Cross-border finance does not suffer only from slow payments. It suffers from unnecessary movement.
Every day, institutions move gross obligations across correspondent accounts, prefunded corridors, stablecoin rails, and settlement networks even when much of that value could be mathematically offset before settlement. The result is trapped liquidity, larger nostro balances, higher intraday funding requirements, and lower return on capital.
Multilateral netting is the mathematical compression of gross financial obligations between three or more participants into a minimum set of net settlement instructions. Where bilateral netting reduces obligations between two parties, multilateral netting finds offsets across an entire group — an institution that owes $50M to one counterparty and is owed $48M from another can settle the $2M residual instead of moving $98M.
The mechanics aren't new. CLS Bank has used multilateral netting to settle 18 currencies and over $8 trillion daily for more than two decades, reducing participants' funding requirements by more than 96 per cent. CHIPS — the largest US dollar clearing system — has run multiple daily netting cycles since the 1970s. The Bank for International Settlements has published research advocating multilateral netting platforms as a structural fix for cross-border payment friction.
Yet most of the global financial system still lacks direct, configurable, open-access multilateral netting infrastructure. CLS reaches around 70 settlement members directly and tens of thousands of third-party participants indirectly — but its access pathway is structured around major-currency FX between large institutions. For participants outside the major-currency CLS perimeter, for asset classes other than spot FX, and for the long tail of regional, Tier-2, and emerging-market institutions, there is no equivalent platform. Those participants prefund every transaction. They tie up capital that compounds at the bank's hurdle rate. And they price that inefficiency into every wire, every remittance, and every stablecoin movement that touches the institutional rail.
This article explains why multilateral netting works, what it requires, and why large parts of the global settlement market still rely on gross settlement even where netting would produce superior capital efficiency.
Bilateral vs. multilateral netting
Bilateral netting reduces obligations between two parties by offsetting their reciprocal flows. If Bank A owes Bank B $100M and Bank B owes Bank A $80M, they can settle the $20M residual rather than moving $180M between them. Most institutional FX markets do this informally; CLSNet and SWIFT GPI's bilateral aggregation services are formalised versions.
The mathematical limit on bilateral netting is the symmetry of the relationship. Two parties can only compress what they owe each other directly. If a third party is involved — say, Bank A is owed money by Bank C, and Bank C owes money to Bank B — the bilateral approach cannot capture that cycle. Bank A and Bank B settle directly; Bank C settles with each. Three settlements occur where one might suffice.
Multilateral netting captures these cycles. The algorithm computes each participant's net position across all counterparties simultaneously: the sum of inbound obligations minus the sum of outbound obligations. The result is a set of net settlement instructions that minimises the number and value of movements, irrespective of which counterparty owes which.
Empirically, the compression jump from bilateral to multilateral is substantial. Bilateral netting typically achieves 30 to 50 per cent compression on a population of obligations. Multilateral netting on the same population routinely exceeds 80 per cent, with well-designed systems reaching 90 to 95 per cent. This is why CLS, CHIPS, and most large-scale clearing infrastructures use multilateral netting as the primary mechanism rather than relying on bilateral aggregation alone.
The mathematical necessity of windows
Multilateral netting requires a window. Obligations must accumulate so the algorithm has a population to work with — zero accumulation means zero offsets to find.
This is the central tradeoff in settlement system design: longer windows produce higher compression but introduce settlement latency. Shorter windows reduce latency but reduce compression. At the limit, a zero-length window — atomic or T+0 settlement — produces zero netting, because every obligation is its own population of one.
The peer-reviewed literature on netting window optimisation is consistent. Work by McLaughlin and colleagues, and research published by the BIS Committee on Payments and Market Infrastructures (CPMI), has examined the efficiency frontier across multiple market structures. The general finding: capital efficiency rises sharply as windows grow from zero to 5 to 10 minutes, continues rising through 30 to 60 minute windows, and asymptotes around end-of-day batches. The marginal return of moving from a 30-minute window to a 4-hour window is small; the cost of moving from a 30-minute window to atomic settlement is substantial.
Production systems reflect this. CLS settles in roughly 5-hour PvP (payment-versus-payment) windows that align with overlapping FX market hours. CHIPS runs multiple daily netting cycles, with end-of-day finality. The European Central Bank's TARGET2 maintains both gross (RTGS) and netted modes for different participant needs.
