Chapter 3 — Settlements: The Rise of Financial Plumbing

A small electronics manufacturer in Singapore needs to pay a supplier in Frankfurt $47,000 for a shipment of circuit boards. The manufacturer walks into their bank on a Monday morning, fills out a wire transfer form, and waits.

Three days later, the money arrives. The manufacturer's bank charged $45 in fees. The supplier's bank took another €15. Somewhere in between, the exchange rate slipped just enough to cost an extra $200 that nobody warned about.

The manufacturer shrugs. This is how international payments have always worked — slow, expensive, and a little opaque. But why? The money isn't traveling by cargo ship. It's not being loaded onto pallets. In a world where a text message reaches Frankfurt in milliseconds, why does the payment take three business days and cost over $260?

The payment doesn't travel through one pipe. It hops through a chain of banks, and every hop adds time, fees, and a slice of the exchange rate. This chapter explains why the chain exists.

Well, the answer isn't technical. It's architectural. The money doesn't flow through one pipe. It flows through a chain of pipes — built over 150 years, by different institutions, in different countries, for different reasons. Some of those pipes are state-of-the-art. Others were designed before the internet existed. And all of them have to cooperate, however imperfectly, every time money crosses a border.

This chapter is about those pipes: how they were built, where they leak, and why the plumbing of global finance looks the way it does today.

The Three-Layer Model: Messages, Obligations, and Finality

In Chapter 1, we established that money is fundamentally a system of trust — a shared belief that the number in your account means something. In Chapter 2, we saw how banks became the custodians of that trust domestically. Now we go international, where the trust problem gets dramatically harder.

Before we trace a payment through the system, you need a mental model that will serve you for the rest of this book. Every payment system in the world — from a Venmo transfer to a $5 billion interbank wire — operates across three layers.

Think of it like a group of friends splitting a restaurant bill. First, someone texts the group chat: "You owe me $30 for dinner" — that's the message. Then everyone checks their records and agrees on who owes what — that's the clearing. Finally, people actually send the money — that's the settlement.

In payments, these three layers have formal names and precise meanings. They are:

1. Messaging: The instruction layer. A bank sends a message to another bank saying: "Please move $47,000 from our account to your client's account." The message contains all the details — amount, currency, sender, recipient, purpose — but no money moves yet. It's just information. Here's something that surprises almost everyone: SWIFT — the Society for Worldwide Interbank Financial Telecommunication, the system most people associate with international payments — operates only at this layer. It carries messages. It doesn't move a single dollar.
2. Clearing: The accounting layer. This is where the institutions involved figure out who owes what to whom. In a simple two-party transaction, clearing is trivial: Bank A owes Bank B exactly $47,000. But when hundreds of banks are sending thousands of payments to each other every day, clearing becomes an optimization problem. If Bank A owes Bank B $100 million and Bank B owes Bank A $93 million, do they really need to move $193 million? Or can they net it down to a single $7 million payment? That netting — calculating the smallest set of actual transfers needed to settle all obligations — is the heart of clearing.
3. Settlement: The finality layer. This is where real money moves — where a bank's balance at the central bank actually changes. Settlement is irrevocable. Once it's done, it's done. No chargebacks, no provisional credits, no "we'll see." The payment is final.

This three-layer model matters because the layers can be separated in time, space, and ownership.

A Belgian cooperative runs the messaging system (SWIFT). A private-sector utility in New York handles the clearing (the Clearing House Interbank Payments System, or CHIPS). And the settlement happens on the books of the Federal Reserve (Fedwire). Three different organizations, three different rule books, three different time zones — all cooperating to move your $47,000.

Table 1: The three-layer model. Every payment system in this book — cards, wires, ACH, real-time payments — maps onto these three layers. When you encounter a new system, the first question to ask is: how does it handle messaging, clearing, and settlement?

This framework will reappear in every chapter that follows.

When we talk about card payments in Part II, you'll see the same three layers — authorization is messaging, clearing is the batch file exchange, and settlement is the money transfer between banks.

When we talk about real-time payment systems in Part V, the innovation is often about collapsing these layers together so they happen simultaneously. But the layers themselves are always there.

Now let's see them in action. The most common way money crosses borders — and one of the oldest pipes still in use — is correspondent banking.

Correspondent Banking: The Original Global Pipeline

Here's a problem that has existed since the 16th century.

Your bank is in Singapore. Your supplier's bank is in Germany. Those two banks don't have accounts with each other. They may never have transacted before. How does the money get from one to the other?

