Stablecoins Do Not Make Remittances Cheap. They Make Them Optional.
Hook
Two hundred dollars. Italy to Brazil. USDC costs 2.70 percent. Wise costs 2.20 percent.
Reverse the corridor. Brazil to Italy. USDC costs 2.21 percent. Wise costs 4.68 to 4.89 percent.
Same asset. Same currency pair. Opposite conclusion.
For two years I have watched this inversion get flattened into one sentence. "Stablecoins make sending money cheap." They do not. Stablecoins make sending money possible, sometimes cheap, and sometimes more expensive than a bank wire with better marketing. The distance between those three claims is a direction. Most commentary deletes the direction because a corridor without a direction does not fit on a slide.
The ledger does not lie, only the interpreters do.
So I read the Bank of Italy's remittance study the way I read a contract โ for what it measured, not for what it implied. What it measured was a reversal. What it implied was an entire industry's pricing model.
Context
USDC is a dollar claim with a blockchain receipt. Circle issues it, holds reserves against it, and redeems it. The token does not float. It does not accrue. It is a transport layer wrapped around a unit of account that predates all of this by two centuries.
That is the entire product. The novelty is not the dollar. The novelty is the rail.
The rail works like this. A sender in Italy buys USDC on an exchange, holds it in a wallet, and transfers it to a recipient in Brazil. The transfer settles in seconds. The recipient now holds a dollar-denominated token on a Brazilian phone. To spend it, the recipient sells it on a local exchange and withdraws local currency to a bank account. The recipient then pays rent, groceries, school fees โ the ordinary destinations of an ordinary wage.
Every step in that chain is a variable. The buy price is a variable. The on-chain fee is a variable, small but non-zero. The sell price is a variable. The withdrawal fee is a variable. The bank settlement time is a variable. The exchange rate used at the point of sale is the largest variable of all, and it is the one most often invisible.
The Bank of Italy's researchers measured USDC transactions from March 2026 against Wise quotes simulated on 14 April 2026. Wise is the cleanest benchmark available, because Wise publishes its rate and its fee separately, and it does not hide margin in the spread the way a bank does. Comparing USDC against Wise is comparing one transparent rail against another transparent rail. That is the only comparison worth running. Comparing USDC against Western Union proves nothing except that Western Union is expensive.
The World Bank's remittance methodology is the correct frame here. Do not measure the fee. Measure the total cost โ what the sender pays out, minus what the recipient receives, divided by what the sender paid. Fees can be zero while the spread eats four percent. Providers know this. It is why "low fee" advertising survives contact with a bad exchange rate.
What the Bank of Italy found was not that stablecoins are cheap. It was that stablecoins are cheap in one direction and expensive in the other. That is a structural finding, and structural findings do not expire with the news cycle. A pricing claim tied to a corridor has a shelf life measured in months. A pricing claim tied to a direction has a shelf life measured in infrastructure cycles, and the infrastructure cycle we are in is long.
Core
Start with the asymmetry, because the asymmetry is the finding.
Italy to Brazil, two hundred dollars: the USDC path costs 5.40 dollars, or 2.70 percent. The Wise path costs 4.40 dollars, or 2.20 percent. Wise wins by half a percentage point.
Brazil to Italy, the same two hundred dollars reversed: the USDC path costs 4.42 dollars, or 2.21 percent. The Wise path costs 9.36 to 9.78 dollars, or 4.68 to 4.89 percent. USDC wins by roughly half.
The cost advantage of a stablecoin rail is not a property of the stablecoin. It is a property of the corridor and the direction of travel. An analyst who reports "stablecoins are cheaper" without naming the corridor has reported nothing. An analyst who reports "stablecoins are cheaper in the Brazil-to-Italy direction" has reported a market structure.
Why does direction matter? Because remittance corridors are bilateral order books, and bilateral order books are not symmetric. Liquidity in BRL/EUR and EUR/BRL is not the same depth on both sides. The correspondent banking relationships that price the Italy-to-Brazil leg are mature, dense, and cheap. The ones that price Brazil-to-Italy are thinner, more intermediated, and more expensive. The stablecoin rail does not care about correspondent banking because it does not use correspondent banking. It uses a token and an exchange. Where the traditional rail is efficient, the token loses. Where the traditional rail is inefficient, the token wins.
