perspective

Beating Gridlock

Cross-border payments aren't slow because technology is slow. They're slow because $10 trillion in capital is trapped in prefunded nostro accounts, waiting for settlement that takes days instead of minutes.

By Anton Titov, Founder · Plexo Institute

Published 2026-05-17

The visible fee is only the surface. The dominant cost is capital locked across prefunding, multi hop routes, FX windows, and batch settlement.

Reader Brief

Cross-border payments aren't slow because technology is slow. They're slow because $10 trillion in capital is trapped in prefunded nostro accounts, waiting for settlement that takes days instead of minutes.

What's Inside

Four moves that separate visible payment fees from the invisible capital cost that dominates institutional payment economics.

The argument starts with the hidden cost of trapped working capital, then maps the four structural causes of gridlock, the mechanisms that compress it, the places capital still gets stuck, and the balance-sheet dividend that motivates settlement modernization.

$10T trapped in nostro prefunding plus 12-18% EM cost of capital plus 3-5 day settlement means the invisible cost dominates institutional payment economics.

Fee data shows 6.36% consumer remittance cost in the latest World Bank RPW release and 1.6-4% B2B EM cost. The bigger number is invisible: working capital locked during settlement. At 15% cost of capital, every locked day costs about 4 bps. Multiply that across thousands of corridors and $10T sits idle. This cost does not appear in any fee table, but it dominates institutional treasury economics [1][2][3].

Four structural causes: prefunded nostro accounts, multi-hop correspondent chains, FX conversion windows, and batch settlement cycles.

Each traps a different kind of capital. Prefunding requires every correspondent relationship to pre-load currency. Multi-hop chains fragment capital across intermediaries. FX windows force capital to wait for liquidity hours. Batch settlement turns instant domestic rails into delayed cross-border settlement. Fixing one in isolation does not solve the problem.

CLS Bank benchmark: 96% netting efficiency at $5T daily across 18 currencies, where multilateral compression replaces gross flows with net positions.

Bilateral settlement of 10 operators means 45 potential pairwise transfers. Multilateral netting collapses that into 10 net positions. At the CLS Bank benchmark, every $100 of gross obligations clears with about $4 of actual movement. The catch is scale: efficiency rises with participant count. A network with three operators provides little benefit; a deep network turns gross obligations into capital-efficient net settlement [5].

Capital efficiency dividend: ~$8T redeployable and ~$640B/year recoverable, but the dividend accrues to balance sheets, not payment operators.

If 80% of $10T in prefunding is freed by minutes-level settlement plus multilateral netting, $8T becomes redeployable. At an 8% blended yield, that implies $640B/year in recoverable capital cost. The dividend accrues to the institutions currently holding the trapped capital - large banks, corporate treasuries, and trade finance providers - not only to payment operators. That is why transition can be slow despite obvious economic benefit.

The Cost You Cannot See

The visible cost is the fee. The invisible cost is working capital locked during settlement.

The visible cost of a cross-border payment is the fee: **6.36%** global average for consumer remittances in the World Bank Q3 2025 RPW release and **1.6-4%** for B2B in emerging market corridors [1]. The invisible cost is working capital locked during settlement. At **3-5 days** average settlement time and **12-18%** annual cost of capital in emerging markets, every day of delay has a real price [2]. This cost does not appear in fee data. It dominates institutional payment economics.

$10T Estimated prefunding capital locked in cross-border nostro accounts globally [3]. This analysis uses this as the headline trapped-capital estimate.
12-18% Annual cost of capital in emerging markets. Each locked day extracts roughly 3-5 bps from margin in this analysis.

Three distinct layers of cost. A piece that conflates them is misleading.

1. **Consumer remittance**: 6.36% World Bank global average in Q3 2025, covering fees and FX paid by individual senders. 2. **B2B infrastructure cost**: 1.6-4% in emerging market corridors, with about 20% operator margin, covering ramps, compliance, FX, and correspondent fees. 3. **Cost of capital on locked working capital**: invisible in fee data, but dominant in institutional economics. Gridlock operates at layer 3. A fintech may charge 1% for a payment, but if settlement takes three days and the client cost of capital is 15%, the effective cost is 1% plus the capital drag. At scale across thousands of payments, the capital drag can exceed the fee.

Why Capital Gets Stuck

Gridlock has four structural causes, and each traps a different kind of capital.

Gridlock has four structural causes. Each one traps capital in a different part of the payment chain. Prefunded nostro accounts trap idle balances before the payment begins. Multi-hop correspondent chains trap capital at each intermediary. FX conversion windows trap capital while the market waits for liquidity. Batch settlement traps capital even when domestic rails on both sides are fast. Fixing any one in isolation does not solve the full problem.