The answer is correspondent banking — a system where banks maintain accounts with each other specifically to facilitate payments on behalf of their customers. These aren't regular checking accounts. They're pre-funded pools of money that sit at the other bank, waiting to be drawn on. These systems originated out of the age of Mercantilism (from the 16th to the 18th century), where there was a need to trade between merchants and borders of different countries and empires.

The terminology is simple once you see it from both sides. When your bank in Singapore holds an account at a large international bank in New York, that account is called a nostro account from Singapore's perspective — "our money held at your bank" (nostro means "ours" in Italian). The same account, viewed from the New York bank's perspective, is a vostro account — "your money held at our bank" (vostro means "yours").

Let's trace the Singapore-to-Germany payment step by step.

1. The Singapore manufacturer instructs their local bank (DBS, let's say) to send $47,000 to the supplier's account at Deutsche Bank in Frankfurt.
2. DBS doesn't have a direct relationship with Deutsche Bank. But DBS does have a nostro account — denominated in US dollars — at JPMorgan in New York. And JPMorgan has a euro-denominated nostro account at Deutsche Bank. This chain of relationships is the corridor.
3. DBS converts the manufacturer's Singapore dollars to US dollars and debits its internal records. It then sends a SWIFT message (an MT103 payment instruction — or now, its ISO-20022 equivalent) to JPMorgan, saying: "Please debit our nostro account and forward $47,000 to Deutsche Bank for credit to the supplier's account."
4. JPMorgan receives the message, debits DBS's nostro account in dollars, converts the dollars to euros, and sends a SWIFT message to Deutsche Bank with the payment instruction.
5. Deutsche Bank receives the message, credits the supplier's account with $47,000, and adjusts its internal records to reflect the debit to JPMorgan's euro nostro account.

The supplier has their money. The manufacturer's account has been debited. And the actual settlement — the movement of real dollars between DBS and JPMorgan — happens later that day through Fedwire or CHIPS, the US wholesale payment systems.

A cross-border correspondent banking payment. Notice how the money doesn't travel directly from Singapore to Germany. It hops through intermediary banks, each maintaining pre-funded accounts with the next. Every hop adds time, cost, and a point where something can go wrong.

This system works. It's been working, in various forms, for centuries. But it has two structural problems that make it expensive and slow. They are:

1. Time zones and FX: The Singapore-to-Germany payment touches three time zones and two currency conversions:
   • DBS operates on Singapore time (UTC+8).
   • JPMorgan operates on New York time (UTC−5).
   • Deutsche Bank operates on Frankfurt time (UTC+1).

   The windows during which all three banks are open and their payment systems are processing overlap by only a few hours. Miss the window, and your payment sits in a queue until the next business day.

   Each currency conversion is also a point of friction. The manufacturer pays in Singapore dollars; JPMorgan converts to US dollars; then euros. Each conversion involves a spread — the difference between the buy and sell rate — which is how the intermediary bank earns a margin on the transaction. Two conversions, two spreads, and neither is visible to the manufacturer until the money arrives.

2. Compliance and network shrinkage: Every bank in the chain must verify that the payment complies with anti-money-laundering (AML) and sanctions rules. JPMorgan screens the payment against US sanctions lists. Deutsche Bank screens it against EU lists. This compliance work is necessary, but it adds processing time and, critically, it makes correspondent banking relationships expensive to maintain.

   The cost of compliance has led to a phenomenon called de-risking — large international banks cutting correspondent relationships with banks in smaller or higher-risk countries because the compliance costs outweigh the revenue from processing their payments.

   The World Bank has documented this trend: the number of active correspondent banking relationships declined by roughly 20% between 2011 and 2022. For banks in the Pacific Islands, parts of Africa, and the Caribbean, losing their correspondent relationship can mean losing access to the global payment system entirely.

Correspondent banking, for all its flaws, remains the backbone of cross-border payments. But it has a vulnerability that nearly brought it all down, in the summer of 1974.

So what was it, and why did it happen?

Herstatt Risk: The Day the Pipe Leaked

On the morning of June 26, 1974, several banks around the world were in the middle of foreign exchange (FX) trades with a small German bank called Bankhaus Herstatt. The trades were routine — Deutschmarks for dollars, settled in two legs. The German leg would settle in Frankfurt. The dollar leg would settle in New York.