That is not a marketing claim. It is a routing claim. It says stablecoin rails are a substitute for the worst parts of the correspondent banking network, and they are a worse substitute for the best parts. Anyone selling the rail as a universal improvement has not run the corridor in both directions.
In 2021 I ran the same logic against Curve Finance's original gauge voting system. I calculated that the reward distribution favored whale wallets because the claim mechanics lacked slippage protection, and I published the math rather than the sentiment. The finding at Curve was the same shape as the finding here: the headline yield was correct on average and wrong for every individual participant. Averages hide the distribution. A corridor average hides the direction. The retail user in a pool is not the average user in a pool, and the remittance sender in a corridor is not the average sender in a corridor. Both facts are invisible in a pie chart and obvious in a spreadsheet.
Now the hidden costs, because the visible ones are the least interesting.
A stablecoin transfer has two cost components that behave completely differently. The first is explicit: the exchange fee, the withdrawal fee, the on-chain gas. These are small, discrete, and countable. The second is implicit: the exchange rate applied at the moment of conversion. This one is large, continuous, and discretionary.
Blockchain settlement controls the movement of the token. It does not control the exchange rate, and it does not control the pricing of any service around it. This is the boundary that breaks the narrative. A sender watching a transfer confirm in eight seconds concludes that the money has arrived. It has not. The token has arrived. The money arrives when the recipient's local bank credits the account, and that credit can take a day. The eight-second confirmation and the one-day settlement are not the same event, and the gap between them is where user expectations are manufactured and then destroyed.
This is the last mile problem, and it has migrated. In 2018 the bottleneck was throughput. In 2026 the bottleneck is the fiat ramp. The Layer 1 has become the easy part. The hard part is a retail conversion desk in a country where the on-ramp and the off-ramp are run by different companies with different fee schedules and no shared settlement standard.
I audited custody architecture for the spot Bitcoin ETF applicants in 2024, before approval, and the finding there rhymes with the finding here. The on-chain custody was elegant. The key management procedures around it did not meet traditional finance standards. The technical layer was ready and the operational layer was not. The same split defines stablecoin remittances. The token layer is ready. The conversion and withdrawal layer is fragmented across exchanges, banks, and local payment systems, each with its own KYC regime, fee schedule, and settlement latency. The recipient does not experience a stablecoin. The recipient experiences whatever the weakest link in that chain happens to be that week.
Trace the actual flow at the infrastructure level.
The sender buys USDC on an exchange. Retail access is exchange-gated. Institutions can mint directly through Circle Mint. The two populations have different fee structures, different KYC burdens, and different redemption rights. A family remitting two hundred dollars does not have a Circle Mint account. The family has a retail exchange account with a spread built into the fill price, and the spread is not on the receipt.
The recipient sells USDC on a local exchange and withdraws to a domestic bank. Whether that is fast or slow depends entirely on how well the exchange is connected to the local payment system. A well-connected exchange settles in hours. A poorly connected one settles in a day or more. The decisive infrastructure is not the blockchain. It is the connected exchange and the fast domestic payment rail behind it. Every dollar of user satisfaction in this model is produced off-chain, by companies most stablecoin commentary never names.
Circle's EEA redemption policy is the tell. Under the European regulatory framework, eligible European holders get a redemption path for USDC that sits outside the ordinary exchange flow. That policy exists because regulatory compliance is now part of the technical stack. A stablecoin is no longer just a smart contract and a reserve. It is a smart contract, a reserve, a legal entity, a redemption guarantee, and a jurisdictional map. Code is law; intent is irrelevant โ but code is also not the whole system, and the parts that are not code are the parts that decide the user experience. The smart contract enforces a transfer. The redemption policy decides whether a holder has a remedy. Those are different guarantees, and only one of them is auditable on Etherscan.
Here is the structural difference that the narrative consistently omits. A bank deposit carries deposit insurance. A stablecoin balance does not. The dollar in a USDC wallet is a claim on a reserve, not an insured deposit. If the reserve is mismanaged, the claim is impaired, and there is no backstop standing behind it. This is not a hypothetical risk. It is the same category of risk that produced the Terra de-pegging sequence I reverse-engineered in 2022, when I traced the oracle manipulation in Anchor's risk parameters and documented the transaction hashes that marked the start of the death spiral. The lesson from Terra was not that algorithmic stability fails. The lesson was that a claim is only as strong as the reserves behind it and the governance that manages them. USDC has real reserves and regulated governance. That is a genuinely different risk profile. It is not zero risk, and the absence of insurance is the precise measure of how far from zero it is.
