Atomic means gross: the unpaid bill in T+0 currency settlement

On 23 June 2026, Chainlink Labs and two banking consortia used the Point Zero Forum in Zurich to present Project Pangea, a working group set up to move currency trading between the euro and the Korean won onto same-day settlement, cleared through regulated stablecoins. The European side is the Qivalis consortium with 37 banks; the Korean side is the UniKA initiative with more than ten commercial banks.

Fernando Vazquez, President of Capital Markets at Chainlink Labs, put the ambition this way: “Project Pangea upgrades the fragmented foreign exchange model of today with direct, atomic currency swaps using stablecoins.”

Nothing has happened since. There has been no live transaction, no start date named by the participants, no further announcement and no public comment from any supervisor, whether the German Federal Financial Supervisory Authority (BaFin), the European Central Bank (ECB) or Korea's Financial Services Commission. The press release itself gives no date for production. The twelve-month horizon that circulates in coverage of the launch comes from secondary reporting rather than from the original announcement.

On its own that is unremarkable. Working groups move slowly and two months is no time at all in payments. What makes the case interesting is something else: the calculation that decides whether this works has yet to be published by anyone. It has nothing to do with blockchain and everything to do with liquidity.

What actually happens today when two banks settle currency

In June 2026 the Bank for International Settlements (BIS) measured how foreign exchange business is settled in practice, rather than surveying turnover alone. The result reframes the debate.

On an average day in April 2025, gross payment obligations of 14,186 billion US dollars were settled. That figure is not FX turnover, which runs at roughly 9,500 billion US dollars a day; it is the sum of obligations that actually had to be discharged. It breaks down as follows:

• 5,157 billion US dollars (36 per cent) settle through payment-versus-payment systems (PvP), where neither leg pays unless both do. Settlement risk is eliminated.
• 4,950 billion US dollars (35 per cent) settle intragroup, between entities of the same banking group.
• 2,167 billion US dollars (15 per cent) pass through pre-settlement netting.
• 496 billion US dollars (3 per cent) settle across bank accounts where the dealer controls the timing.
• 1,416 billion US dollars (10 per cent) settle on a gross, bilateral basis, with no mitigation whatsoever. This is the share that matters for systemic risk.

The BIS draws a firm conclusion: “In April 2025, 90% of the average daily settlement was via methods which eliminate or minimise FX settlement risk.” The existing plumbing works considerably better than its reputation suggests.

The netting effect is the number that decides everything

Buried in that breakdown sits a figure absent from every Project Pangea announcement, and it is the most important one in the whole venture. Of the 2,167 billion US dollars that pass through netting each day, 337 billion US dollars remain to be paid once the offsetting is done.

That is a reduction of 84 per cent. Netting is not a marginal technique. It is the reason global currency trading functions on a fraction of the liquidity its notional volumes would imply, and the same logic underpins PvP systems such as Continuous Linked Settlement (CLS), which also offset before settling.

The conflict follows directly. An atomic swap settles each trade individually, on a gross, real-time basis. By construction it cannot net, because netting means waiting and combining. Move to same-day atomic settlement and you remove settlement risk at the price of an intraday liquidity requirement: both stablecoins have to sit fully funded across the trading day.

For a treasurer that is the number against which the venture will be judged. It appears nowhere in the public communication. No one has put a number on that requirement in public.

The won is CLS-eligible. It still rarely settles through PvP

A common misreading holds that the Korean won cannot be settled through Continuous Linked Settlement. It can. The same BIS study lists the eligible currencies: “As of June 2026, CLS-eligible currencies are: AUD, GBP, CAD, DKK, EUR, HKD, HUF, ILS, JPY, KRW, MXN, NZD, NOK, SGD, ZAR, SEK, CHF and USD.” The won has qualified since 2004, with the Bank of Korea as its central bank.

The real gap sits elsewhere and can be measured precisely. More than four fifths of daily settlement, some 12,200 billion US dollars, involves currency pairs that are eligible in principle. Only 40 per cent of that volume actually settles through PvP systems. For pairs that are not eligible the figure is 12 per cent.

Eligibility has long since been granted for the won. What goes unused is the eligibility itself, and that distance is created by membership requirements, third-party access arrangements, cut-off times and each bank's own nostro structure. Pangea targets that gap, and rightly so. What it describes, though, is a problem of access and process rather than one of technology.

What is genuinely difficult about the won

Korea's currency is only partially convertible. Offshore trading therefore runs largely through non-deliverable forwards, settled in cash rather than by delivery, with London as the largest venue. Transfers abroad above one million US dollars require approval under Korea's Foreign Exchange Transactions Act.

One development bears directly on the timetable: from December 2026 cross-border crypto transfers also fall under Korean foreign exchange control and must pass through the Bank of Korea's reporting network. An atomic swap in a won stablecoin has to fit into that reporting regime, which adds friction rather than removing it.

The trade corridor is real. Goods trade between the European Union and South Korea came to 124.25 billion euros in 2025. The figure of 150 billion US dollars that circulates in specialist coverage probably rests on a wider definition that includes services.

Four questions that come before the technology

Assessed from a bank's perspective, the venture raises points that the communication so far leaves untouched.

First, whose credit risk? A stablecoin swap removes the risk that one side pays while the other fails to deliver. In its place comes a claim against the issuer of each token and against the assets backing it. The risk does not disappear; it changes address.

Second, which regime applies? The euro stablecoin planned by Qivalis would be an e-money token under the Markets in Crypto-Assets Regulation (MiCA), which carries a right to redemption at par and imposes requirements on the reserve. That is the starting point for accounting and prudential treatment.

Third, is settlement final? A smart contract that has executed is not automatically final within the meaning of Settlement Finality Directive 98/26/EC, to which German law refers in section 1 (16) of the Banking Act for its definition of a system. Whether a bank may derecognise the position depends on that answer.

Fourth, what does the headline metric tell us? The consortia put the combined assets under management of the participants at more than ten trillion US dollars, by their own account. Assets under management say nothing about currency settlement volume. The figure that would matter is these institutions' share of the euro-won payment flow, and nobody publishes it.

The participants are spreading their bets

How the Korean members behave puts the working group's binding force into perspective. Kbank has been running its own cross-border payments venture with Ripple since April 2026, aimed at the corridors to the United Arab Emirates and Thailand. Jeonbuk Bank adopted Ripple Payments on 18 August 2026, independently. Neither has any connection to Pangea.

None of that argues against the participants. It shows that they are keeping several routes open, and it suggests reading membership of the working group as exploration rather than commitment.

Project Agora at the BIS, a central-bank-led venture with a smaller field of seven central banks and more than 40 institutions, offers a yardstick for how long this takes: the path from prototype to testing with real value ran until May 2026.

The silence of the supervisors

For an initiative of this size, no supervisory authority and no named practitioner from the currency markets has commented publicly. The scepticism that can be found comes from crypto analysts and addresses token valuations around the project rather than the settlement mechanics.

That absence is itself informative. A working group that has settled nothing gives supervisors little to respond to. Read the other way round: the regulatory assessment of this model is still entirely outstanding.

Recommendations