By midmorning in Frankfurt, the German side of the trades had completed. Banks had delivered their Deutschmarks to Herstatt through the German payment system. The dollars were supposed to flow the other direction — from Herstatt's accounts in New York — later that same day, when New York's banking day opened.

At 3:30 PM Frankfurt time — 10:30 AM in New York — German regulators pulled the plug. They revoked Herstatt's banking license and ordered it liquidated. Immediately.

The banks that had already paid their Deutschmarks were frozen. The New York payment leg hadn't been processed yet. Herstatt's New York correspondent, Chase Manhattan, received the closure notice and stopped all outgoing dollar payments from Herstatt's account. Banks that had delivered their Deutschmarks hours earlier — some with exposures as high as $156 million — were left holding nothing.

The contagion was swift. CHIPS, the interbank clearing system that processed the dollar payments, effectively seized up. Banks stopped sending payments to each other because nobody knew who else had exposure to Herstatt. For months afterward, the global clearing system operated at reduced efficiency, with banks demanding immediate settlement instead of the standard next-day processing.

The failure exposed a vulnerability hiding in plain sight: settlement risk in FX transactions.

When you trade one currency for another, the two legs settle in different countries, in different time zones, through different systems. Between the moment you pay one leg and receive the other, you're exposed. If the counterparty fails during that gap — as Herstatt did — you lose everything you've paid and receive nothing in return.

This risk — now formally called Herstatt risk, or more broadly, FX settlement risk — isn't a theoretical concern. The Bank for International Settlements (BIS) estimates that trillions of dollars in FX trades remain exposed to settlement risk on any given day. The gap between paying and receiving can last from a few hours to more than a full business day.

The fix took nearly three decades to build. The concept is called Payment versus Payment (PvP): both legs of an FX trade settle simultaneously, or neither does. If one side can't deliver, the whole transaction rolls back. No more paying your Deutschmarks and hoping the dollars show up.

PvP found its institutional home in CLS Bank, which began operations in 2002. CLS acts as a settlement utility for FX trades, holding both legs of the transaction in escrow and releasing them simultaneously. Today, CLS settles an average of $7.9 trillion per day across 18 currencies, with over 70 direct members and more than 37,000 third-party participants.

The Herstatt timeline. The gap between 10:30 AM New York time (when Herstatt was closed) and the point at which dollar payments would have settled left counterparties exposed. Banks had paid Deutschmarks but would never receive dollars.

Diagram 4: Payment versus Payment (PvP) via CLS Bank. Both sides submit their currency to CLS. Only when both legs are confirmed funded does CLS release the payments simultaneously. If either side fails, both get their money back. This is the engineering fix for Herstatt risk.

CLS didn't eliminate all FX settlement risk — not every currency pair is eligible, and some trades still settle outside CLS. But it turned the Herstatt scenario from an ever-present danger into a manageable, monitored exposure. The principle of PvP has since been applied to other settlement systems around the world.

Herstatt's failure also had a lasting institutional legacy. Within months, central bankers created the Basel Committee on Banking Supervision — the body that would go on to write the capital adequacy rules (Basel I, II, III) that govern banks worldwide. A $620 million German bank that most people had never heard of ended up reshaping the entire architecture of global financial regulation.

The Basel Committee doesn't make laws. It issues standards — recommendations that member jurisdictions are expected to translate into national regulation. Its membership has grown from 10 central banks in 1974 to 45 institutions across 28 jurisdictions today, covering the world's major financial centers. The Committee's core mandate is simple: ensure that banks hold enough capital to absorb losses without requiring government bailouts. What sounds like a narrow technical brief has, over five decades, produced one of the most consequential bodies of financial rulemaking in history.

Each iteration of the Basel framework was a direct response to a crisis or a recognized gap — not an exercise in proactive design. Basel I was a product of Cold War anxiety and the Latin American debt crisis. Basel II was the profession's attempt at sophistication, which failed spectacularly in 2008. Basel III was the reckoning that followed. Each version added layers — more capital, more nuance, more complexity — but the underlying architecture remained the same: banks must hold a cushion of loss-absorbing equity calibrated to the riskiness of their assets.

For payments specifically, Basel matters because the banks that move money through Fedwire, CHIPS, and correspondent networks are subject to these rules. Capital requirements affect how much liquidity banks are willing to pre-fund in their nostro accounts. Liquidity coverage ratios determine whether a bank can sustain a 30-day stress scenario — including a disruption to its payment flows. The plumbing and the regulation are not separate systems. They co-evolved.

