The $1.4 Trillion Gap: Why FX Settlement Is Still Exposed
Foreign-exchange settlement risk is the risk that one side of a currency trade pays what it owes but does not receive the currency it bought. This series examines why that risk has not disappeared after decades of investment in safer settlement systems, and what the latest evidence says about the remaining exposure. The practical question is not simply whether the FX market is safer than it was. It is where protection reaches, where it stops, and why that boundary matters.
The issue is easy to underestimate because settlement is usually invisible to the person watching an exchange rate. A trade can be agreed in seconds, yet its two payments may move through different currencies, banks, time zones and operating schedules. The market’s price may be settled, but the payments that complete the trade still have to arrive in the right order—or be made conditional on one another.
When One Currency Arrives First
The classic illustration is the failure of Bankhaus Herstatt on 26 June 1974. German authorities ordered the bank into liquidation and closed it at 3:30 p.m. Frankfurt time. Earlier that day, counterparties had already delivered Deutsche marks to settle dollar–mark trades. After the closure, Herstatt’s New York correspondent suspended outgoing dollar payments. The counterparties were therefore left having performed their side of the bargain without receiving the currency owed to them.
That is the distinctive danger of FX settlement: the exposure can be the full value of the currency already delivered, not merely a fee, a delayed confirmation or the cost of correcting a clerical error. The failure was also more than a collection of bilateral losses. Market participants began delaying their own payments until they knew that counterparties had paid. Cross-border payments and FX trading slowed sharply, threatening a wider loss of confidence. Major central banks eventually signalled that they were ready to provide liquidity.
The lesson is about sequencing. Suppose a firm sells currency A and buys currency B. If the payment of A becomes final before the payment of B, the firm has temporarily financed the other side of the trade. If the counterparty defaults during that interval, the firm can lose the principal it delivered. The longer or less predictable the interval, the more difficult it becomes for participants to trust the payment chain in a period of stress.
This risk is sometimes called “Herstatt risk,” after the 1974 episode. It is not the same as every risk associated with an FX trade. A trade can be settled successfully and still create a replacement-cost loss if the original transaction must be recreated at a worse market price. A firm can also face a liquidity problem if it does not receive an expected currency on time. But settlement risk comes first in a very literal sense: it asks whether the two purchased currencies are exchanged without leaving one party’s principal hanging in the air.
Payment-versus-payment, or PvP, addresses the sequencing problem by making the final payment of one currency occur if and only if the final payment of the other currency occurs. The condition changes the exposure itself: neither side is supposed to complete only half of the exchange. The protected result depends on more than the existence of a PvP concept, however. The relevant counterparties, currency pair, trade type, legal arrangements, liquidity and operating window must all connect to a system that can apply the condition.
The Shrinking Share That Still Equals $1.4 Trillion
The April 2025 BIS Triennial Survey offers the clearest current scale of the problem. It measured actual settlement during the month—not FX turnover—and covered two-way-payment trades reported by dealers in 49 jurisdictions. On an average day, the survey recorded $14.186 trillion of gross financial obligations.
The largest protected slice was $5.157 trillion, or 36 percent, settled through PvP systems. Another $7.6 trillion, or 54 percent, used methods that mitigated but did not eliminate settlement risk. These included intragroup settlement, pre-settlement netting and settlement over bank accounts with timing controls. Together, the two groups account for about 90 percent of average daily settlement that either eliminated or reduced the settlement-risk exposure.
The remaining 10 percent was not small in absolute terms. About $1.416 trillion per day settled on a gross bilateral basis, through traditional correspondent-banking relationships and without a method that mitigated FX settlement risk. The survey therefore describes those trades as exposed for their full value. That does not mean $1.416 trillion is expected to be lost each day. It means that, if a counterparty fails after one currency has been delivered, the settlement-risk exposure is measured against the full amount of the trade.
