Herstatt risk is the risk that you pay one currency in an FX trade and the other party never delivers the currency it owes. Also called FX settlement risk. The payments settle on separate systems, so one side pays first and trusts the other to follow. The loss is the full principal, not a price move.
What happened at Bankhaus Herstatt in 1974?
On June 26, 1974, the German authorities ordered Bankhaus Herstatt into liquidation. The bank closed at 3:30 pm Frankfurt time. Its New York correspondent bank then suspended all outgoing US dollar payments from Herstatt's account, according to the BIS Quarterly Review of June 15, 2026.
Herstatt's counterparties had already paid it Deutsche marks in Frankfurt that day to settle dollar-mark trades. The dollars never arrived, so they were exposed for the full value of the marks delivered.
The BIS describes what followed as a loss of trust: participants delayed payments until they had confirmation of their counterparties' payments, which slowed FX trades to a halt.
Why the full principal is exposed
An FX trade has two legs, one in each currency, and each leg settles in its own country's payment system on its own hours. Nothing ties one payment to the other, and yours becomes final whether or not the counterparty's ever happens.
The exposure window runs from the moment you can no longer recall your payment until you can confirm the other currency has arrived. Three things stretch it:
Time zones. Asia settles before Europe and Europe before New York, so the earlier payer waits.
Cut-offs and holidays. A second leg that misses its cut-off settles the next business day, and a Friday trade can wait until Monday. Windows by rail are in real-time cross-border settlement.
Intermediaries. A payment through correspondent banks can be held at each hop without telling you where the money is.
This is principal risk. Replacement-cost risk is the smaller loss of redoing a trade at a new price after a counterparty defaults, while Herstatt risk is the loss of everything you delivered.
It is also rare. The BIS counts settlement fails at 0.01% of gross obligations settled in April 2025: $121 million among reporting dealers and $1.4 billion with other financial institutions. A fail is a delivery that did not settle on time, which is not the same as a counterparty collapsing, so the figure measures how often legs go wrong, not how large the losses were.
A rate move of 2% costs 2%. A leg that never arrives costs the amount you sent.
PvP and CLS: how the two legs get tied together
Payment versus payment (PvP) links the legs: the first currency is released only if the second is released too. If either side cannot pay, neither leg settles, and nobody loses principal.
CLS is the largest PvP system for FX. The BIS dates its launch to 2002, and CLS describes itself as settling payments in 18 of the most actively traded currencies, with more than USD 8 trillion settled on an average day (as read on CLS's own pages on November 1, 2026).
Other PvP systems cover currencies CLS does not. The BIS lists B3's foreign exchange clearinghouse in Brazil, which has settled the Brazilian real against the US dollar since 2002, CCIL in India since 2015, and CHATS in Hong Kong since 2000. What matters is whether a PvP arrangement exists for your pair and whether you can access it.
How much FX settlement is still exposed?
About a tenth of it, in the latest data. The BIS split more than $14 trillion of average daily FX obligations, from its 2025 Triennial Survey, by how they settled in April 2025.
How FX obligations settled on an average day in April 2025 (BIS Quarterly Review, June 15, 2026)
Settlement method | Share | Value | Effect on settlement risk |
Payment versus payment | 36% | Just over $5 trillion | Eliminated |
Netting, intragroup and in-house methods, timing controls | 54% | About $7.6 trillion | Reduced, not removed |
Gross bilateral settlement | 10% | More than $1.4 trillion | Fully exposed |
The 54% row is a sum of three mitigants, not netting alone. The BIS puts pre-settlement netting at about 15% and intragroup settlement at about 35%, and the rest used bank accounts with timing controls and similar mitigants. All of it is mitigated, not eliminated.
The exposed tenth is not spread evenly. Gross bilateral settlement made up 22% of obligations with non-financial customers and 19% with other financial institutions, against roughly half of that between external dealers.
For CLS-eligible currency pairs, 40% of trades settled through PvP; for pairs outside CLS, only 12% settled through any other PvP system. Of the gross bilateral amount, about three quarters had no PvP route. The most common reason, per the BIS, was a counterparty without direct or indirect access (57% of those trades, with overlapping reasons). The other quarter, about $347 billion a day, was eligible and still settled gross. The BIS points to missed cut-offs and operational problems, so part of the exposure is operational, not access.
For Latin America, CLS lists the Mexican peso among its 18 currencies, though a company paying a supplier does not trade it directly through CLS. The Brazilian real and the Colombian peso are not on that list.
A USDC (a dollar stablecoin) payout settles through none of these systems. A stablecoin (what a stablecoin is) converts and lands in a bank account through a local rail, with a provider in between. The useful question is how long the provider is your counterparty. The liquidity side is in how stablecoin FX liquidity is structured.