Table: The Basel framework at a glance. Each iteration tightened the definition of qualifying capital and added new dimensions of risk measurement. The consistent theme: crises drove reform, not foresight.

The pipes were getting stronger. But they still needed a common language to talk to each other. That's the story of SWIFT — and it starts with one fundamental misconception that persists to this day.

SWIFT: The World's Messaging Backbone (And Nothing More)

Ask most people what SWIFT does, and they'll say it moves money between banks. It's one of the most persistent misconceptions in finance. Headlines reinforce it: "Russia cut off from SWIFT" sounds like Russia's money pipeline was severed. The reality is more nuanced — and understanding the nuance is essential to understanding how cross-border payments actually work.

SWIFT doesn't move money. It is a messaging framework. It carries messages. That's it.

The Society for Worldwide Interbank Financial Telecommunication (SWIFT) is a cooperative, headquartered in Belgium, that operates a secure messaging network connecting over 11,000 financial institutions in more than 200 countries. When DBS in Singapore sends a payment instruction to JPMorgan in New York (as in our correspondent banking example), that instruction travels over the SWIFT network. But the actual money moves separately, through the wholesale settlement systems we'll discuss in the next section.

Think of SWIFT as the postal service for banks. It delivers the letter that says "please move $47,000." It doesn't deliver the cash. The distinction matters because cutting a bank off from SWIFT doesn't freeze their money — it makes them unable to communicate efficiently with other banks.

They could, in theory, send payment instructions by email, fax, or carrier pigeon. It would be slow, error-prone, and operationally painful, but the money itself isn't locked.

Decoupling from SWIFT: An uncertain future?

That said, disconnection from SWIFT is devastating in practice. In February and March 2022, the EU Council disconnected several Russian banks from SWIFT as part of sanctions following the invasion of Ukraine. More banks were added in subsequent rounds. Without SWIFT, those banks lost their ability to process cross-border payments at any meaningful scale. The sanctions worked not because they froze money directly, but because they severed the messaging infrastructure that makes modern banking possible.

SWIFT's other major story is the migration to ISO-20022 — a new messaging standard that replaces the legacy MT format that banks had used for decades. The coexistence period ended on November 22, 2025, and all cross-border payments on the SWIFT network now use the richer, more structured ISO-20022 format. The old MT103 (customer credit transfer) and MT202 (bank-to-bank transfer) messages have been formally retired.

For reference, here is a working glossary of the most common MT message types you will encounter in cross-border payments.

The MT numbering scheme is not arbitrary: the first digit identifies the category (1xx = customer payments, 2xx = financial institution transfers, 3xx = FX and derivatives, 9xx = account statements and queries). Within each category, the remaining digits identify the specific message function.

Table: Common SWIFT MT message types. All MT codes were formally retired on 22 November 2025 and replaced by their ISO 20022 equivalents (pacs, camt, pain message families). The table above reflects the legacy taxonomy that remains essential context for understanding how cross-border payment instructions were structured and routed for five decades.

Why does the format matter? Because ISO-20022 messages carry significantly more data than their MT predecessors. A payment instruction can now include structured remittance information, purpose codes, and richer identification fields — all of which reduce the manual investigation work that banks do when payments are flagged for compliance review. The richer data also means better straight-through processing: fewer payments stuck in queues, fewer phone calls between operations teams, fewer days of delay.

SWIFT itself is governed by its member banks and overseen by the National Bank of Belgium, with cooperative oversight from G10 central banks. This governance structure matters because SWIFT's decisions about who can and cannot use its network are subject to both regulatory pressure (like EU sanctions) and the cooperative interests of its members.

To conclude, messaging and settlement are different layers, operated by different entities, governed by different rules. You can restrict one without directly touching the other — but in practice, restricting the messaging layer makes the settlement layer nearly unusable.

So if SWIFT carries the messages, who actually moves the money? That's the job of the wholesale payment systems — the big pipes.

The Big Pipes: Fedwire, CHIPS, and CIPS

Every currency has at least one system for settling large-value payments between banks with finality. These are the wholesale pipes — the infrastructure that moves trillions of dollars, euros, yen, and renminbi between financial institutions every day. Three systems illustrate the design choices that matter.

1. Fedwire, USA: Fedwire is the United States' real-time gross settlement (RTGS) system, operated by the Federal Reserve. Every payment on Fedwire settles individually, in real time, with immediate finality. When JPMorgan needs to move $500 million to Citibank, the Federal Reserve debits JPMorgan's reserve account and credits Citibank's reserve account — right now, irrevocably. There's no netting, no batching, no waiting. In 2024, Fedwire processed an average of 836,000 transactions per day, with a daily value of approximately $4.5 trillion.

