Part IV — Digital Money

Chapter 25 · CBDC and Cross-Border Settlement

Everything domestic that this book has traced — UPI, the e-rupee, the wallets and rails — works inside one currency, one regulator, one law. Cross the border and the machinery reverts to a nineteenth-century design: chains of correspondent banks, prefunded accounts, days of delay, six per cent costs. This chapter prices that legacy, maps the three routes being built against it, and follows the route that became geopolitics: mBridge, the multi-CBDC platform the BIS walked away from and China kept building.

📖 Chapter 25 of 35 ⏱ ~70 min read 🎓 Session 31 · CILO 4 Core 60% Stable 0% Volatile 40% 📝 Quiz · 30 MCQs

🎯 Learning Objectives

1. The Legacy: Correspondent Banking

A domestic payment settles on one ledger inside one jurisdiction. A cross-border payment cannot: your bank in Mumbai holds no account at the recipient's bank in Manila. So the money travels a chain. Your bank holds an account at a correspondent — a large international bank — which holds one at another correspondent, which holds one at the destination bank. Each hop is a balance on somebody's ledger (the nostro/vostro accounts), moved by messages across SWIFT's network, cut by fees and foreign-exchange spreads at every step. SWIFT itself is only a messaging system — the post office, not the money — but the balances the messages describe must exist in advance, which forces banks to park substantial funds as idle liquidity in nostro accounts across the world: capital that earns settlement nothing, sitting there because the architecture cannot promise when a payment will arrive.

The design's costs follow from its shape. Speed: multiple hops, cut-off times, and time zones stack delays into days. Cost: each intermediary prices its leg, and the FX spread hides inside the exchange rate the customer never sees itemised. Opacity: a payment in flight can be invisible to both sender and receiver until it lands. And retreat: as compliance costs rose after the de-risking wave, correspondents abandoned small corridors — entire regions lost banking doors, and remittance costs there rose further. The G20 built a whole roadmap (2020) around these four frictions — speed, cost, transparency, access — because they are structural, not incidental.

📝
Scale of the prize: cross-border payments moved an estimated 190-plus trillion dollars in 2024, and migrant remittances alone exceeded 900 billion — flows that exceed foreign direct investment to low- and middle-income countries. Every percentage point of cost shaved off that flow is development finance that stays with families.
⚠ Volatile content — Reviewed: November 2026 · Next review: February 2027

2. The Cost Evidence

The World Bank's Remittance Prices Worldwide is the scoreboard the G20 targets hang on. The latest full reading (Q3 2025, published 2026):

Measure (sending $200)Average costNote
Global average, all services6.36%SDG target: 3% by 2030; no corridor above 5%
Banks14.99%The most expensive channel by far
Money transfer operators~4.7%Traditional plus fintech players
Digital remittances4.59%Against 7.30% for non-digital
Cheapest qualifying services (SmaRT)3.29%The cheapest three per corridor — near the target
Sub-Saharan Africa (receiving)8.46%The most expensive receiving region
South Asia (receiving)5.30%Cheap — competitive Indian corridors

Read the table twice. The first reading is failure: two decades of G8, G20 and UN targets have not moved the global average below six per cent — still more than double the 3 per cent SDG goal, with a BIS speech in 2026 conceding that none of the four G20 targets has been achieved for remittances. The second reading is the tell: the cheap services already exist. Digital channels at 4.59 per cent and the SmaRT average at 3.29 per cent prove the technology can land near the target — the expensive averages are the distribution of access, not the physics of payment. The question the rest of this chapter pursues: which architecture makes the cheap service the normal one?

3. Three Routes to a Fix

Everything being built against the correspondent chain reduces to three architectures — a taxonomy to carry through the chapter.

  1. Bilateral fast-payment links. Two countries connect their domestic instant payment systems directly, corridor by corridor: India's UPI–PayNow link with Singapore (live February 2024) is the model, and UPI has similar arrangements in the neighbourhood. Cheap, quick to build, and exactly as scalable as bilateral diplomacy: every new country needs a new bespoke link — the N×(N−1)/2 problem, a spaghetti of one-off connections.
  2. The multilateral hub. Every instant payment system connects once to a single platform and reaches all the others: Project Nexus, the BIS-designed scheme now being built by six Asian central banks including the RBI. Solves the spaghetti problem by design; raises the governance question of who runs the hub. Settles in bank money — the same deposits, now moved across borders instantly.
  3. The multi-CBDC platform. Participating central banks put wholesale CBDC on one shared ledger, so cross-border settlement happens in central bank money, peer-to-peer, with no correspondent chain at all: mBridge and its Western-leaning counterpart Agorá. The most radical surgery on the legacy — and, as Section 4 shows, the one that collides with geopolitics.