So what is the stablecoin rail actually selling?
Not price. The price advantage is corridor-dependent and direction-dependent, and the data proves it in both directions.
Not speed. The on-chain leg is fast, but the money leg is not, and the money is what the recipient spends.
The real product is optionality.
A stablecoin recipient can convert part of the balance and hold the rest. Convert two hundred dollars partially, keep fifty as dollars, spend the rest in local currency. A traditional remittance service forces full conversion at the moment of receipt. There is no hold function, because there is no dollar balance to hold. Wise converts and delivers local currency. Western Union converts and delivers local currency. Neither one hands the recipient a dollar-denominated account as a side effect of the transfer.
That is the differentiated feature, and it is not a pricing feature. A stablecoin rail turns a remittance into a currency decision, and hands that decision to the recipient.
A recipient who receives dollars can choose to hold them through a period of local currency weakness. A recipient who receives a Wise transfer cannot. The optionality has real value, and that value is invisible in a fee comparison, which is exactly why the fee comparison is the wrong scoreboard. The Bank of Italy's own data gestures at this โ the report notes that recipients can convert partially, and that flexibility has no clean analogue in the incumbent rail. It is also the one feature the fee table will never capture, because it does not change the cost of the transfer. It changes who owns the timing decision after the transfer lands.
I want to be precise about the limit of the claim. Optionality is only worth something if the recipient can exercise it. Exercising it requires an exchange account, KYC, and enough familiarity with the tool to hold a token balance on purpose. For a recipient who already uses a crypto app, the rail is smooth. For a parent who does not, the rail is a wall. The same transfer produces two different products. One is a faster dollar account. The other is a homework assignment. User segmentation is not a footnote in stablecoin remittances. It is the main variable. Any cost model that averages across those two populations is measuring a customer that does not exist.
Contrarian
The bulls are wrong about the price and right about the product, and the distinction matters more than either camp admits.
The honest reading of the Bank of Italy data is that stablecoins lost the cheaper corridor and won the more important one. Italy-to-Brazil is a corridor where the incumbent rail is already efficient. Brazil-to-Italy is a corridor where the incumbent rail is expensive, and it is also the corridor where the sender is more likely to be the migrant worker sending money home. The stablecoin rail wins exactly where the incumbent rail is weakest, which is the only place a new rail has any business winning. A product that only beats the incumbent in the incumbent's strongest market is a product with no market. This is not that product.
The second thing the bulls got right is that optionality is genuinely hard to copy. Wise could add a dollar-hold feature. To do it, Wise would have to become a dollar custodian, which drags it into a different regulatory regime, a different reserve model, and a different set of capital requirements. The feature is not a product tweak. It is a business model change. That friction is real, and it gives the stablecoin rail a durable structural edge that has nothing to do with fees.
The third thing the bulls got right is the direction of travel on the last mile. The bottleneck is closing. Exchanges are integrating faster domestic rails. Withdrawal latency is compressing. Each improvement to the off-ramp improves the entire product, and the improvements compound because they sit at the point where the user forms the experience. History repeats, but the gas fees change โ the rail that looked slow in 2024 is a different rail in 2026, and it will be different again.
Where the bulls are wrong is the scoreboard. As long as the industry sells stablecoin remittances as cheap, it will be graded on price, and on price it will lose half the corridors it enters. As long as it sells optionality, it is graded on a feature the incumbent cannot match without rebuilding itself. The marketing error is not cosmetic. It selects for the wrong user, in the wrong corridor, with the wrong expectation, and it manufactures the disappointment that feeds the next round of "stablecoins do not work" commentary. That disappointment is a self-inflicted wound, and it is priced into every skeptical headline that follows.
Takeaway
The next catalyst in stablecoin payments will not be a cheaper fee. It will be a faster last mile, and the signal to track is exchange withdrawal latency, not headline transfer cost. Watch for deposit-insurance-equivalent structures, because the absence of a backstop behind a dollar claim is the one risk that stabilizes everything else in the model. And watch who bears the reserve risk when the corridor goes quiet.
Trust is a bug, not a feature โ at least until someone can count it.