Stablecoins for payment service providers. 6 chapters and a final quiz.

Chapter 1. Framing the problem: what a stablecoin rail solves.

A stablecoin rail is not justified everywhere, and it loses within a single currency area. Pix in Brazil, SPEI in Mexico, SEPA Instant Credit Transfer in the euro area, and the Faster Payments Service in the UK already settle in a few seconds, at almost no cost. The token wins elsewhere, where the payment goes through a correspondent bank. Only there does round-the-clock availability become a measurable advantage.

The fit test, in five questions

• Does the flow go through a correspondent bank? Without a correspondent, there is nothing to replace. A domestic instant transfer always beats the on-chain rail.
• Does the beneficiary wait outside business hours? Saturday-morning settlement is the one argument that neither cards nor wholesale transfers can match.
• Do you have to prefund an account in the destination country? Then the gain is measured in trapped working capital, not in transaction fees.
• Does a local exit leg exist, and at what price? A corridor without a reliable off-ramp is not a corridor. It is a stranded balance.
• Does the merchant accept irreversibility? There are no disputes on-chain. What goes out stays out.

Chapter 2. Vetting the issuer as a counterparty.

Accepting a token means accepting a debtor. The balance you receive is a claim on a private company, backed by no public guarantee in either the European Union or the US. The due diligence looks like the one you would run on a new partner bank: legal identity, license, assets against liabilities, exit terms. You run it once, then review it on a fixed schedule.

The word “audit” gets used loosely here. Issuers publish attestations. An attestation confirms the tokens in circulation and the reserve composition on a given date, with no opinion on the financial statements or on internal controls. It proves that the assets existed on that date, not that they exist every day. The question for the provider is about the interval between reports: how often, which firm, and what detail is published in between. Circle publishes a monthly attestation signed by Deloitte & Touche LLP, backed by a weekly disclosure and its portfolio detail, while Tether publishes a quarterly attestation signed by BDO.

Six clauses to secure before you sign

• Price and timestamp: a named reference rate, its source, the fixing time, and the maximum spread applied to the conversion
• Maximum conversion time: the point beyond which the orchestrator, not you, bears the price gap
• Exposure cap per issuer: a written exposure limit, reviewed as often as a bank counterparty limit
• Notice before a token is delisted: how much time you get if the provider stops supporting it
• Freeze notification: who notifies you, how quickly, and what recourse exists
• Orderly exit: how long it takes to wind down balances and switch to a second asset already under contract

Chapter 3. Checking that the rail is legal on both sides: MiCA, the GENIUS Act, and local regimes.

You cannot assess a stablecoin setup under a single legal system, because four jurisdictions matter in every corridor: the payer's, the beneficiary's, the token issuer's, and that of the provider handling the conversion. A token that is perfectly legal where the beneficiary sits may be ineligible where the payer sits. Write the assessment memo before the first API call, not after the first rejection.

MiCA, Regulation (EU) 2023/1114, applies in stages: Titles III and IV since June 30, 2024, and Title V since December 30, 2024. Under MiCA, a stablecoin backed by a currency is an e-money token. Three articles govern day-to-day operations. Article 49 grants a right to redemption at par, at any time, with no fees beyond the cost of execution. Article 50 bans paying interest to holders. Article 54 requires at least 30% of funds to be held as deposits with credit institutions, or 60% for a token classified as significant.

The US GENIUS Act was signed into law on July 18, 2025. It takes effect on the earlier of two dates: 18 months after enactment (January 18, 2027), or 120 days after federal banking regulators publish final rules. The law limits issuance to supervised entities and sets a $10 billion threshold for tokens outstanding. Above it, a state regime is no longer enough. The framework comes with no public backstop: no federal deposit insurance and no direct access to the central bank.

The practical impact is already visible among providers. Circle National Trust received final approval from the OCC on July 10, 2026. Bridge, Stripe's infrastructure subsidiary, received conditional approval in February 2026, and Anchorage Digital Bank has been issuing USA₮ for Tether's US arm since January 27, 2026. Your provider's status can therefore change mid-contract, and its supervisor with it. Include a notification clause, and review the counterparty profile at every change.

Chapter 4. Costing the corridor: on-ramp, off-ramp, and true cost.

Chain fees are measured in cents. They are not the cost of the rail. The cost sits at the two ramps, in the FX conversion and the local payout. A provider that highlights its network fees is steering you away from the only figure you can hold it to: the amount the beneficiary receives relative to the amount debited from the payer.

amount sent                           100,000.00 EUR

1  FX conversion in       25 bp           250.00 EUR
2  token purchase         10 bp           100.00 EUR
3  chain fees             flat              0.50 EUR
4  FX conversion out      20 bp           200.00 EUR
5  local payout           flat              8.00 EUR
                                       -------------
   total cost                             558.50 EUR
                                        = 55.9 bp   = 0.559%

comparison: transfer through a correspondent bank
   fixed fees                              30.00 EUR
   FX spread              60 bp           600.00 EUR
                                       -------------
                                          630.00 EUR
                                        = 63.0 bp   = 0.630%

saving on the on-chain rail                71.50 EUR per transaction
                                        = 7.1 bp

NOTE: these values are exercise assumptions, not market prices.
The only figure you can hold a provider to is the AMOUNT RECEIVED
by the beneficiary, measured on a real send timestamped end to end.

