Why "Safe" Currency Trades Can Still Lose Money Overnight

The Two Hours That Broke a Market

Picture a Tuesday afternoon in Frankfurt, June 1974. A small bank called Bankhaus Herstatt is closed by regulators at 3:30 pm local time. Its New York correspondent bank then stops sending out dollars from Herstatt's account. But here's the problem: Herstatt's trading partners had already paid their Deutschmarks that morning, expecting dollars back later the same day. They got nothing. They lost the full value of what they'd sent.

That single afternoon gave the world a term — "Herstatt risk" — and it's the clearest illustration of foreign exchange (FX) settlement risk. In plain terms: FX settlement risk is the chance that you deliver the currency you sold but never receive the currency you bought. The full trade value is on the line, not just a small slice of it.

The fallout didn't stop at those counterparties. Confidence in who would pay whom froze up. Market participants started holding back payments until they'd confirmed their counterparties had paid. Cross-border payments stalled, and central banks eventually had to signal they'd supply liquidity to calm things down. A single bank failure nearly locked up the plumbing of global finance.

That's why regulators and banks have spent the last fifty years building protections — and why it's worth understanding which ones actually do something.

Where the Risk Lives: Not Quite Two Days, Not Quite One Moment

Settlement risk isn't a single instant. It's a window that opens and closes. It opens the moment your payment instruction for the currency you sold can no longer be cancelled on your own — no take-backs. It closes when the currency you bought lands in your account with finality: received, irrevocable, unconditional.

Between those two points, you're exposed. That window can run one to two business days, with another day or two of "did it actually arrive?" uncertainty layered on top. In the 1990s, supervisors found that many banks were treating this exposure as a trivial intraday blip — assuming it would resolve before the day ended. It often doesn't. Exposure can sit open overnight, through weekends and holidays, when time zones mean your counterparty's markets are shut and yours are open. During that gap, you're carrying an unhedged position, which means exchange-rate moves can pile onto the original loss.

The exposure isn't small either. The 2025 BIS Triennial Survey measured more than $14 trillion in gross payment obligations settling on an average day in April 2025. Of that, roughly $1.4 trillion — about 10% — settled on a gross bilateral basis, which is the fully exposed, unprotected route. Twenty years earlier, that share was 32%. So the trend is genuinely good, but the absolute number is still enormous.

What Payment Versus Payment Does — and Doesn't

Here's the elegant part. Payment versus payment, or PvP, is a mechanism where the final payment of one currency happens if and only if the final payment of the other happens. The two legs are linked. Nobody pays unless everybody pays.

Think of it as a handshake where both hands only close at the same moment. Swap the cash for the cash, or the whole thing unwinds.

The largest and best-known PvP system is CLSSettlement, launched in 2002. It settles payments in 18 of the most actively traded currencies, and it doesn't just remove settlement risk — it multilaterally nets payments, which shrinks the funding needed to settle by over 96%. Others exist too: B3 in Brazil, CCIL in India, CHATS in Hong Kong SAR, Buna in the UAE. Each covers different currency pairs and different local market participants.

Now the important caveat, because this is where people quietly get it wrong. PvP eliminates settlement risk. It does not eliminate replacement cost risk — the risk that you have to go back into the market at current prices to replace a trade that didn't happen, potentially at a worse rate. It also doesn't eliminate liquidity risk: the risk of not receiving a currency when you expected it, which can strain your funding even if you lose nothing on principal. PvP is not the same as a central counterparty, which steps in as the counterparty to everyone. Different tool, different job.

Time Zones, Cutoff Times and the Trades That Slip Through

Here's a genuinely useful fact about the 2025 survey: of the $1.4 trillion settling on a gross bilateral basis, about 25% — roughly $347 billion per day — was actually eligible for a PvP system. The currency pair qualified. The counterparty had access. It just didn't happen.

Why? Operational friction, mostly. Trades can miss the cutoff time for submitting settlement instructions. PvP systems generally operate on specific cutoffs designed for next-day or two-day settlement, so same-day trades often can't use them at all. Managing credit exposure to the correspondent banks used as intermediaries adds another wrinkle. And here's a forward-looking one worth noting: as more jurisdictions move to one-day securities settlement cycles, demand for same-day FX swaps to fund cross-border securities trades may rise — and those trades currently sit outside PvP's reach. So this particular leak in the pipe could widen.