Cross-border payment route map showing capital locks at sender treasury prefunding, correspondent chain, FX window, and settlement batch. — Gridlock is not one delay; it is four different capital locks sitting in four different parts of the route.

Prefunding is the dominant form of trapped capital. Every correspondent relationship requires a pre-loaded nostro account.

A nostro account is a correspondent banking account held by one bank with another, pre-funded with the settlement currency of the destination market. To move $1M from the US to Kenya, a US bank must maintain a nostro account at a Kenyan bank with enough balance to cover expected volume. If the bank maintains $10M prefunded and processes $100M per month, the capital turnover is 10x. At 15% cost of capital, that $10M costs **$1.5M/year** just to sit there. Multiply across thousands of corridors and the source arrives at McKinsey's estimate: **~$10 trillion** in global nostro prefunding [3]. Every dollar of it is idle. Every dollar carries its own cost of capital.

Multi-hop chains lock capital at each intermediary until the full chain clears.

When a payment routes through three or four correspondent banks, each intermediary debits its nostro at the previous bank and credits its nostro at the next. Each link requires its own prefunding. Each link adds its own settlement window. The payment is not "in transit" in a single place; it is fragmented across multiple banks' balance sheets, with capital locked at each point. SWIFT gpi reduced this friction for banks that adopted it, cutting average settlement time for enabled corridors to under five minutes [4]. But gpi only covers participating banks. Emerging market corridors with few gpi participants remain stuck in legacy settlement windows.

FX windows force capital to wait for liquidity that only exists during certain hours.

Cross-border FX markets are 24/5, but individual currency pairs have deep liquidity only during overlapping business hours of the two markets. A USD-to-PHP conversion executed outside Asian/US overlap windows pays wider spreads or waits. For institutional flows, this creates settlement risk and capital drag: funds committed to a conversion may sit overnight before execution, unable to be redeployed. This is the problem CLS Bank was designed to solve for major pairs, with 96% netting efficiency at about $5T daily settlement [5]. But CLS covers only 18 currencies. Emerging market currency pairs remain unsolved.

Batch settlement means domestic real-time rails do not translate into cross-border finality.

Most national payment systems settle in near real time within their jurisdiction, including RTP in the US, SEPA Instant in the EU, UPI in India, and PIX in Brazil. But cross-border settlement between them still relies on end-of-day or batch netting cycles. A payment that is "instant" domestically on both ends can still take days cross-border because the settlement layer between the two systems has not been modernized. This is the gap BIS Project Nexus, Project Agora, and Project mBridge are attempting to close [6].

What Breaks The Gridlock

Three technologies and one architectural shift compress capital lock-up, but none solves the problem alone.

Three technologies and one architectural shift compress capital lock-up. Stablecoin transit shortens the settlement window. Multilateral netting compresses gross obligations into net positions. Atomic PvP removes settlement-risk capital. A shared compliance layer reduces repeated counterparty work. None solves the problem alone. Together, they materially change the economics.

Mechanism	What it unlocks	Capital savings
Stablecoin transit	Settlement in minutes replaces a 3-5 day window	3-5 days x cost of capital per transaction
Multilateral netting	Net obligations compress gross flows 20-25x	~96% reduction in actual value movement
Atomic PvP settlement	Both legs clear simultaneously, with no settlement risk	Eliminates Herstatt risk capital
Shared compliance layer	Single KYC per counterparty, not per corridor	Reduces compliance friction at each hop

Stablecoin transit is the fastest-deploying fix. Settlement in minutes replaces settlement in days.

A fiat-sandwich stablecoin payment - fiat in, stablecoin transit, fiat out - completes the transit leg in minutes. The on-ramp and off-ramp legs are gated by banking speed, but the cross-border middle is near instant. For a B2B operator processing $100M per month on three-day settlement, switching to minutes-level transit frees about $10M of working capital (3/30 x $100M). At 15% cost of capital, that is **$1.5M/year** in capital savings. This is the cost-of-capital dividend that stablecoin transit unlocks [7]. The Fiat Sandwich explains the transit architecture.

Multilateral netting compresses gross flows. The CLS benchmark is 96% efficiency at $5T daily volume.

Bilateral settlement means that if 10 operators each owe each other various amounts across corridors, 45 separate pairwise transfers may need to execute. Multilateral netting through a shared clearing layer calculates each operator's net position across all counterparties. Only 10 net transfers execute. At CLS Bank's 96% efficiency, every $100 of gross obligations requires only about $4 of actual value movement [5]. Applied to cross-border stablecoin settlement, a multilateral clearing network with 50 operators across 20 corridors could reduce settlement capital requirements by 20-25x compared with bilateral settlement. The catch is that netting efficiency scales with participant count. A network with three operators in two corridors provides minimal benefit. This cold-start problem makes multilateral networks hard to build but valuable once built.