FX dominates. In this exercise, the two spreads cost €450 while chain fees come to €0.50, a ratio of 900 to one. Negotiate this line against a reference rate named in the contract, with a fixing time. Without a written reference, the spread is whatever the provider decides at the time of the transaction. It is the most profitable line in the provider's offer, and the least discussed in an RFP.

The corridor test protocol

• Three amount tiers: a small ticket, the expected average ticket, and an amount ten times larger. Liquidity is not linear.
• A timestamp at every step: payer debit, FX execution, on-chain broadcast, token sale, beneficiary credit.
• One transfer on a weekday, one on a Saturday, one on a local public holiday. Three measurements, three different delays.
• The amount received, not the advertised fees. It is the only metric that survives the sales pitch.
• A comparison with the existing rail, on the same dates and for the same amounts.
• A second provider on critical corridors, tested with the same protocol and kept active.

Chapter 5. Holding tokens and managing operational risk.

The best custody is often no custody. A token that passes through the system in a few minutes exposes almost no one. A balance that sits for a week leaves the PSP carrying the issuer's credit risk and FX risk. Choose the model before the technology, and justify the choice in writing.

Controls that stop an irreversible payout from going to the wrong place

• Address allowlist, with a cooling-off period between adding an address and the first transfer to it
• Test transfer of a small amount before any first payout to a new address
• Dual approval above a written threshold, by two people using different authentication methods
• Chain check: the same address can exist on several networks, and a transfer on the wrong network is lost
• Contract address check: it is the only proof of which asset you received. A token with the same name at another address is a different asset
• Decimals read from the contract: USDC and USDT declare six, other tokens 18. The mismatch throws an amount off by a factor of a trillion
• Gas token balance monitored on every chain: without it, no payout goes out

def credit_deposit(event):
    # 1. THE ASSET: the contract address, never the displayed symbol
    if event.token_contract not in AUTHORIZED_CONTRACTS:
        return reject("unrecognized asset")

    # 2. THE CHAIN: an authorized contract on a chain we have not opened
    #    is not a valid deposit
    if event.chain not in OPEN_CHAINS:
        return quarantine("chain not open")

    # 3. THE AMOUNT: decimals read from the contract, never hard-coded
    amount = event.amount_raw / 10 ** decimals(event.token_contract)

    # 4. FINALITY: a confirmation threshold set for each chain
    if event.confirmations < MIN_CONFIRMATIONS[event.chain]:
        return wait()

    # 5. THE COUNTERPARTY: screen BEFORE crediting, never after
    if not screening_ok(event.from_address):
        return escalate("address needs review")

    # 6. THE MATCH: one deposit address per invoice;
    #    the chain carries no standardized remittance reference
    invoice = invoice_for_address(event.to_address)
    if invoice is None:
        return manual_review("unidentified deposit")

    return credit(invoice, amount)

A depeg hits you over a window of time, not as a matter of principle. In March 2023, USDC fell to $0.87 after $3.3 billion of its reserves were trapped at Silicon Valley Bank, and it took about three days to return to par. A business that converts its collections within the hour carries one hour of exposure; one that holds them all week to convert on Friday carries five days. Holding time is the only variable the PSP truly controls.

Chapter 6. Reconciling, accounting, and deciding.

Reconciling an on-chain collection is no harder than reconciling a card batch. It is just less forgiving. With no disputes and no recall of funds, an undetected discrepancy is permanent. So three data sources must match every day on a shared key, and any difference still open at the end of the day becomes an open claim, not a line to watch.

Three-way reconciliation

• The on-chain transaction: hash, contract address, chain, amount in base units, decimals, confirmations, deposit address.
• The provider's conversion confirmation: quantity of tokens sold, rate applied, timestamp, fees deducted, transaction reference.
• The bank credit in the account currency: amount, value date, payment reference, payout rail used.
• The matching key: the deposit address dedicated to the invoice, carried in all three sources. The chain carries no standardized remittance information.

What to settle with the auditor before the first close

• How the balance is classified: a claim on a private issuer, with no public guarantee, and not a cash equivalent by default
• Valuation source: which venue, what time of day, and what rule applies if a price gap persists
• FX treatment: collecting in a dollar token at an entity that keeps its books in euros creates a difference to document
• Segregation of client funds: which addresses, which accounts, what proof that assets are not commingled
• Proof of key control: who signs, with what device, and how to prove it without exposing the keys
• Audit trail retained: transaction hash, contract address, timestamps, conversion confirmations, bank statements
• Treatment of a frozen token: how to account for a frozen balance, and what the merchant must be told