   The advantage of RTGS is certainty: the recipient knows the money is final the moment it arrives. The disadvantage is liquidity: every payment requires the sender to have the full amount sitting in their reserve account at the moment of transfer. That's a lot of cash to keep on hand.

2. CHIPS (The Clearing House Interbank Payments System), USA:

   CHIPS takes a different approach. Operated by The Clearing House (a private-sector consortium of major US banks), CHIPS uses a netting algorithm to dramatically reduce the amount of cash that actually needs to move. Instead of settling each payment individually, CHIPS continuously calculates the net positions of its participants and settles the smallest possible amounts.

   The efficiency is remarkable. CHIPS processes approximately $1.9 trillion per day, but its patented netting algorithm achieves a ratio of 29:1 — meaning $1 of funding supports $29 in settled payment value. That frees up enormous amounts of liquidity that banks would otherwise need to hold in reserve. The tradeoff is that CHIPS payments aren't individually final in real time; they're final when the netting cycle completes, typically by end of day.

3. CIPS (The Cross-Border Interbank Payment System), China: CIPS is China's answer to the question: how do you settle renminbi payments internationally? Launched in 2015 and operated by the People's Bank of China, CIPS uses a hybrid model: it offers both RTGS for urgent payments and deferred net settlement (DNS) for standard flows. CIPS operates on an extended schedule — roughly 5×24+4 hours to cover time zones from Asia through Europe to the Americas — and connects to banks in over 110 countries.

   CIPS matters because it's the infrastructure backbone for the renminbi's growing (if still modest) role in international payments. As we'll see in the next section, China's push to internationalize the RMB requires plumbing that doesn't depend on the US-centric correspondent banking network.

4. TARGET2 (Trans-European Automated Real-time Gross Settlement Express Transfer System), Europe:

   TARGET2 is the eurozone's RTGS system — the euro equivalent of Fedwire, operated by the Eurosystem (the ECB together with the participating national central banks). When Deutsche Bank in Frankfurt needs to settle a large-value euro payment with BNP Paribas in Paris, or when an Italian bank needs to fund its overnight reserve position, those transactions move across TARGET2. Each payment is individually final and irrevocable the moment it completes. In 2024, TARGET2 settled approximately €2.1 trillion per day across around 1,000 direct participants — making it the world's second-largest wholesale settlement system by value after Fedwire.

   What makes TARGET2 structurally interesting is who owns and operates it. Unlike Fedwire — which is run by a single central bank — TARGET2 was built and maintained by a consortium of four national central banks (the Deutsche Bundesbank, Banca d'Italia, Banque de France, and Banco de España) on behalf of the entire Eurosystem. This cooperative model reflects the political reality of the eurozone: monetary union without full fiscal or political union. The euro is one currency, but the banks that settle it are supervised by 20 different national regulators.

   TARGET2 was migrated in 2023 to a next-generation platform called T2 (sometimes referred to as TARGET2-New Generation). T2 brought ISO-20022 messaging natively into the settlement layer, harmonised liquidity management across national central banks, and introduced a Central Liquidity Management (CLM) module that lets banks pool reserve balances across borders — a meaningful improvement for treasury operations of pan-European banking groups.

   One design feature worth flagging: despite operating on a single shared platform, T2 maintains distinct national component accounts for legal and supervisory purposes. A German bank's reserve account is technically held at the Bundesbank, not the ECB. Settlement between participants is mutual and immediate — but the legal architecture underneath it is layered across 20 jurisdictions. This creates nuances in cross-border eurozone payments that have no direct US equivalent, particularly around legal finality and insolvency treatment.

   For our Singapore-to-Germany payment, TARGET2 is where the euro leg finally lands. After JPMorgan has converted dollars to euros and sent a payment instruction to Deutsche Bank, Deutsche Bank settles the interbank obligation through TARGET2 — adjusting the reserve balances between JPMorgan's euro correspondent account and Deutsche Bank's own account. It's the last pipe in the chain before the supplier sees the funds.

In summary:

Table 2: Wholesale pipe comparison. Notice the design tradeoff: RTGS systems (Fedwire, TARGET2) offer immediate finality but require more liquidity. Netting systems (CHIPS) are more capital-efficient but settle with a delay. CIPS tries to offer both.