The percentages become more meaningful when their denominators are kept visible. The 10 percent is a share of the $14.186 trillion that actually settled in April 2025; it is not a share of all financial assets, and it is not the same thing as the $7.5 trillion average daily FX turnover reported for 2022 in a New York Fed account of the market’s evolution. The latter figure is a turnover measure, while the BIS settlement survey counts obligations that reached settlement during its observation period.
The difference between a percentage and an amount is central to the apparent paradox. The global FX market grew from about $1.5 trillion of average daily turnover in 1998 to $7.5 trillion in 2022. Much of the more recent growth came from FX swaps used for funding and hedging, with shorter maturities that must be rolled over more often. A smaller exposed share applied to a much larger and more active market can still produce a very large daily amount.
That is why the right conclusion is neither “FX settlement is unsafe” nor “the problem is solved.” The first ignores the reach of PvP and other mitigants. The second mistakes a lower share for zero exposure. The useful question is how the obligations are distributed across the settlement architecture, and what prevents the remaining trades from entering its strongest protection.
The Real Question Is Coverage
The $1.416 trillion is best understood as the visible edge of a coverage problem. A settlement system can be highly effective for the transactions it accepts while leaving other transactions outside its perimeter. The BIS survey’s hierarchy makes this distinction explicit: protection depends on the method used, not merely on the fact that a trade was arranged by a large bank or involved a familiar currency.
Some of the apparent gap sits inside the nominal perimeter. Of the gross bilateral amount, $347 billion, or 25 percent, was eligible for settlement in an applicable PvP system. The currency pair and product were eligible, and the counterparty had direct or indirect access. Yet those trades still settled bilaterally. The BIS gives operational explanations rather than a complete causal breakdown: an instruction may miss a cutoff, a bank may face difficulties managing credit exposure to correspondent banks used to reach the PvP system, or the system’s payment schedule may be hard to meet when local-currency funding is less liquid.
The larger portion was not eligible for an applicable PvP system. Among those trades, the reported reasons overlap. In 57 percent, the counterparty lacked direct or indirect membership. In 36 percent, the currency pair was not eligible. In 37 percent, the trade type was not eligible. These figures cannot be added to form a pie chart: one trade can have more than one barrier. But they show the shape of the problem. Access is the most common reported obstacle, while currency coverage and product design leave other trades outside the protected path.
The perimeter also varies by relationship. More than four-fifths of the surveyed obligations involved CLS-eligible currency pairs, and 40 percent of those were settled through PvP. For non-CLS-eligible pairs, the PvP share was 12 percent. External inter-dealer settlement used PvP more often than settlement involving other financial institutions or non-financial customers. That pattern does not prove that a counterparty category caused a particular trade to settle gross bilaterally. It does show that membership, currency coverage and the structure of a relationship affect whether the strongest method is available.
The market itself makes the coverage problem harder to freeze in place. The New York Fed describes a more electronic and fragmented FX ecosystem, a rise in trading with other financial counterparties, and greater internalisation by dealers. Principal trading firms and other non-bank financial institutions have become important participants in spot FX. These changes can improve choice and lower transaction costs, but they also mean that a growing share of activity involves participants and execution paths that do not resemble the dealer-to-dealer arrangements around which major settlement infrastructure was built.
This history also shows that the remaining gap is not proof of failed reform. Compared with the 2006 CPSS survey, the share of gross bilateral settlement fell from 32 percent to 15 percent on a basis that excludes intragroup settlement. PvP rose modestly from 54 percent to 56 percent, while pre-settlement netting rose from 12 percent to 23 percent. The comparison has methodological limits, but it records genuine progress. The exposed share declined; the market and the demands placed on its infrastructure changed around it.
The next question is therefore precise. How does conditional payment remove principal exposure, and what should replace or supplement it when the relevant participant, currency, product or operating window cannot use PvP? Answering that requires looking closely at the settlement methods between PvP and gross bilateral settlement—because “mitigated” is not the same as “eliminated,” and because the difference explains why the final 10 percent remains economically important.