Note what distinguishes routes two and three, because commentary blurs it: Nexus moves bank deposits cross-border (the money is a commercial bank's liability throughout); mBridge settles in central bank money (the liability is the sovereign's, final on a shared ledger). Chapter 22's taxonomy, applied at the border. India is deliberately building positions on all three routes at once — Section 6.

⚠ Volatile content — Reviewed: November 2026 · Next review: February 2027

4. mBridge: Platform and Politics

mBridge began in 2021 at the BIS Innovation Hub's Hong Kong centre with four partners: the People's Bank of China's Digital Currency Institute, the Hong Kong Monetary Authority, the Bank of Thailand and the Central Bank of the UAE. A 2022 pilot ran real-value transactions; by June 2024 the platform reached minimum viable product — a shared ledger on which each central bank operates a validating node and commercial banks settle directly in wholesale digital currencies, cross-border, around the clock. The measured promise: settlement in seconds rather than days, at a fraction of correspondent cost. Saudi Arabia's central bank joined as a full participant in June 2024. More than thirty authorities observed — including the ECB, the New York Fed's innovation centre, and the RBI.

Then the politics arrived. In October 2024 the BIS exited, handing the platform to the partner central banks. The BIS called it graduation — four years of support, partners ready to carry on alone — and its general manager rejected the suggestion that mBridge could serve as a BRICS sanctions-bypass rail. Skeptics noted the timing and the pressure reported from Washington. In September 2026 the Saudi exit became public: SAMA said it had completed its proof of concept on 13 May 2025 and was no longer a participating member — "as planned," it insisted, while commentators read US sensitivity to non-dollar settlement. The platform itself kept moving: Macao joined as a full member and went live on 2 June 2026 with eleven approved banks; Mongolia joined with a steering-committee seat; Bank of China moved a record 11.3 billion yuan single transfer; Dubai's government executed its first digital-dirham transaction. Cumulative volume passed roughly 55–70 billion dollars across a few thousand transactions — real money, but a rounding error against dollar clearing, and an estimated 95 per cent of it settled in digital yuan.

Two lessons, one technical and one strategic. Technically, mBridge works: the multi-CBDC design does collapse the correspondent chain into one shared ledger. Strategically, the project exposed what might be called the neutrality problem: a settlement rail's neutrality lasts exactly as long as every member's willingness to keep it neutral. A platform that could settle outside dollar clearing invites the question of who else might use it that way — and the BIS, accountable to central banks on every side of every divide, could not underwrite that answer as membership drifted east. Its response was to build the alternative: Project Agorá, launched the same year it left mBridge, with seven G7-aligned central banks and forty-plus private institutions — JPMorgan, Citi, HSBC, SWIFT itself — testing tokenised bank deposits alongside wholesale CBDC on a unified ledger, with findings due through 2026. Same technology, different coalition. Settlement infrastructure, it turns out, is foreign policy by other means.

⚠
Warning: Do not accept either headline — "mBridge proves de-dollarisation" or "mBridge failed". The honest reading is narrower: the platform is technically sound, commercially marginal (a few thousand transactions, 95 per cent one currency, no Western member), and politically radioactive enough to cost it its two most symbolically important participants. Volume, not volume announcements, is the metric to watch at each review.
⚠ Volatile content — Reviewed: November 2026 · Next review: February 2027

5. Nexus: The Hub Being Built

While mBridge made headlines, Nexus made a company. Conceived at the BIS Innovation Hub's Singapore centre in 2022, the project delivered its full blueprint — scheme rulebook, governance, commercial model, ISO 20022 message specifications — in July 2024, when the BIS and five first-mover central banks (Malaysia, the Philippines, Singapore, Thailand, and India, with UPI the world's largest instant system) signed the founding agreement in Basel. In April 2025 those five incorporated Nexus Global Payments — a not-for-profit company in Singapore, owned by the central banks, to run the scheme; the BIS handed over the keys and stepped back to advise. In February 2026 the build began in earnest: a PayNet–NETS joint venture (Malaysia and Singapore's domestic operators) was appointed technical operator, with AWS on cloud and Endava on engineering, and Indonesia joined as the sixth member. Go-live is targeted for 2027, connecting fast-payment systems that together reach roughly 1.7 billion people — with payments designed to arrive within sixty seconds and the final amount shown before sending. The ECB is watching closely: the Eurosystem is weighing a euro-area connection through TIPS alongside a possible bilateral UPI link.

