What a Cross-Border Stablecoin Transfer Actually Costs

Five phases, one of which is on a blockchain

A cross-border stablecoin transfer looks like a single action on a screen. In cost terms it is five: getting money into the system, buying the stablecoin, moving it on-chain, selling it for local currency, and withdrawing that currency to a bank account. Banca d’Italia’s mystery-shopping study — published in July 2026 as No. 86 in its Markets, Infrastructures and Payment Systems series — executed real transfers of 200 USDC across ten corridors between Italy and Argentina, Brazil, South Africa, the United Arab Emirates and Japan on 24 and 26 March 2026, and priced every phase.

The result overturns the usual framing. The on-chain leg ranged from 0.01% to 1.40% of the amount. The purchase leg reached 5.36% in the most expensive direction, the sale leg reached 2.84%, and the withdrawal leg 3.40%. Total end-to-end cost ranged from 0.30% to 8.96%.

In other words, the part of the process that is genuinely novel — a settlement layer that works the same at 3 a.m. on a Sunday — is the cheapest part. The expensive parts are the same fiat conversions that a remittance company charges for.

The corridor table, read properly

The study published a per-corridor total for eight of its ten corridors, and the spread is wide enough to make the point without any decomposition at all.

Two rows deserve attention.

The 0.30% row is not a benchmark. The study is explicit that the remarkably low cost of the Italy-to-Argentina transfer must be interpreted with caution: the received peso amount converted back at the official rate produces a flattering number, and a more realistic market conversion would reduce or erase the advantage. A reader who takes 0.30% as “what stablecoins cost” has taken an exchange-rate artefact as a technology claim.

The 8.95% row is the counter-example. In the reverse UAE direction, the only funding option available was a credit card, carrying a 3.8% surcharge that the study folds into a combined funding and purchase cost of 6.17%. The on-chain leg in the same corridor was 1.40%, the sale 0.81% and the withdrawal 0.58%. Against a World Bank benchmark of 2.65% for that sending country, the stablecoin route was more than three times as expensive.

That is the honest summary of the study: stablecoins are not systematically cheaper, they are differently distributed.

Where the delays come from

Speed is not a blockchain property either. The study recorded end-to-end settlement under 20 minutes where instant payment systems existed at the domestic ends — Brazil’s Pix, the euro area’s TIPS, Argentina’s Transferencias 3.0 — and one to two business days where standard bank transfers were required.

The pattern is consistent across the corridors: the chain leg confirmed quickly and the fiat leg set the clock. A recipient whose bank posts credits in batches on the next business day receives the stablecoin quickly and the money slowly.

That has a practical consequence for anyone quoting a settlement time to a counterparty. The number worth quoting is the destination bank’s cut-off and posting schedule, not the chain’s block time.

The stablecoin itself is the boring part

USDC’s own properties are the least variable element in this picture. Circle describes USDC as fully backed and redeemable 1:1, with reserves reported weekly. That is a claim about the issuer’s balance sheet, and it matters for a different question — whether the token holds its peg — rather than for how much a transfer costs.

What it does mean is that the asset in the middle of the transfer is not the risk. The risk is the two conversions around it, which is why how on-ramp fees differ from spread is the more useful reading for anyone planning a transfer than any comparison of block times.

The failure that costs more than the fee

The one structural advantage of a stablecoin transfer — irreversibility — is also its worst failure mode. A transfer sent to the wrong network, or to an exchange that does not credit that chain, does not bounce.

Binance’s published recovery page illustrates the price of that class of error. For deposits on unsupported EVM networks it quotes about 20 USDT and roughly 30 working days, and it states that successful recovery of funds is not guaranteed. On a 200 USDC transfer that is a 10% charge and a month of waiting — an order of magnitude worse than the most expensive corridor in the central bank study.

The cheapest control is a five-second check: the receiving venue’s network list, and the memo or tag requirement if the destination is a shared address.

What the study does and does not show

The study is a small, careful piece of evidence with clear limits, and it is worth stating them because the topic attracts confident generalisations.

It is small. Ten corridors, one sending country at the centre, 200 USDC per transfer, two days of execution. It is a benchmark, not a census.

It was conducted in calm conditions. The authors note the transfers ran during a period of low Ethereum congestion, and that execution times and gas fees could differ under load.

It prices one asset on one chain family. Ethereum was the dominant rail in the exercise, and a transfer on a cheaper chain would change the on-chain line — though the study’s own table suggests that line was rarely the binding constraint.

It compares against national averages, not against the best available provider. The World Bank benchmarks used are country averages; a specific remittance company on a specific corridor may beat both.

The conclusion the authors draw is narrower than either camp in the debate would like: stablecoins are complements to, not substitutes for, efficient domestic payment infrastructure. Where the domestic rails are good, the stablecoin adds a wrapper. Where they are broken, it substitutes for them — and charges accordingly at the edges.

A checklist before quoting a cross-border price

1. Price the funding rail first. A card surcharge can dominate everything else, as the UAE corridor shows.
2. Price the destination conversion. The spread on the local-currency sale is the second-largest line in most corridors.
3. Check the network the recipient actually credits, not the one with the lowest fee.
4. Add the withdrawal fee at the destination, which in the study ranged from zero to 3.40%.
5. Quote a settlement time from the destination bank’s schedule, not from the chain’s finality.
6. Assume no refunds. Recovery exists at large venues and is priced in dollars and weeks.

For the narrower question of how USDC moves between chains, CCTP fast versus standard covers the mechanism and what the fee buys. For the charge that gets confused with the chain fee, withdrawal fees versus network fees separates the three charges that a transfer receipt shows as one.

Risk note. Every figure in this article comes from Banca d’Italia’s No. 86 (July 2026) mystery-shopping study, from Circle’s USDC page, or from Binance’s recovery page, all read on 2026-10-31. The study describes transfers executed in March 2026 under specific conditions and does not predict what any corridor costs today. Nothing here is investment or remittance advice.