In a Lumx off-ramp the provider is that counterparty between the TRANSFERRING_STABLECOIN and TRANSFERRING_FIAT statuses, and no PvP system links the two.
How rails and stablecoins change the exposure window
They shorten it. Herstatt risk is a timing problem, and a payment that settles in seconds leaves less time for a counterparty to fail between the legs.
The local-currency leg is where the rail matters. Pix (Brazil's instant payment system, run by the Central Bank) and SPEI (Mexico's interbank transfer system, run by Banxico) are listed in Lumx's docs as instant and 24/7/365, so on paper a payout over them has no banking-day cut-off. A wire does. The same docs page lists Banxico's non-business days for SPEI, with transactions queued to the next business day, and says Brazilian holidays can still affect Pix dispute windows. These are the cut-offs and settlement targets published in Lumx's docs, read on November 1, 2026:
Cut-off and settlement target by rail, from docs.lumx.io on November 1, 2026
Rail | Cut-off | Settlement target | Operating days |
Pix (BRL) | None | Instant | 24/7/365 |
SPEI (MXN) | None | Instant | 24/7/365, but Banxico non-business days queue to the next business day |
FEDWIRE (USD) | 3:00 PM ET | Same business day, typically within hours | US business days |
ACH (USD) | 2:00 PM ET | 1 to 2 business days | US business days |
SWIFT (USD) | 3:00 PM ET | 1 to 5 business days | Subject to intermediary bank hours |
The targets are not guarantees, and a request after the cut-off, on a weekend or on a holiday is processed on the next business day. More on the calendars in payout cut-offs by rail. Lumx publishes live pages for USDC to BRL and USDC to MXN, and the payment coverage by country page shows which currencies are supported before you read a window into one.
Two things shorten the window. A rail with no cut-off in the docs removes the cut-off wait on the local leg, and a stablecoin settles on a blockchain at any hour, so the dollar leg does not wait either. Neither makes the legs atomic. A Pix credit lives in Brazil's central bank system, outside any blockchain, so a stablecoin-to-bank payout still has an order: one leg first, the other after. Swapping two stablecoins on the same blockchain can be atomic, because both legs live in one ledger.
On the Lumx side, an off-ramp moves through TRANSFERRING_STABLECOIN, then TRADING, then TRANSFERRING_FIAT, then SUCCESS, and each step emits a webhook (offramp.transferring_stablecoin, offramp.trading, offramp.transferring_fiat, offramp.success). The order of the legs is visible to you, and the time between the first status and the last is the window. A failed off-ramp arrives as offramp.failed, and a flagged one is held for an RFI (request for information, a compliance hold that asks for a document before a transaction clears).
A Friday payment, leg by leg
Illustrative, not a quote: a company owes a supplier in São Paulo the reais equivalent of $20,000 and pays at 4:10 pm New York time on a Friday.
Illustrative exposure window for the same payment by two routes
Route | When your leg leaves | When the other leg lands | Who you rely on in between |
USD wire, converted by a bank in Brazil | Friday, but after the 3:00 PM ET cut-off, so the wire is processed Monday | Up to five business days after that, per the SWIFT target above | Every bank in the chain, one after another |
Stablecoin payout over Pix | Friday evening | Minutes on a clean path, because Pix has no cut-off; an RFI or the receiving bank can hold it | Lumx and its banking partner, then the receiving bank |
For up to a week the money in the first route sits in a chain of banks. That is transit and correspondent exposure over the same window, the same shape of risk but not the 1974 mechanism, which is the risk between the two FX counterparties. Neither route is PvP. The difference is the size of the window and the number of counterparties inside it. Cost and speed are in stablecoins versus SWIFT for B2B payments.
I recommend refusing the word "atomic" for any payout that ends in a bank account, whoever is selling it, and asking for the order of the legs instead. I would write the exposure window down in minutes for each rail the company uses, because minutes are what a treasury team can price and compare. I would also ask who holds the funds between the legs and what the provider does when the fiat leg fails, before looking at the rate.
How do you reduce Herstatt risk?
Ways to reduce FX settlement exposure and where each one stops
Tool | What it does | Where it stops |
PvP access | Ties the legs so neither settles alone | Needs direct or indirect access and an eligible currency pair; missing access is the most common reason trades settle gross |
Pre-settlement netting | Cuts several gross amounts to one smaller net amount | Needs enforceable netting agreements, and the net still settles by another method |
Shorter windows | Uses rails with no cut-off and avoids sending before a weekend | Shrinks the time, does not link the legs |
Counterparty limits | Caps how much sits with one provider or bank at once | Caps the loss, not the chance |
Questions about holds | Asks who holds the funds between legs and what happens when a leg fails | Only as good as the written answer |
With Lumx, a locked quote fixes the price, Pix and SPEI have no cut-off in the docs, and each leg sends a webhook, so you can measure the window and set your own cap on what is in flight. None of that links the two legs or removes the risk.