Two honest caveats. First, a target is not a launch: the RBI's 2024 announcement said "live by 2026"; it is 2027 now, and no first live Nexus payment had been confirmed as of late October 2026. Second, the bilateral bridge already exists and is growing while the hub is built: UPI–PayNow recorded close to S$21 million in the first half of 2026 — six times the same period of 2025 — real money, but a rounding error against the corridor's total. Nexus's bet is that the hub's economics (connect once, reach every member) beat the spaghetti's economics (connect again for every country) at scale. It is the right bet on paper; the 2027 go-live and the tariffs that follow will price it in practice.

6. India's Three-Track Strategy

Stand back and India's position across this landscape is unusually coherent — a portfolio, not a bet. Track one, bank money, bilateral: UPI links country by country — live with Singapore, extended across the neighbourhood — moving deposits instantly where corridors justify the build. Track two, bank money, multilateral: founding membership of Nexus, contributing UPI's scale and the RBI's operating experience to the hub that generalises track one. Track three, central bank money: the e-rupee's cross-border agenda from Chapter 24 — MoUs with Singapore and the UAE, bilateral and multilateral pilot commitments for FY27, and observer status at mBridge to watch the multi-CBDC frontier without boarding it. India joined the Western-leaning Agorá? It did not — the RBI has kept its central-bank-money diplomacy Asian and bilateral, which is itself a signal.

The strategy's logic is sequencing. Bank-money routes deliver benefits to remitters and merchants now, on rails that already exist and scale through diplomacy. The CBDC route is an option on the future — kept warm through pilots and observation, activated if and when multi-CBDC settlement proves necessary rather than merely possible. A country with UPI's scale can afford patience; a country whose domestic instant system is the world's largest also has the most to gain from the hub model, because every new Nexus member is a new destination for UPI. The wager is that connectivity, not currency, is the near-term contest — and that the currency contest, if it comes, rewards those holding options on every side.

7. Comparing the Routes

The chapter's summary table — and the method the hands-on will run on a real corridor.

RouteMoney settledScalingStatusGovernance
Correspondent / SWIFTBank money, hop by hopAlready globalIncumbent; improving (gpi) but structurally unchangedCommercial banks + SWIFT standards
Bilateral IPS link (UPI–PayNow)Bank money, one hopLinear — each corridor a new buildLive 2024; small but growing volumeTwo central banks per link
Nexus hubBank money, instant, multilateralConnect once, reach all membersBuilding; go-live targeted 2027NGP, central-bank owned, not-for-profit
Multi-CBDC (mBridge)Central bank money on shared ledgerRequires aligned membershipLive but marginal; geopolitics volatilePartner central banks post-BIS; neutrality contested
Multi-CBDC (Agorá)Tokenised deposits + wholesale CBDCCoalition-basedTesting; findings through 2026BIS + G7-aligned coalition + private firms

What would settle the contest among routes? Not technology — every architecture above works. The deciders are governance (who sets the rules and who trusts them), compliance (sanctions and AML regimes must be able to live with the rail), liquidity (settlement requires funded balances in the currencies members actually trade), and tariffs (what providers ultimately charge users). Watch those four, not the announcements. Part IV continues next with the asset class that shares CBDC's rails but not its issuer — stablecoins, and the regulatory fight over them, in Chapter 26.

🛠 Hands-On: Settlement Route Comparison

Three corridors, four routes. For each scenario, compare the routes and recommend one — with the arithmetic and the trade-off stated.

  1. Scenario A — The remitter: a worker in Singapore sends ₹20,000 (about S$310) home to Tamil Nadu every month.
  2. Scenario B — The importer: a Mumbai importer must settle a $2 million invoice from a Dubai supplier within 48 hours.
  3. Scenario C — The policy desk: India's payments diplomacy must choose where to place the next three years of effort — a new bilateral UPI link, deepening Nexus, or joining a multi-CBDC platform.

Routes available: (1) correspondent banking via SWIFT; (2) bilateral UPI–PayNow; (3) Nexus (assume post-2027 live); (4) a mBridge-style multi-CBDC platform. Cost data from Section 2 applies.

Steps

  1. For Scenario A, compute the annual cost of sending under the global average (6.36%), the digital average (4.59%) and the SmaRT average (3.29%). How many months of remittances does the cheapest route save per year, versus the average?
  2. For Scenario B, rank the four routes on speed, finality, and counterparty risk. Which single route do you recommend today — and which is blocked by what?
  3. For Scenario C, build the decision matrix: effort, benefit, time-to-live, and geopolitical exposure for each of the three options. Recommend a sequence.
  4. The trap: a colleague argues "multi-CBDC platforms make sanctions unenforceable — so India should refuse all of them." Steelman the opposite in three sentences, then give your own position.
View Solution / Walkthrough

1. Remitter arithmetic. Annual flow: ₹2,40,000 (≈ $2,860·year at S$310/month ≈ $3,720 — use the rupee figure consistently; the percentages apply either way). At 6.36%: ₹15,264 a year. At 4.59%: ₹11,016. At 3.29%: ₹7,896. The cheapest route saves ₹7,368 a year against the global average — which is roughly 1.5 monthly remittances recovered every year for the family. The SmaRT–digital gap (₹3,120) is what route choice alone can capture today with no technology change at all — the distribution point of Section 2.

