At 3:30 in the afternoon on June 26, 1974, German regulators shut down a relatively small bank in Cologne. Unfortunately, they closed it in the middle of the world’s business day.
Several of the bank’s counterparties had already delivered Deutsche marks in Europe. They were waiting for the bank to deliver dollars in New York, where the morning had only just begun.
Those dollars never arrived.
The money had not disappeared while traveling through a tunnel. There was no broken cable and no forged payment instruction. The problem was more fundamental: what appeared commercially to be one currency exchange was legally and operationally two separate payments.
One had been completed. The other had not.
The failed institution was Bankhaus Herstatt. Its collapse gave the financial world a lasting term—Herstatt risk—for the possibility that one participant in a foreign-exchange transaction delivers the currency it sold but never receives the currency it purchased.
Half a century later, the technology is faster, the amounts are vastly larger and the payment messages travel almost instantly. But the underlying problem has not disappeared.
In fact, stablecoins, tokenized deposits and instant-payment systems are forcing the industry to confront the same question in a new form:
When two assets are being exchanged, how can each party be certain that it will not pay unless the other party pays too?
A trade that looks simpler than it is
Suppose a Turkish importer needs to purchase €10 million from a European bank.
The importer agrees to deliver the Turkish-lira equivalent. The bank agrees to deliver €10 million.
Economically, this looks like a single exchange:
Lira go one way; euros come the other way.
Operationally, however, the transaction may involve two payment systems, two correspondent-banking chains, different operating hours and multiple intermediaries.
The lira payment might settle through Turkey’s domestic payment infrastructure. The euro payment might settle through a European correspondent or central-bank money in the euro area.
One currency can therefore become final before the other.
Imagine that the importer’s bank sends the lira at 10:00 a.m. The receiving institution obtains irrevocable control of the funds. The euro payment is scheduled for later that afternoon.
At noon, the receiving institution fails.
The importer’s bank does not merely face a delay. It may have lost the entire principal amount it delivered.
This is what distinguishes FX settlement risk from ordinary market risk.
If the euro exchange rate moves against you before a trade is completed, you may lose the difference between the contracted price and the replacement price. That is market or replacement-cost risk.
If you deliver US$10 million and receive nothing in return, you may lose the US$10 million itself.
The principal exposure can be many times larger than the profit expected from the trade.
What happened at Herstatt?
Bankhaus Herstatt had become heavily involved in foreign-exchange trading during the turbulent period following the collapse of the Bretton Woods fixed-exchange-rate system.
Currency values were moving more freely. That created opportunities for trading profits but also exposed poorly controlled institutions to substantial losses.
Herstatt accumulated losses large enough to overwhelm its capital. German supervisors withdrew its banking license and ordered it closed on June 26, 1974.
The timing proved disastrous.
Because European payment systems had already been operating for hours, some counterparties had completed their side of dollar-related currency trades by delivering Deutsche marks to Herstatt. But New York was several hours behind Cologne. Herstatt’s corresponding dollar payments had not yet been made.
When the bank closed, its US correspondent stopped outgoing dollar payments.
The Bank of England describes the event simply: Herstatt had already received foreign-currency payments, but its counterparties did not receive the dollars they were owed and suffered substantial losses.
The failure was not enormous by modern banking standards. Its influence came from what it revealed.
International banking had grown faster than the systems for supervising and settling it. A bank could trade around the world while regulators, payment systems and legal rules remained largely national.
Later in 1974, the central-bank governors of the Group of Ten established what eventually became the Basel Committee on Banking Supervision. Herstatt was not the only reason, but its failure was one of the defining shocks behind the initiative.
A small bank had exposed a large gap in international finance.
Payment messages are not the same as settlement
People often say that a payment has “gone through” when they see a confirmation screen or receive a message from their bank.
In wholesale finance, that language can be dangerously vague.
Several events may occur at different times:
A trade is agreed.
Payment instructions are submitted.
Banks verify and match the instructions.
Funds are reserved or debited.
