you. If cashing out costs more than a bank’s FX spread, it is not, and the honest place to find that out is before you switch a supplier over, not after.
The off-ramp: where the saving disappears
The saving survives if the recipient keeps the payment in stablecoin, and it shrinks the moment they convert to local currency. The on-chain transfer is nearly free; the exit to fiat is a separate, priced step, and it decides whether crypto actually beat the wire.
With Speend, a payment settles in crypto on the merchant’s balance by default, and withdrawal happens on demand: to a self-custody wallet, an exchange, or a supported bank channel. The funds credited to that balance stay the merchant’s own; the balance sits with the provider until the merchant moves it.
That structure creates two clean cases, and which one you are in depends on the business, not the rail.
Holding the balance in stablecoin makes sense when the money does not need to become local currency right away. A business that pays its own overseas suppliers in USDT, keeps part of its treasury in stablecoin, or runs recurring B2B payments can receive, hold, and spend without ever touching the off-ramp. For that business, the cost genuinely stops at the gateway fee, and most of the theoretical saving is real.
Cashing out is unavoidable when local currency has to land in a local account, for payroll, tax, or domestic operating costs. Speend routes fiat through partner banking channels in select jurisdictions, and the list of those channels is not public. The cost of that route is variable and provider-specific, which is exactly why it belongs on its own line rather than folded into “almost free.” A business that ends every payment in a conversion should price that line first and treat the on-chain “cents” as the smallest part of the story.
Put the two endpoints in dollars on the same $10,000. Held in stablecoin, the payment costs about $50, the gateway’s 0.5% plus on-chain cents. Converted to a different payout currency at Speend, it costs about $150 before the off-ramp, the 0.5% plus a 1% conversion, and then the off-ramp to a local bank account on top of that. The bank wire’s $135–420 sits between those two numbers, which is the whole argument: crypto wins clearly on the held payment and turns into a real comparison on the cashed-out one.
Want to accept crypto payments on your website?
Fast setup and KYC/KYB, fee starts from 0.5%
Contact UsThe practical read: if your model lets you hold and spend stablecoins rather than convert on every payment, most of the saving is yours to keep. If every payment ends in a conversion to fiat, price the off-ramp before you assume crypto won.
Which network to send on for a cross-border flow
The network you send USDT on sets the on-chain cost, and the standard advice, “use TRC-20, it is the cheapest,” is out of date as of September 2026. On a direct transfer without staked energy, ERC-20 is currently cheaper than TRC-20.
Measured on 2026-09-04, an ERC-20 USDT transfer cost about $0.015, with a 90th-percentile figure near $0.34, while Ethereum gas held below 1 gwei. A TRC-20 transfer burned roughly $1.05–2.13 in TRX without staked energy. Solana and Polygon came in under $0.01. Public gas trackers through late August 2026 showed the same reversal: ERC-20 near two cents, TRC-20 at a dollar or more on burn.
The catch keeps TRC-20 in the game. Staking TRX for energy in advance drops a TRC-20 transfer toward zero, below ERC-20 again. So the honest rule is narrow: on a plain burn transfer, ERC-20 wins right now; TRC-20 wins only when the sender has staked energy. For a business moving USDT on a schedule, that choice, staking energy or routing over a low-fee network, is a real lever rather than a rounding error.
Cost is not the only axis, though on a cross-border flow it is the loudest. Speend settles USDT on five networks, and each has a practical fit:
- TRON (TRC-20) has the widest exchange acceptance for USDT, which matters if your recipient withdraws to an exchange that credits Tron deposits first. Its true cost depends on energy.
- Ethereum (ERC-20) is the cheapest on burn right now and the default for counterparties who already hold USDT on Ethereum, common in B2B.
- BNB Smart Chain (BEP-20) suits recipients inside the Binance ecosystem.
- Polygon fits low-value, high-frequency flows such as subscription billing.
- Solana offers sub-second finality for high-throughput checkouts.
On-chain costs move fast, so re-check the live figure before you send anything sizeable; the numbers above are a dated snapshot, not a constant.
Speed is a cost too, while the money is in transit
Settlement time is not just convenience; it is exposure. A payment sitting in a correspondent chain for one to five business days is a payment exposed to the exchange rate moving against you the whole time it travels. A stablecoin transfer that finalizes in seconds closes that window almost entirely.
A bank wire over SWIFT typically settles in one to five business days, though newer routing has compressed part of that. A stablecoin transfer finalizes in seconds to a few minutes, depending on the network. For a single invoice the difference is minor. Across a book of supplier payments in volatile currency pairs, the shorter exposure window is a cost the “$10–30 vs $45” framing never captures, and it sits on the crypto side of the ledger, not the bank’s.
What stays on the business’s side
Two things stay with the business whatever rail it picks: where it is registered, and verification. Neither is set by the language on a website or the country you read from; both are settled once, at onboarding.
Speend verifies a business through KYB (Know Your Business), usually within 1–3 business days once documents are in. The jurisdiction where the business is registered is part of that check, decided case by case. A payment provider cannot promise to connect a business from anywhere sight unseen, and the registration jurisdiction is the reason why: it is a KYB question, resolved at onboarding rather than assumed from a marketing page.
That is the whole of it on the business side. Rail choice and network choice are technical; registration and verification are the two facts that travel with the company itself.
FAQ
How much cheaper is crypto than a bank wire?
On a $10,000 payment kept in stablecoin, about 0.5% plus on-chain cents, roughly $50, against a wire’s $135–420. If you convert and cash out to local fiat, add the gateway’s 1% conversion where it applies and a variable off-ramp cost, and the gap narrows. The honest answer depends on whether you hold or cash out.
Is the on-chain transfer really “almost free”?
The transfer itself is, cents on most networks. The gateway fee (from 0.5%) and any off-ramp to fiat are not. “Almost free” describes one leg of the payment, not the payment.
Where exactly does the saving get lost?
At the off-ramp, the conversion from stablecoin to local currency through a bank channel. Keep the balance in stablecoin and the saving stays intact; convert on every payment and part of it goes back.
Which USDT network should I use for a B2B payment?
As of September 2026, on a plain transfer without staked energy, ERC-20 (Ethereum) is cheaper than TRC-20 (TRON), reversing the usual advice. TRC-20 becomes the cheapest only with TRX energy staked in advance. Re-check live costs before moving size.
Can I connect from any country?
The jurisdiction where your business is registered is checked during KYB and decided case by case. It is set by where the company is registered, not by the language you read in or the country you sign up from.
Does the network I send on change what my recipient receives?
It changes the on-chain fee and which wallets and exchanges will accept the transfer, not the dollar value of the USDT. One rule protects the money: the sending network must match the network your recipient’s address and exchange expect. Sending TRC-20 USDT to an ERC-20 address, or the reverse, can lose the funds.
Should I keep the payment in stablecoin or cash out to fiat?
Keep it in stablecoin to preserve the saving, and cash out only when you genuinely need local currency. Before you convert, ask what the off-ramp costs in your payout currency; that number decides whether the crypto route was actually cheaper.
The one number to ask for
Before switching a supplier or contractor payment to crypto, ask any provider a single question: what does the off-ramp cost in the currency you pay out? The on-chain fee is cents, the gateway fee is published, and the FX spread on a bank wire is the hidden cost you are trying to escape. The off-ramp is the line that decides whether you escaped it. A crypto payment gateway that answers that question plainly is doing the honest version of this math for you.
