Most stablecoin use cases sound better on a slide than they look in a ledger. Cross-border value transfer is the exception. It is the flow where the numbers consistently survive contact with a real corridor, a real treasury team, and a real reconciliation process. So it is worth understanding exactly why this one works, because the reasons are narrow and they tell you where the case stops working too.
The short version: the savings are real, but they do not come from the part everyone points at. The on-chain transfer is nearly free and nearly instant. That was never the expensive part of correspondent banking. The cost lives at the edges, in converting fiat to a token and back again. Whether a corridor pencils out depends almost entirely on what those two conversions cost.
Why the legacy rail is expensive
A traditional cross-border payment is slow and costly because of structure, not technology. The average B2B cross-border payment touches roughly four intermediaries, takes about three days, and loses 3 to 6 percent to fees and FX spread along the way. Each correspondent bank in the chain holds the value briefly, takes a cut, and adds latency.
Retail remittance is worse. The World Bank's Q3 2025 data put the global average cost of sending a remittance at 6.36 percent, with banks averaging 14.99 percent and digital channels around 4.59 percent. Settlement runs one to five business days because the message and the money move on separate, batch-processed systems that do not run nights or weekends.
This is the baseline a stablecoin corridor has to beat. It is a high baseline, which is why even a clumsy stablecoin implementation can look good.
Where the stablecoin saving actually comes from
The on-chain leg is the cheap, boring part. A transfer of USDC or USDT on a low-cost chain like Solana or Base settles in well under a minute, around the clock, for cents in network fees. The Federal Reserve's March 2026 analysis put per-transaction stablecoin cost in a range of roughly $0.01 to $1.00 with sub-minute settlement.
That speed removes float. Money that used to sit in transit for three days is now usable in three minutes, which matters as much to a treasury team as the headline fee. Indian IT firms invoicing US clients have reported materially faster cash flow on stablecoin settlement for exactly this reason.
But the Fed note is also blunt about the limit: stablecoins eliminate the messaging and intermediary cost, not the conversion cost. You still have to get into the token and out of it.
The off-ramp is the bottleneck
To pay a supplier in Lagos or Manila, someone has to convert the stablecoin into naira or pesos at the far end. That conversion is the off-ramp, and in thinner markets it carries a wider spread than the dollar leg ever did. Converting to currencies like NGN, IDR, PKR, or BDT runs through shallower liquidity and developing banking partnerships, which partially erodes the on-chain advantage.
So a real corridor cost is roughly: on-ramp spread, plus near-zero on-chain fee, plus off-ramp spread, plus any provider processing fee. Providers typically charge 0.3 to 2 percent in processing, with conversion spreads of 0.5 to 2 percent per side. The on-chain step you were excited about is a rounding error inside that stack.
A worked example
Take a US company paying a $50,000 invoice to a vendor in Brazil.
On the legacy rail, a wire plus correspondent fees plus FX markup at, say, 3 percent all-in costs roughly $1,500 and clears in two to three days, with the vendor unsure of the exact landed amount until it arrives.
On a stablecoin corridor: the US firm converts $50,000 to USDC (on-ramp spread near zero through a regulated provider), sends it on-chain for under a dollar in minutes, and a local partner converts USDC to Brazilian reais. If the off-ramp spread and processing land at around 1.2 percent combined, the total cost is roughly $600 and the vendor is funded the same day.
That is a saving near 60 percent and two days of float recovered. It is real, and it is repeatable. But flip one variable. Route the same payment to a corridor where the only off-ramp partner charges 4 percent because local liquidity is thin and compliance overhead is high, and the all-in cost converges back toward the wire. The chain did not change. The edge did.
This is why measured results cluster where they do. Across B2B implementations, total cost reduction lands in roughly the 30 to 50 percent range once fees, FX spread, float, and intermediary deductions are all counted, not the 99 percent that "cents on-chain" framing implies.
What makes a corridor pencil out
Three conditions have to hold together.
First, a high legacy baseline. The corridor has to be genuinely expensive today, which means emerging-market and lower-volume lanes, not London-to-Frankfurt where rails are already cheap and instant.
Second, a real off-ramp. There must be a licensed local partner with enough liquidity to convert at a tight spread. Coordination layers like Circle Payments Network, which launched in May 2025 and settles in USDC or EURC, exist mostly to solve this last-mile and compliance problem by connecting vetted institutions across markets like Brazil, Colombia, and Nigeria. The token was never the hard part. The off-ramp network is.
Third, regulatory cover at both ends. The US GENIUS Act, enacted in July 2025, set reserve and disclosure standards for issuers, and the far-end jurisdiction needs its own clear treatment. This connects to the wrappers and reserves we cover in earlier modules. A corridor that is cheap but legally ambiguous on the receiving side is not a corridor a treasury team can use at scale.
Where it does not pencil out
It does not work on already-cheap, already-fast lanes. The savings come from beating a bad baseline, and major-currency corridors no longer have one.
It does not work where the off-ramp is thin or monopolized, because one expensive conversion partner eats the entire advantage.
And it does not work as a domestic story dressed up as cross-border. If both ends sit in the same currency and the same banking system, you have added a token round-trip and two conversions to a payment that did not need them.
Takeaway
Cross-border is the use case that pencils out because it attacks a structurally expensive, slow, multi-intermediary baseline, not because tokens are magic. Run the math on the edges, not the chain. If the corridor is expensive today and you can name a real, liquid, licensed off-ramp at the far end, the case is strong and durable. If you cannot, the on-chain leg being free will not save you, and you are better off skipping it.