The lesson: windows are not a flaw to be eliminated — they are the design that makes netting work.
What multilateral netting requires
Three components are required for multilateral netting to operate at institutional scale:
1. A trusted operator. Some entity must compute the net positions and dispatch settlement instructions. In CLS, this is a SIFI-designated Special Purpose Bank with Federal Reserve oversight. In CHIPS, it is The Clearing House Payments Company. In multi-CBDC platforms like the BIS Innovation Hub's Project Agora and Project mBridge, central banks act as joint operators.
2. Settlement finality. Once the netting computation is complete and net positions are dispatched, those positions must be legally and operationally final. CPMI-IOSCO Principles for Financial Market Infrastructures (PFMI) Principle 8 — Settlement Finality — specifies what this means in regulated context: the settlement instruction is irreversible at a defined moment, supported by legal opinions that bind in each participating jurisdiction.
3. Standardised message formats. Net settlement instructions must be expressible in a format every participant's downstream systems understand. ISO 20022 — specifically the CBPR+ (Cross-Border Payments and Reporting Plus) usage guidelines that became the institutional baseline for cross-border payments and reporting following the November 2025 end of the SWIFT MT and ISO 20022 coexistence period — is the institutional standard for these messages. The pacs.008 (Financial Institution to Financial Institution Customer Credit Transfer) and pacs.009 (Financial Institution Credit Transfer) message types are the two most relevant for net settlement dispatch.
The first requirement — trusted operator — is the structural reason multilateral netting has not scaled to all institutions. CLS membership requires a substantial equity buy-in, an investment-grade credit rating, central bank sponsorship, and a per-jurisdiction legal opinion. These are reasonable controls for a SIFI-designated infrastructure, but they exclude tens of thousands of institutions that would benefit from the underlying mechanism.
Capital efficiency: the institutional case
For a financial institution, settlement infrastructure is fundamentally a capital decision. Every gross prefunding requirement is capital that cannot be deployed elsewhere. Basel III liquidity supervision, including the BCBS 248 monitoring tools for intraday liquidity management, makes those exposures visible to supervisors and to bank treasury teams — and the lower the intraday exposure, the lower the liquidity that must be held idle against settlement obligations.
Consider a payment service provider running a single cross-border corridor. The corridor processes $100M per month gross, distributed across business days. To settle gross, the PSP must maintain prefunded nostro accounts in the destination jurisdiction equivalent to several days of expected flow — typically $10 to 20 million of locked capital.
At an institutional cost of capital of 8 to 12 per cent, that locked capital costs the PSP between $800,000 and $2.4 million per year for a single corridor. Multiplied across the 30 or more corridors a global PSP might operate, the annual cost of prefunding can exceed $30 million — purely as a function of settling gross rather than netted.
Multilateral netting compresses this. A 90 per cent compression rate reduces effective settlement value to 10 per cent of gross. The required nostro buffer scales accordingly. Capital that was locked is freed for productive deployment, lending, or returned to shareholders. Intraday liquidity exposures fall, improving the bank's profile under supervisory monitoring frameworks. The same transaction revenue produces a higher return on equity.
This is why the institutional case for netting infrastructure is not "faster settlement" — it is capital efficiency at scale.
Why most cross-border flows still settle gross
If multilateral netting is so capital-efficient, why does it cover only a fraction of global cross-border settlement volume? Three structural reasons:
Currency coverage gaps. CLS supports 18 currencies. The remaining roughly 150 actively traded currencies — including the Chinese renminbi, the fifth most-traded currency globally — have no equivalent multilateral netting infrastructure. Trades involving these currencies settle gross or rely on bilateral relationships with limited compression.
Institutional access constraints. CLS reaches around 70 direct settlement members and tens of thousands of third-party participants. But access is mediated through CLS member banks, structured around eligible FX flows in supported currencies, and gated by membership economics that exclude most regional and Tier-2 institutions from direct, configurable participation. The result is concentration: a small set of major banks have full access; everyone else has partial, mediated, or no access at all.