Here's the tie-back to our Singapore-to-Germany payment: that $47,000 transfer wasn't one payment. It was a chain of payments, each jumping from one domestic wholesale system to the next. The Singapore dollar leg settled through MAS (the Monetary Authority of Singapore's RTGS system). The US dollar leg settled through Fedwire or CHIPS. The euro leg settled through TARGET2. Three currencies, three wholesale systems, three sets of rules — all chained together by correspondent banking relationships and SWIFT messages.

This is why international payments are slow and expensive. Not because the technology is bad, but because the architecture requires coordination across independently operated systems that were never designed to work together seamlessly. The pipes work fine individually. It's the joints between them that leak.

And at the center of all these pipes sits one currency that everything else flows through. That's next.

Why the Dollar Sits at the Center

Here's a number that explains almost everything about global payments: 89.2%.

That's the share of all foreign exchange transactions that have the US dollar on one side, according to the BIS Triennial Survey of April 2025. Out of $9.6 trillion in daily FX turnover, nearly nine out of every 10 trades involve the dollar.

Remember our Singapore manufacturer's payment? Even though neither party was American, the money flowed through New York, converted into dollars along the way. That's the gravitational pull of the USD at work.

The dollar's dominance isn't an accident, and no single mechanism maintains it. It's a self-reinforcing loop with at least four interlocking components:

1. Liquidity breeds usage: Because so many people trade dollars, the dollar market is the deepest and most liquid in the world. Tight bid-ask spreads mean lower transaction costs. Lower costs attract more participants. More participants deepen the market further.
2. Trade invoicing consolidates around the dollar:

   Even when neither the buyer nor the seller is American, many international trade contracts are denominated in dollars. A Brazilian soybean exporter selling to a Chinese buyer will often invoice in USD — not because either party prefers it, but because both can hedge their dollar exposure cheaply and efficiently. International Monetary Fund (IMF) research through 2023 shows that roughly 40% of global trade invoicing is denominated in dollars, far more than the US share of global trade.

3. Reserves anchor around the dollar:

   Central banks hold reserves to defend their currencies, finance imports in a crisis, and signal stability. As of Q3 2025, 56.9% of the world's allocated foreign exchange reserves were held in US dollars, according to the IMF's Currency Composition of Official Foreign Exchange Reserves (COFER) data. The euro was a distant second at 20.3%. The Chinese renminbi? Just 1.9%.

4. The "petrodollar" as a mechanism bundle: The term "petrodollar" gets thrown around loosely, but the underlying mechanics matter. Since the 1970s, oil producers have predominantly invoiced and settled in dollars, creating a constant global demand for USD. Oil-exporting countries that accumulate large dollar surpluses recycle them into US Treasury securities and dollar-denominated assets — which, in turn, finances US government borrowing and deepens dollar capital markets further. It's not a conspiracy. It's a set of interlocking economic incentives that have persisted for 50 years.

Table 3: The USD's reinforcing roles. Each role reinforces the others, creating a self-sustaining loop that makes the dollar's position extremely difficult (though not impossible) to displace.

The dollar's self-reinforcing loop. Every role feeds the next: liquidity attracts usage, usage drives invoicing, invoicing anchors reserves, reserves deepen capital markets — which deepens liquidity again. Commodity pricing feeds the loop from outside.

The reinforcing loop means that displacing the dollar doesn't require just building a better payment system. It requires simultaneously displacing the dollar in FX markets, trade invoicing, central bank reserves, commodity pricing, and international debt markets. No single currency or technology can do that overnight. But that hasn't stopped people from trying — and recent geopolitics has given them more reason than ever.

From One Big Network to a Multi-Rail World

For decades, the global payment system was, for practical purposes, a dollar-centric network. Almost every cross-border payment eventually touched a US bank, a US clearing system, or a dollar-denominated account. This architecture gave the United States an unusual form of power: the ability to cut individuals, companies, and entire countries off from the global financial system by restricting their access to dollar clearing.

After 2022, the United States exercised that power at unprecedented scale. The disconnection of Russian banks from SWIFT, the freezing of Russian central bank reserves held in Western institutions, and the broad sanctions imposed on Russian financial entities sent a signal that reverberated far beyond Moscow: the dollar-centric system can be weaponized.