The bigger chunk, three-quarters of gross bilateral settlement, isn't eligible at all. About 57% of that was because the counterparty simply lacked direct or indirect access to a PvP system — more common with non-bank financial institutions and corporate customers than between dealer banks. About 36% was because the currency pair wasn't eligible, which is the story of emerging market currencies. And about 37% was because the trade type wasn't eligible. These overlap, so they don't sum neatly.

Widening currency coverage isn't just a technical upgrade. It requires robust legal frameworks that guarantee settlement finality, workable convertibility, reliable payment infrastructure, and willing liquidity providers in local currencies. That's a capital markets modernisation project, not a software update.

The Middle Ground: Netting and Timing Controls

Between the full protection of PvP and the full exposure of gross bilateral settlement sits a middle layer. In April 2025, about $7.6 trillion per day — 54% of total settlement — went through methods that reduce but don't eliminate risk.

Pre-settlement netting is the big one. Instead of paying each trade separately, two counterparties offset multiple matching trades to arrive at a single smaller net amount per currency. In April 2025, netting reduced roughly $2.2 trillion in average daily obligations down to $337 billion. That's substantial compression. But two caveats matter: the netting agreement has to be legally enforceable if your counterparty defaults, and the remaining net amount still has to settle somehow — often by traditional correspondent banking, which carries its own counterparty credit risk. Also, netting is likelier to compress inter-dealer flows than customer flows, simply because dealers' positions are more bidirectional while customers tend to be directional.

Intragroup settlement covers trades between entities in the same banking group. The 2025 survey found 35% of daily settlement — over $4.9 trillion in inter-dealer trades — moved this way. The odds of trouble are low because a group has strong incentives to coordinate internally. But in severe stress, liquidity can get trapped inside a single entity through ring-fencing measures or payment moratoriums, and that trapped liquidity can disrupt the whole group's internal payment sequence.

Settlement timing controls are the smallest category at 3%. Here a dealer effectively settles only once the other side has paid with finality or posted collateral. Sensible for the dealer. Less symmetric for the client, who may not have equivalent controls — and in stressed markets these controls can amplify delays and reduce trading activity, which is worth being honest about.

How Much Goes Wrong

The 2025 survey collected settlement fail data for the first time: trades due to settle in April 2025 that were still unsettled at month end. Fails accounted for just 0.01% of gross obligations. Among reporting dealers, $121 million; with other non-reporting financial institutions, $1.4 billion. That gap likely reflects different levels of operational sophistication and controls.

Tiny percentage, real money. And it's worth watching, because shorter settlement cycles reduce the window to fix confirmation or settlement problems.

The pre-settlement side of this story deserves its own mention, because it's where the risk is easiest to underestimate. Before any of these settlement protections apply, you've already committed to a trade — you've agreed a price, a currency pair and a value date, and you're now carrying exposure to your counterparty until settlement completes. That pre-settlement window is a real credit exposure in its own right, and netting agreements only help if they're legally enforceable in the counterparty's jurisdiction. Jurisdiction matters enormously here. Netting that holds up in one country may not in another, and that difference can decide whether protection actually works when you need it.

The Honest Bottom Line

The architecture that exists today — PvP systems, netting services, supervisory guidance, the FX Global Code's settlement hierarchy — is the product of fifty years of deliberate, mostly successful work. The gross bilateral share has more than halved since 2006. That's a real achievement.

But PvP adoption itself has risen only modestly, from 54% of average daily settlement in 2006 to 56% in 2025. The gains came mostly from netting, which mitigates rather than eliminates. And the rising share of non-CLS-eligible currencies and the growing role of non-bank financial institutions both push in the other direction.

One more thing worth sitting with: in ordinary times, everyone's estimates look roughly fine. In a crisis, they tend to be wrong in the same direction — and that's precisely when it matters most. Nobody can tell you with certainty how these protections will hold under conditions we haven't seen yet, and rules differ meaningfully by jurisdiction. What you can do is understand which layer you're relying on for any given trade, and what that layer actually protects.