When Herstatt risk is the wrong lens
A single-currency payment has no second leg, so it carries ordinary payment risk. A dealer trading large volume should use CLS or another PvP system, which a payout API does not replace. The lens fits when amounts are large, the window crosses a weekend or the pair has no PvP route.
What a locked quote and a Pix payout look like with Lumx
A locked quote fixes the price, not the order of the legs. Rate movement between quote and execution is market risk, and a lock removes it. The window between the stablecoin leg and the fiat leg is settlement exposure, and a lock does not touch it.
Lumx offers locked rates for 30 seconds, one minute or five minutes. The 30-second lock has no fee, and the longer locks include a fee to cover volatility risk. You request the rate, receive an exchangeRateId, and pass it on the off-ramp request. An expired lock needs a new rate.
Lumx is stablecoin payments infrastructure for businesses that move money between Latin America and the rest of the world: one API to collect, hold, convert, and pay out in BRL, MXN, COP, USD, EUR, and GBP or in USDC and USDT, over local rails such as PIX, SPEI, PSE, Bre-B, ACH, FEDWIRE, SEPA, and Faster Payments, with SWIFT and on-behalf-of payments and collections (POBO and COBO) in USD, EUR, and GBP, plus named virtual accounts, custodial wallets, and KYB/KYC built in.
On a Pix or SPEI payout the fiat leg has no cut-off in the docs, so the exposure that remains is a hold, not a missed window, apart from Banxico's non-business days on SPEI. Transactions flagged by Lumx's monitoring are held for an RFI, and the receiving bank can hold funds for its own review. The docs say Lumx treats an off-ramp as complete once the funds leave its banking partner. On a USD wire rail the window follows the table above.
Related: stablecoin off-ramp, stablecoin corridor and the global payments product page.
Methodology and sources
Read on November 1, 2026: the BIS Quarterly Review of June 15, 2026, "Uncovering FX settlement risk", for the 1974 account, settlement shares, PvP figures, settlement fails and other PvP systems; the CLS Group page on eligible currencies and its About page for the 18-currency count and list; and the Lumx documentation pages on exchange rates, transactions, webhooks, request for information and payment rail cut-off times at docs.lumx.io.
Verified on November 1, 2026. Operational context, not legal, tax, or investment advice.
Cover photo: Venti Views on Unsplash.
What is Herstatt risk?
Herstatt risk is the chance that you deliver your currency in a foreign exchange trade and never receive the currency you bought. It is named after Bankhaus Herstatt, a German bank closed in 1974 before it paid the dollars it owed. The loss is the full principal, not a rate move. Lumx locks the rate on a payout for 30 seconds, one minute or five minutes, which covers the rate side and leaves the window between the legs visible in the off-ramp statuses.
Does CLS cover the Brazilian real or the Mexican peso?
CLS settles 18 currencies and lists the Mexican peso among them. It does not list the Brazilian real or the Colombian peso. Brazil has its own PvP system, B3's foreign exchange clearinghouse, which has settled the real against the US dollar since 2002, according to the BIS. Lumx pays BRL over Pix and MXN over SPEI, neither of which has a cut-off in the docs.
Do stablecoin payments remove Herstatt risk?
They shorten the window but do not remove the risk when one leg is a bank transfer. A stablecoin settles on a blockchain at any hour without waiting for a banking day, and a rail like Pix has no cut-off in the docs, so the time between legs gets short. The two legs still settle in a sequence, so for that time you rely on the provider to deliver. A Lumx off-ramp runs the stablecoin transfer first, then the trade, then the fiat transfer.
How do you reduce Herstatt risk?
Use PvP where you can get access, net trades where the agreements are enforceable, keep the window short, cap what sits with any one counterparty, and ask your provider who holds the funds between legs and what happens when one fails. Lumx gives you a locked quote, Pix and SPEI payouts with no cut-off in the docs, and a webhook for each leg to measure the window. It does not remove the risk.
Who carries the risk in a stablecoin payout?
The sender carries it between the stablecoin leaving their balance and the local currency arriving in the recipient's account, and the payment provider is the counterparty during that window. The provider's holds and its handling of a failed fiat leg matter as much as blockchain speed. With Lumx, a transaction under review is held for an RFI, the offramp.failed webhook reports a failed off-ramp, and offramp.success marks a completed one. The docs add that an RFI left unanswered for 24 hours can end with the transaction refunded to the sender.