The response has been a quiet but significant push to build alternative pipes:

1. Russia's SPFS (System for Transfer of Financial Messages): Built as a SWIFT alternative, SPFS now connects 584 organizations, including 177 non-resident participants from 24 countries. It handles domestic Russian messaging and an increasing share of cross-border communication with partner countries. But SPFS is a messaging system — it doesn't settle anything. Russian banks still need correspondent banking relationships and nostro/vostro balances in the currencies they want to transact in.
2. India's Special Rupee Vostro Accounts (SRVAs): Building local-currency corridors that bypass the dollar entirely. As of early 2025, 123 correspondent banks from 30 countries had opened 156 SRVAs with Indian banks, allowing trade to be invoiced and settled in rupees rather than dollars. In August 2025, the Reserve Bank of India (RBI) simplified the process further, removing the requirement for prior approval to open new SRVAs. The volumes are still modest — total vostro balances reached about $1.6 billion by late 2024 — but the infrastructure is being laid.
3. CIPS: China's Belt and Road Initiative creates natural demand for RMB settlement, and CIPS provides the plumbing.

Meanwhile, the US Treasury's Office of Foreign Assets Control (OFAC) has flagged SPFS participation as a potential sanctions risk — signaling that the act of joining an alternative messaging network could itself trigger scrutiny.

The right frame for this isn't "de-dollarization" — a term that implies the dollar is being replaced. The data doesn't support that. The dollar's share of FX turnover increased from 88.4% to 89.2% between 2022 and 2025. Its reserve share, while declining gradually over decades, is still nearly three times larger than the euro's.

The better frame is multi-rail reality — a regression back toward the Age of Mercantilism, when the world ran on distinct, overlapping payment zones.

The dollar pipe isn't shrinking. New pipes are being added alongside it — for specific corridors, specific trade relationships, and specific geopolitical reasons. A payment from Russia to India might now travel on SPFS + SRVA rails rather than SWIFT + USD correspondent rails. On the other hand, payment from China to a Belt and Road partner might use CIPS directly.

Diagram 7: From one pipe to many. The traditional architecture routes nearly everything through a USD correspondent. The emerging multi-rail architecture adds parallel pipes for specific corridors — but the USD path remains the widest and most liquid.

The world isn't abandoning the dollar plumbing. It's building resiliency, and redundancy. For some countries, that redundancy is a matter of sovereignty — the ability to transact even if cut off from the Western financial system. For others, it's about cost — settling trade in local currencies avoids the double conversion (local currency → USD → local currency) that makes correspondent banking expensive.

Either way, the architecture of global payments is becoming more complex, not simpler. And the newest pipes being proposed are the most radical yet.

Looking Ahead: Stablecoins, Tokens, and New Settlement Ideas

Every generation of payment infrastructure has tried to solve the same problem: how do you move money faster, cheaper, and with less risk? The latest proposals aim to do something more ambitious — collapse the three-layer model itself.

What Exists Today

Stablecoins — digital tokens pegged to a fiat currency, typically the US dollar — have moved from crypto curiosity to institutional reality. Visa launched USDC stablecoin settlement in the United States in December 2025, with Cross River Bank and Lead Bank settling through the Solana blockchain. The program has reached a $3.5 billion annualized run rate globally, spanning pilots across Latin America, Europe, and Asia Pacific.

The appeal is concrete: stablecoin settlement can operate 24/7, including weekends and holidays, when traditional banking systems are closed. For issuers and acquirers, this means faster fund flows and reduced overnight exposure. It's not about replacing banks — it's about extending the hours during which settlement can happen.

But the BIS has sounded a cautious note. Stablecoins, the BIS argues, lack two properties that make bank money safe: singleness (the guarantee that a dollar in one bank is worth exactly a dollar in another) and elasticity (the ability of central banks to create money in a crisis to prevent liquidity crunches). A stablecoin is only as good as the reserves backing it and the governance of the entity issuing it. If the issuer fails or the peg breaks, holders bear the loss — there's no central bank backstop.

The IMF takes a more measured view: stablecoins can improve cross-border payment speed, particularly in corridors where correspondent banking is expensive or unavailable. But the binding constraints remain governance, compliance (AML/KYC for on-chain transactions), and convertibility — the ability to move seamlessly between stablecoin balances and traditional bank deposits.

What's Being Designed

Tokenized settlement goes further. The idea, championed by the BIS in its "next-generation financial market infrastructure" research, is to represent traditional assets — government bonds, commercial bank money, central bank reserves — as programmable tokens on a shared ledger. If you can do that, you can potentially collapse messaging, clearing, and settlement into a single atomic operation: the token moves from one owner to another, and that is the settlement. No separate clearing step, no batch processing, no T+1 delay.

