Herstatt Risk: The Inseparability of Liquidity and Counterparty Risk

On June 26, 1974, a mid-sized German bank called Bankhaus Herstatt did something that nearly annihilated global finance: It failed.

It didn’t even have the decency to do it spectacularly; it simply ran out of money. At 3:30pm local time, German regulators shut it down. That moment was the proverbial butterfly whose wing-flaps make a hurricane.

Several of Herstatt's counterparties had already sent Deutsche Marks to the bank that day, expecting to receive US dollars in return. But because of the time difference between Frankfurt and New York, the dollar leg of those trades hadn't settled when regulators pulled the plug.

The resulting chaos led to a near-seizure of the New York clearing system and gave this particular variety of danger its permanent name: Herstatt risk.

Fifty years later, that term still describes the same fundamental problem: In any foreign exchange transaction, the two legs of the trade settle at different times, in different places and through different systems. The window in-between is where things can go spectacularly wrong.

Fragmentation ups the ante

The FX market has evolved considerably since Herstatt. CLS Bank, the Continuous Linked Settlement system established in 2002, was designed to address settlement risk by synchronising the two legs of a trade, ensuring that payment in one currency only happens if payment in the other currency also occurs. For the currencies and institutions it covers, it works well. Daily CLS settlement volumes regularly exceed $6 trillion.

But CLS covers just 18 currencies, whereas there are 180-odd currencies in active global use. Everything outside that charmed circle still settles the old way, sequentially, with full Herstatt-style exposure sitting between execution and settlement..

Here, liquidity fragmentation and counterparty risk become inseparable. The more fragmented your liquidity, the more counterparties you're dealing with across more jurisdictions and in more currency pairs, the more settlement exposures you're carrying.

A payment service provider moving money across several emerging market corridors might be dealing with a dozen counterparties: Local banks in Nigeria, Brazil, the Philippines, a regional broker in the Gulf, a non-bank liquidity provider for Southeast Asian currencies. Each relationship involves a settlement window, and each settlement window represents capital that is at risk as long as it remains in flight.

To manage these exposures prudently, institutions must hold buffers against the possibility that a counterparty fails during a settlement window. The more counterparties, the more buffers. The more buffers, the more capital tied up doing nothing except sitting there. In thin-margin businesses like payments and remittances, this all translates into cost.

Unlike 1974, today's fragmented market operates at extraordinary speed and scale. The volume of trades that can accumulate settlement exposure in a single day dwarfs anything Herstatt's counterparties dealt with. The infrastructure is faster and more interconnected, which is great, but when something does fail, the knock-on effects can propagate further, faster.

So we’ve improved the controls. But the architecture, the source of the problem, is still with us.

What stablecoins can and can’t change

When studying settlement risk, it’s tempting to consider stablecoins as a clean solution, in part because of atomic settlement. If both legs of a trade settle simultaneously on-chain, the window between them disappears. No window, no Herstatt risk. Problem solved. Right?

Atomic settlement is when a transaction either completes fully or not at all, with no intermediate state. This genuinely does eliminate the specific risk that Herstatt exposed. If you're swapping one stablecoin for another on a properly designed system, there is no moment where you've sent your side but haven't received theirs. Simultaneity is enforced by the protocol.

But stablecoins don't eliminate counterparty risk, they merely relocate it.

The question shifts from "will my counterparty deliver the second leg of this trade?" to "is this stablecoin actually worth what it says it is?"

In 1974, the risk was Herstatt's solvency. With stablecoins, the analogous risk is the issuer's solvency, the quality of their reserves, and whether redemption actually works under stress.

We've seen what stablecoin stress looks like. The 2022 collapse of TerraUST, which lost its dollar peg catastrophically, wiped out tens of billions in value in days. That wasn't Herstatt risk, but it was still counterparty risk, repackaged and made modern.

Even well-designed, fully-reserved stablecoins carry basis risk, the possibility that the stablecoin trades at a discount to its peg during periods of market stress. Stablecoins also come packing smart contract risk, the possibility of bugs or exploits in the settlement infrastructure itself. This type of operational risk doesn’t exist in traditional settlement systems.

None of this is an argument against stablecoins. For certain corridors, particularly those outside CLS coverage where traditional settlement infrastructure is thin, atomic settlement via stablecoins represent genuine improvement over sequential settlement with multiple counterparties. What you need to understand is that the risk profile is different.

Every settlement system makes a choice about where to place potential failures. Traditional FX settlement puts it in the settlement window, which CLS compresses for major currencies. Stablecoins relocate it to the issuer.

Herstatt happened because a bank failed during a settlement window that the market's infrastructure wasn’t designed to handle. What we draw from this isn’t that we need to find a hermetically risk-free system; it’s that we need to understand where in a system our risk lives, and whether we can manage it.