2. Importer ranking. Speed: multi-CBDC (seconds) and Nexus (under a minute) > bilateral link (instant in corridor) > correspondent (hours to days). Finality: multi-CBDC settles in central bank money with final settlement on-ledger — strongest; correspondent is final only per-hop, with nostro counterparty exposure in between. Recommendation today: the bilateral corridor rails where they exist (India–UAE arrangements) over correspondent banking; the multi-CBDC option is blocked not by technology but by India's non-membership of mBridge — the platform exists in that corridor and the RBI is only an observer. Nexus is blocked by the calendar (go-live 2027, retail focus).

3. Decision matrix. Bilateral link: low effort, corridor-specific benefit, live in months, negligible geopolitical exposure — best marginal return per corridor with real flow. Nexus: medium effort, network benefit compounding with each member, live 2027, low exposure — the strategic centre of gravity for a country with UPI's scale. Multi-CBDC: high effort, uncertain benefit (marginal volumes, 95% one-currency concentration), high exposure — wrong to join mBridge now, right to keep observing and to continue bilateral e-rupee MoUs. Sequence: deepen Nexus first, keep adding bilaterals opportunistically, hold the multi-CBDC option warm. That is, in fact, India's revealed strategy — the exercise should rediscover it from first principles.

4. The sanctions steelman. Against the colleague: settlement platforms do not create sanctions evasion — cash, gold and existing correspondent chains already move what states want moved; a shared ledger with central-bank validating nodes is more auditable than a nostro chain, because every participating central bank sees the records it validates; and refusing to engage cedes the standard-setting to others — India's influence on AML rules for such platforms is zero from outside. Own position (defensible either way, argued not asserted): engage selectively, insist on compliance rules as a condition of membership, and remember that the neutrality problem is real — mBridge's own history shows even observers must ask who else the rail serves.

Key Takeaways

1

Correspondent banking is slow and expensive by architecture, not accident: chains of hops, prefunded nostro liquidity, unitemised FX spreads.

2

The global average cost of sending $200 is 6.36 per cent against a 3 per cent target — but digital (4.59%) and cheapest-service (3.29%) averages prove the target is reachable. Access, not physics, is the barrier.

3

Three architectures: bilateral links (spaghetti), the Nexus hub (connect once, bank money), multi-CBDC platforms (central bank money on a shared ledger).

4

mBridge works technically, remains commercially marginal (95 per cent digital yuan, no Western member), and cost itself the BIS and Saudi Arabia — the neutrality problem in action. Agorá is the same technology with a different coalition.

5

India runs all three tracks at once: UPI bilaterals now, Nexus as the strategic hub, e-rupee diplomacy as the option on central-bank-money settlement.

Test Your Understanding

Choose one best answer per question. Section 2's cost table and the route-comparison table are fair data for any item. Aim for 24/30 before moving to Chapter 26.

1. SWIFT, correctly described:

2. Why must correspondent banks hold nostro balances in advance?

3. The G20 cross-border payments roadmap targets four frictions:

4. The global average cost of sending $200, against its target:

5. The most expensive remittance channel, on average:

6. The most expensive receiving region, at 8.46 per cent:

7. What the 3.29 per cent SmaRT average proves:

8. The scaling weakness of bilateral fast-payment links:

9. The money distinction between Nexus and mBridge:

10. mBridge's founding partners did NOT include:

11. mBridge reached minimum viable product in:

12. The BIS left mBridge in October 2024, describing the exit as:

13. Saudi Arabia's exit, accurately stated:

14. Post-BIS expansion of mBridge has included:

15. The currency-composition fact about mBridge volume:

16. The neutrality problem, stated:

17. Agorá's coalition, in one line:

18. Nexus Global Payments is:

19. India's position in Nexus:

20. Nexus's go-live status as of late 2026:

21. Nexus membership after the founders:

22. The UPI–PayNow bilateral link:

23. India's three-track strategy consists of:

24. India's mBridge status:

25. What will settle the contest among routes?

26. The development stakes of remittance costs:

27. The de-risking wave's effect on correspondent banking:

28. The hub model's core promise:

29. The honest three-part reading of mBridge:

30. Chapter 26 takes up: