BIS Working Paper 1385 compared bank wires, money transfer operators, and stablecoins on all-in cost. The five findings, the numbers, and what they mean.
BIS Working Paper 1385, published in October 2026, compares bank wires, fintech money transfer operators, and stablecoin transfers on all-in cost. Its verdict: bank FX wires are expensive, fintechs have cut costs sharply, and stablecoins price the FX leg near wholesale rates but add new fees and intermediaries. Stablecoins help. They are "not a panacea."
The paper is "Competing rails for cross-border payments: banks, fintechs, and stablecoins" by Wenxin Du, Catherine Huang, and David Scharfstein of Harvard Business School (BIS, October 2026). It is the most detailed public cost comparison of the three rails so far. If the term "stablecoin sandwich" is new, start with what a stablecoin sandwich is.
Key takeaways
It measured the all-in cost of sending money across borders on three rails, using one formula for all of them.
All-in cost = FX markup against a Bloomberg benchmark rate + explicit fees, both in basis points of the amount sent. A basis point is 0.01%, so 100 basis points is 1%.
The three rails:
The data mixes several sources. MTO prices were collected every 30 minutes from public websites starting January 2026, across about 100 corridors and four sizes ($200, $500, $10,000, $50,000). Bank wire costs came from a small sample of the authors' own transfers. Stablecoin prices came from centralized exchange data via CoinDesk, exchange fee schedules, gas fee data, and DeFi trade data. The World Bank's Remittance Prices Worldwide database filled in historical remittance costs.
The paper lists five takeaways, in this order.
| # | Finding | The number behind it |
|---|---|---|
| 1 | Retail costs come from payment architecture, not thin FX markets | Bank FX markups of 200 to 400 bps vs wholesale spreads of a few bps |
| 2 | Fintech MTOs compressed costs | About 40 to 150 bps on average, falling as size rises |
| 3 | Stablecoins cut settlement frictions but keep intermediaries | FX leg near benchmark; trading, withdrawal, and ramp fees added |
| 4 | Stablecoins have an edge in capital-controlled markets | Stablecoin-implied rates beat official rates in several emerging markets |
| 5 | Liquidity limits scale | USD stablecoin to EUR: about $200 million a day vs $2.7 trillion EUR FX turnover |
Each one changes how a payments team should read a stablecoin pitch. The sections below take them one at a time.
Because of how the rail is built and priced, not because currencies are hard to trade.
Wholesale FX is deep. The paper cites the 2025 BIS Triennial Survey: about $9.5 trillion of average daily FX turnover in April 2025, with EUR, JPY, and GBP trading against the dollar at bid-ask spreads of 6 to 9 basis points. Yet the authors' own wires through large international banks showed FX markups of about 230 to 470 basis points for EUR and USD pairs, before any wire fee.
The paper points to two causes. The fixed cost of running correspondent relationships. And banks' pricing power in retail FX. For a wire-by-wire breakdown of where the money goes, see hidden fees in international wires.
By paying out of local balances they already hold, so most transfers never cross a border at the moment the customer sends.
MTOs keep accounts in each country, offset customer flows going in opposite directions, digitize KYC (know your customer checks), and show a fixed fee plus a disclosed rate. The paper finds median costs of roughly 200 basis points for $200 to Tier 1 currencies (USD, EUR, JPY, GBP) and nearly 300 basis points to Tier 4 currencies. At $10,000, costs fall to about 40 to 60 basis points for Tier 1 and Tier 2 currencies and about 80 to 130 basis points for Tier 3 and Tier 4.
The catch: MTOs still rely on partner banks, pre-funded local liquidity, and wholesale FX to settle imbalances. They re-optimize the old architecture. They don't leave it.
The FX leg can be close to free for major currencies. The rest is not.
The paper splits a stablecoin transfer into four costs: the FX cost of converting between fiat and stablecoins, exchange trading fees, blockchain gas fees, and fiat on- and off-ramp fees such as withdrawal charges.
The paper's verdict: centralized exchanges play "a role analogous to correspondent banks." The rail swaps one set of intermediaries for another. That matches what the Federal Reserve's March 2026 FEDS Note said: stablecoins can shorten the intermediary chain, but large banks may keep a role in liquidity and compliance.
Because of capital controls, not better technology.
In several emerging-market corridors, the stablecoin rate was more favorable than the official benchmark for money sent from the US. The paper calls this a stablecoin access premium: where residents face limits on buying dollars, they pay extra for dollar stablecoins, so a sender converting into local currency gets a better rate. The authors are blunt: these advantages "arise from capital control distortions, not from technological efficiency."
Very little next to traditional FX.
The USD stablecoin to euro pair is the deepest, at about $185 million a day on centralized exchanges and about $15 million on decentralized exchanges. Traditional FX turns over about $2.7 trillion a day in EUR. Beyond the euro, only five currencies (KRW, TRY, BRL, GBP, and THB) averaged more than $10 million a day against USD stablecoins.
Thin markets mean volatile prices and slippage on large tickets. The paper concludes stablecoins may compete for retail-sized payments, while their ability to displace traditional rails in high-value institutional transfers "is currently limited." For how this plays out in a payout quote, see onchain liquidity vs off-ramp liquidity.
A caution on comparability, and a second cost estimate.
Tobias Adrian of the IMF, discussing the paper (included in the same BIS publication), noted that bank figures exclude wire fees while fintech figures are all-in, so the three rails are not measured identically. He cited forthcoming IMF research covering 270 corridors that put the average cost of a $200 stablecoin transfer at about 2.2%, with a range of roughly -2% to 8%. In that work, the cheapest traditional option still beat the cheapest stablecoin option in about one-third of corridors.
Three practical rules fall out of the paper.
Our own cost per delivered dollar method applies the same all-in logic to a single payout.
BlindPay prices the whole conversion up front. A payout quote locks the exchange rate and fees for a specific bank account and expires after 5 minutes, and the response shows the exact amount the recipient receives (payout quotes). BlindPay combines a monthly plan with per-transaction fees that vary by rail and network, disclosed in the live quote before execute. Payouts land over local rails: Pix, PIX Safe, and TED in Brazil, SPEI in Mexico, ACH Colombia, Transfers 3.0 in Argentina, ACH, wire, and RTP in the US, SEPA, Faster Payments, and SWIFT (POBO/COBO) with UETR tracking and MT103 confirmations.
Pick your three busiest corridors. For each, get a same-minute quote from your bank, a money transfer operator, and a stablecoin payout provider at your typical payment size, then compare the amount the recipient receives. That is the comparison the BIS paper says matters.
Always cheaper? No intermediaries? Recipient needs a wallet? 10 common claims about cross-border stablecoin payments, checked against the Fed and the BIS.
Seven places blockchain payments beat bank rails: contractor payroll, remittances, B2B suppliers, marketplaces, 24/7 treasury, bill pay, and PSP payouts.
The dates that turned stablecoins from a trading tool into cross-border payment rails: USDC, Pix, MiCA, the GENIUS Act, ISO 20022, and Visa settlement.