Asset class fragmentation. CLS handles foreign exchange. Some specific markets do have specialist clearing — FICC's General Collateral Finance Repo Service nets a portion of US Treasury repo, for example — but there is no universal cross-border, cross-participant netting layer for the global repo market or its growing tokenised equivalents. Tokenised asset markets (BlackRock's BUIDL, Ondo's OUSG, Franklin Templeton's BENJI) settle bilaterally. Stablecoin platforms — including Circle's Cross-Border Payment Network, Ripple's RLUSD network, and Visa's USDC settlement layer — generally settle obligations gross rather than through multilateral compression. Trade finance settles document-by-document. Energy settlement runs on bilateral nominations.
The pattern is consistent: multilateral netting works extraordinarily well where it has been deployed (FX between major banks), and is structurally absent everywhere else.
What modern netting infrastructure must address
The next generation of multilateral netting infrastructure has to address the structural gaps of the previous generation. Four architectural requirements stand out:
Multi-currency from day one. CLS's currency expansion has been slow because adding a currency requires central bank sponsorship, legal opinions, and a shareholder vote. A protocol-based approach can support currencies modularly without requiring per-currency governance approval, opening corridors for the 150-plus currencies CLS does not cover.
Open access without sponsorship gating. A multi-million-dollar equity buy-in is rational for SIFI risk concentration, but excludes most of the institutional market. Settlement infrastructure designed for the long tail of regional, Tier-2, and emerging-market institutions must accept participants based on technical and compliance criteria — KYB tiers, sanctions screening, settlement track record — rather than membership equity.
Configurable settlement timing. A single window length cannot serve all use cases. A trade finance corridor might settle daily; a high-frequency FX corridor might settle every 10 minutes; an urgent treasury obligation might require fast-lane bypass. Configurability — per corridor, per obligation priority, per asset class — is the architectural answer to the false debate between netting and atomicity.
Cross-asset interoperability. Fragmented netting infrastructures (FX in one system, repo in another, tokenised assets in a third, stablecoins in a fourth) leave compression value on the table. Cross-asset and cross-corridor netting captures offsets that single-asset systems cannot — particularly in environments where the same institution is active across multiple asset classes simultaneously.
The opportunity is not to rebuild CLS for the currencies and institutions it already serves well. CLS solved the problem it was designed for — multilateral netting of major-currency FX between major banks — and it solved it durably. The opportunity is to extend the capital-efficiency primitive of multilateral netting into corridors, participants, and asset classes where no equivalent infrastructure exists.
For emerging-market corridors, payment service provider networks, stablecoin-assisted settlement, tokenised deposits, and tokenised asset markets, the missing layer is not another payment rail. It is an obligation-compression layer that determines how much value actually needs to move before any settlement rail is used. That distinction matters. Faster gross settlement still requires gross liquidity. Net settlement reduces the liquidity requirement itself.
What institutions should demand
For treasurers, risk officers, and central bankers evaluating settlement infrastructure, four questions matter most:
- What is the demonstrated multilateral compression rate, on what corridors, at what gross volume? Vendor claims should be measurable. Testnet data and pilot results should be reviewable. Compression of 90 per cent or more across multiple corridors and a meaningful gross volume sample is the institutional benchmark.
- What is the settlement finality framework? Legal opinions, jurisdictional coverage, alignment with CPMI-IOSCO PFMI Principle 8. Settlement that is technically final but not legally final is not settlement.
- What is the regulatory engagement posture? Active dialogue with central banks, sandbox participation, supervisory observability. Settlement infrastructure operating outside regulator awareness is a structural risk, not an architectural feature.
- What is the configurability model? Window length, urgency lanes, asset coverage, participant onboarding terms. Infrastructure that cannot be configured per corridor and per obligation type cannot serve the diversity of institutional flow.
Multilateral netting is not new infrastructure. It is established, peer-reviewed, regulator-respected technology that has compressed the largest FX settlement system in the world for two decades. The question is no longer whether it works — CLS proved that. The question is whether the next generation of netting infrastructure will reach the institutions and asset classes the previous generation could not.
FiatRails is building for that gap — a configurable settlement-compression layer for institutions, corridors, and asset classes underserved by existing netting infrastructure. The infrastructure exists; the question for institutional buyers is which corridors and asset classes will get access to it, and on what terms.
This article is part of the FiatRails Insights series on institutional settlement infrastructure.