The receiving institution is credited.
The transfer becomes legally final and cannot ordinarily be reversed.
A message stating that a bank intends to pay is not the same as final settlement.
This distinction also applies to blockchains. A visible transaction may be confirmed on-chain, but that does not automatically resolve every legal or commercial question. One must still ask:
Which asset was transferred?
Who issued it?
Is the transfer final under the relevant rules?
Can the token be frozen?
Can it be redeemed for bank money?
Does the recipient have access to that redemption channel?
Was the asset on the other side of the trade delivered with equivalent finality?
Speed improves convenience. It does not, by itself, create a safe exchange.
The solution: payment-versus-payment
The cleanest protection against FX settlement risk is payment-versus-payment, usually abbreviated as PvP.
PvP links the two currency legs so that the final transfer of one currency occurs if—and only if—the final transfer of the other currency also occurs.
Think of two locked doors connected to the same mechanism. Neither opens alone. Both open when the required conditions have been satisfied.
For example:
Bank A must deliver US$100 million.
Bank B must deliver €85 million.
The system verifies that the required funding is available.
If both sides can settle, both payments are completed.
If one side cannot settle, neither principal payment is released.
“Payment-versus-payment” does not necessarily mean that two entries are recorded in the same technical microsecond. It means that settlement is conditional: one final transfer cannot occur without the corresponding final transfer.
This removes the classic Herstatt principal risk.
It does not eliminate every risk. A failed trade can still create liquidity problems, operational costs or replacement losses. But it prevents the worst outcome—paying away the full amount while receiving nothing.
How CLS changed the FX market
The most important institutional response is CLS, originally known as Continuous Linked Settlement.
CLS began operations in 2002, almost three decades after Herstatt. It connects settlement across eligible currencies and provides PvP protection.
Its role is easier to understand through a simplified example.
A global bank may have hundreds of thousands of FX obligations during a day. Without netting, it might need to send enormous gross amounts in each currency.
CLS calculates what each participating institution ultimately needs to pay into the system. A bank with US$30 billion of dollar payments and US$28 billion of dollar receipts may need to fund a much smaller net amount during the settlement process, subject to the system’s rules and schedules.
This offers two benefits:
PvP reduces principal settlement risk.
Multilateral netting reduces funding requirements.
CLS currently provides PvP settlement for 18 currencies. In 2025, it settled an average daily value exceeding US$8 trillion, including a record US$22.9 trillion in payment instructions on a single day in December.
Those numbers demonstrate the scale of protection. They also reveal why CLS is considered critical financial infrastructure: a system responsible for synchronizing trillions of dollars cannot be treated like an ordinary technology vendor.
It requires exceptional operational resilience, liquidity planning, legal certainty and central-bank oversight.
Why has Herstatt risk not disappeared?
CLS does not cover every currency, every counterparty or every type of FX transaction.
Emerging-market currencies present particular challenges. Some have capital controls, limited offshore liquidity or restrictions affecting who may hold and transfer them. Domestic payment systems may not remain open during the necessary settlement window. The relevant central bank may not permit or support participation in a global PvP arrangement.
Smaller institutions may also access settlement indirectly through larger banks, adding cost and dependency.
The BIS estimated that in April 2022 almost one-third of deliverable FX turnover—approximately US$2.2 trillion per day—remained exposed to settlement risk. That was higher in absolute terms than the estimated US$1.9 trillion in 2019.
More recent BIS survey work indicates substantial improvement: CLS reported that the 2025 BIS survey found approximately 90% of average daily FX settlement was processed through methods that eliminate or reduce settlement risk. Nevertheless, about 10% by value still settled on a gross bilateral basis.
Ten percent sounds small until it is applied to the enormous size of the global FX market.
Risk also tends to concentrate where protection is hardest to arrange: less-liquid currencies, smaller counterparties, emerging markets and payment corridors with complex correspondent-banking chains.
The safest part of the market becomes safer, while residual exposure collects around its edges.
Netting helps, but it is not PvP
Suppose two payment companies trade repeatedly during the day.
Company A owes Company B US$20 million. Company B owes Company A US$18 million. Instead of sending US$38 million gross, they agree that Company A will send the US$2 million difference.
This is netting.
Netting reduces the amount that must be transferred. It therefore reduces liquidity requirements and the size of the possible loss.
But netting alone does not guarantee that the remaining US$2 million payment will be exchanged safely against another currency.
A company can net thousands of trades and still face settlement risk on the final balance.
This matters in fintech proposals because the words “net settlement” are sometimes presented as if they solve the entire problem. They do not.
A complete design must answer two different questions:
How much must each participant pay?
How are the two currencies exchanged without one side paying first?
Netting addresses the first. PvP addresses the second.
Why instant settlement can create new pressure
It is tempting to believe that every financial transaction should settle instantly.
Sometimes that is correct. Shorter settlement periods can reduce the time during which a party is exposed to counterparty failure or market movements.
But immediate gross settlement can also increase liquidity requirements.
If every FX trade must settle individually and instantly, institutions need the correct currency in the correct account at precisely the right moment. They have less time to net incoming and outgoing transactions or arrange funding.
A bank may be solvent but temporarily unable to deliver a particular currency at 2:03 p.m.
This is why good settlement design balances three objectives:
Reducing credit risk;
Conserving liquidity; and
Preserving operational resilience.
“Faster” is not a complete financial architecture. The real objective is safe finality with manageable liquidity needs.
Stablecoins rediscover an old problem
Stablecoin advocates sometimes describe blockchain settlement as atomic: either both sides of a transaction occur or neither does.
Within a single technical environment, smart contracts can indeed coordinate asset transfers in ways that resemble delivery-versus-payment or payment-versus-payment.
But the protection depends on the boundaries of the transaction.
Consider an exchange between a US-dollar stablecoin on one blockchain and euros in a commercial-bank account.
The stablecoin transfer may be completed in seconds. The euro payment may still depend on banking hours, correspondent institutions and fiat-payment finality.
Herstatt risk has not been eliminated. One side of it has merely been accelerated.
Cross-chain transactions introduce other complications. A bridge, oracle, custodian or market maker may become the point of failure. Two tokens can transfer atomically while still carrying different issuer, redemption and legal risks.
Even exchanging two stablecoins is not necessarily equivalent to exchanging two dollars. One token may be redeemable directly by the holder; another may require an account with the issuer. One may hold short-term government securities; another may depend on a bank deposit, affiliated market maker or offshore legal structure.
Technical atomicity cannot make economically unequal assets identical.
The proper questions are therefore:
What constitutes final delivery on each side?
Can the recipient use or redeem what it receives?
Does the transaction cross blockchains or move between tokenized and conventional money?
Which intermediary supplies liquidity?
What happens if that intermediary fails midway?
Which law recognizes the transfer as final?
The labels have changed. The underlying discipline has not.
The larger lesson
Herstatt risk teaches us that every exchange contains a moment of vulnerability.
Commerce speaks as though assets are swapped: dollars for euros, cash for securities, stablecoins for bank deposits. Payment infrastructure sees separate instructions moving through separate systems.
The job of settlement design is to make those separate movements behave like one exchange.
That is why payment finality matters. It is why time zones matter. It is why a payment confirmation should not be confused with money received. And it is why the most important part of a cross-border transaction may be the brief interval between the two sides.
Bankhaus Herstatt disappeared in 1974. The risk bearing its name remains wherever one participant must pay first and trust that another payment will follow.
Practical takeaway: When reviewing a cross-border payment, FX or stablecoin structure, draw both settlement legs separately. Identify the sending institution, correspondent chain, settlement system, operating window, point of legal finality and party exposed during any timing gap. Then determine whether the structure uses genuine PvP, simple bilateral netting or merely an operational promise to pay later. If one party can irrevocably deliver value while the other side remains pending, Herstatt risk is still present.