This is genuinely novel. The three-layer model we introduced at the start of this chapter exists because historically, information and money couldn't move at the same time. Messaging had to happen first, then clearing, then settlement. Tokenization proposes to make the information be the money — or at least, be so tightly bound to it that the layers collapse.

It's early. Central bank digital currencies (CBDCs), programmable money, and tokenized securities are all in various stages of experimentation. None has achieved the scale or reliability of Fedwire or CHIPS. The question isn't whether the technology works in a lab — it does — but whether it can replicate the trust, governance, and resilience that the existing plumbing has built up over decades.

The honest answer: we don't know yet. What we do know is that the design questions being asked today — Who governs the ledger? Who provides the backstop? How do you ensure compliance? — are the same questions that have shaped every generation of payment infrastructure since the clearing houses of 1770s London.

In Summary

This chapter traced an arc from a restaurant bill analogy to $9.6 trillion in daily FX trades. Let's step back and see the pattern.

The plumbing of global finance was built to solve three problems.

1. Distance: How do you move money between people who are separated by oceans, borders, and legal jurisdictions? Correspondent banking solved this by creating chains of trust — banks holding accounts with other banks, linked by messaging and netting.
2. Timing and risk: How do you ensure that both sides of a transaction actually deliver? Herstatt showed what happens when they don't, and CLS Bank's Payment versus Payment mechanism was the engineering fix — make both legs settle simultaneously, or not at all.
3. Scale and standardization: How do you handle billions of transactions per day without the system collapsing? SWIFT standardized the messages. CHIPS optimized the netting. Fedwire provided finality. ISO-20022 made the data richer and the processing faster.

Each generation of plumbing was built on top of the last, often in response to a crisis or a failure. The result is a system that works — reliably, at enormous scale — but that's also layered, fragmented, and expensive to operate across borders.

Seen across the whole of Part I, the trust architecture has evolved in layers — each one built on top of the last:

Diagram: The historical evolution of payment trust layers. Each era answers the same question — why should a stranger accept my payment? — with a new institution. The institutions change; the question never does.

The future isn't one pipe replacing all the others. It's multiple overlapping pipes — dollar corridors, renminbi corridors, local-currency corridors, stablecoin rails, tokenized settlement layers — coexisting and competing. The strategic advantage won't belong to whoever builds the fastest pipe. It will belong to whoever controls the interoperability points — the junctions where one rail connects to another.

That's the view from 30,000 feet. In Part II, we're going to zoom all the way in — to a single moment, at a single coffee shop, where a customer holds their phone near a terminal and a beep triggers a cascade of authorization, clearing, and settlement across the very pipes we've just described. The payment takes two seconds. Understanding what happens in those two seconds — and why the same pattern repeats across every electronic payment rail on the planet — is the work of the next few chapters — and it will change how you see every tap, swipe, and checkout from now on. Let's dive in.

Sources

• Bank for International Settlements (BIS), Triennial Central Bank Survey of Foreign Exchange and OTC Derivatives Markets (April 2025) — FX turnover data, USD share of 89.2%, $9.6 trillion daily volume
• International Monetary Fund (IMF), Currency Composition of Official Foreign Exchange Reserves (COFER) (Q3 2025) — USD reserve share of 56.9%, euro at 20.3%, RMB at 1.9%
• CLS Group, Operational Data (H1 2025) — $7.9 trillion average daily settled value, 18 currencies, 70+ settlement members
• Federal Reserve, Fedwire Funds Service Annual Statistics (2024) — 836,000 transactions/day, $4.5 trillion daily value
• The Clearing House, CHIPS Statistics — $1.9 trillion daily volume, 29:1 netting efficiency ratio
• World Bank, Correspondent Banking Data — 20% decline in active correspondent banking relationships between 2011 and 2022
• Bank of Russia, National Payment System Statistics — SPFS: 584 participating organizations, 177 non-resident participants from 24 countries
• Reserve Bank of India, Special Rupee Vostro Account Data (2025) — 156 SRVAs opened by 123 correspondent banks from 30 countries
• Visa Inc., Stablecoin Settlement Program Announcement (December 2025) — USDC settlement via Solana blockchain, $3.5 billion annualized run rate
• BIS, The Future Monetary System and Blueprint for the Future Monetary System — singleness, elasticity, and tokenized settlement concepts
• SWIFT, ISO-20022 Migration Updates — coexistence period ended November 22, 2025; MT103/MT202 retired