Atomic PvP eliminates settlement risk capital. No more Herstatt capital buffers.

Payment-versus-payment settlement ensures that both legs of an FX transaction clear simultaneously or not at all. The alternative, unsynchronized settlement, creates Herstatt risk: one party delivers while the counterparty fails. Banks hold capital against Herstatt exposure. Atomic PvP through smart contracts or shared ledgers eliminates this risk and the capital buffer that backs it. BIS Project Agora, involving seven central banks and 41 private institutions in this analysis, is building this primitive for wholesale tokenized settlement [6].

What Still Traps Capital

Even with stablecoin transit and netting, capital remains stuck in last-mile payout, chain liquidity, and trade finance.

Even with stablecoin transit and netting, capital remains trapped in three places. The last-mile off-ramp still depends on local banking speed. Liquidity fragments across chains when each chain needs its own float. Trade finance remains a credit-and-documentation problem with much longer holding periods. Solving these is the next frontier.

The last-mile off-ramp: fiat payout still takes banking time.

Stablecoin transit is minutes. Converting stablecoins back to local fiat and delivering to the recipient's bank account can take hours to days, depending on the destination market's payment rails. In markets with instant payment systems such as PIX, UPI, M-Pesa, and SEPA Instant, off-ramp can be near real time. In markets without them, it reverts to correspondent banking delays. The implication: stablecoin transit unlocks the middle leg, but the last mile remains capital-locked in markets without modern domestic rails.

Liquidity fragmentation across chains: every chain needs its own float.

USDC exists on Ethereum, Solana, Arbitrum, Base, Polygon, and other chains. USDT exists on Tron, Ethereum, BSC, and others. An operator serving multiple corridors must hold float on each chain to avoid bridge delays. This fragments capital across chains instead of consolidating it. CCTP V2, Circle's burn-and-mint protocol, and equivalent cross-chain primitives partially solve this by enabling native transfer between chains without bridge liquidity [8]. But for USDT and other stablecoins without native cross-chain protocols, fragmentation persists.

Trade finance: the deepest pool of trapped capital and the slowest to modernize.

Letters of credit, bank guarantees, and documentary collections involve extended holding periods, often 30-180 days. The ADB estimates a **$2.5 trillion** global trade finance gap, driven largely by banks' reluctance to commit capital against SME trade exposure [9]. Stablecoin payments do not directly address this gap because the gap is about credit, not settlement. But the capital savings from faster settlement in other flows can be redirected toward trade finance. Every dollar freed from nostro prefunding can theoretically support new trade finance. The $2.5 Trillion Gap explores this in detail.

The Capital Efficiency Dividend

When settlement moves from days to minutes, the working capital freed is measured in trillions.

If the global cross-border payments market moves from 3-5 day settlement to minutes-level settlement, the working capital freed is measured in trillions. This is the capital efficiency dividend that motivates central banks, large institutions, and clearing networks to invest in the transition.

Capital efficiency release path showing gross prefunding, minutes transit, net positions, capital release, and an objection rail asking whether the lock moved or shrank. — The capital dividend is conditional: the lock has to shrink through transit, netting, and reconciliation, not simply move to another balance sheet.

The numbers: how much capital is actually unlocked.

Base estimates in this analysis: - Global nostro prefunding: **~$10T** [3] - Average cost of capital across institutions: **~8-12%**, weighted across advanced and emerging markets - If 80% of prefunding is freed by faster settlement and netting: **$8T** redeployable - Annual yield at 8%: **$640B** in recoverable capital cost These are order-of-magnitude estimates, not precise projections. The point is that the capital efficiency dividend is larger than the direct fee savings from faster payments. The second-order effect, redeploying freed capital, exceeds the first-order effect, cheaper fees.

Who captures the dividend: not necessarily the payment operators.

The capital efficiency dividend accrues primarily to the balance sheets that currently hold the trapped capital: large banks with extensive correspondent networks, corporations with global treasury operations, and trade finance providers. Payment operators capture a smaller portion through faster turnover of their own working capital. The larger prize goes to the institutions whose treasury economics improve when settlement accelerates. This has policy implications: the institutions with the most to gain from the transition are also the ones with the most operational friction to switch. This is why the transition is slow despite the clear economic benefit.

Counter-Arguments & Limitations

Every perspective has boundaries. Here are the strongest challenges to this analysis.

Two objections define the boundary of this perspective. The first challenges the headline prefunding number. The second argues that stablecoin transit does not free capital; it simply moves the lock from nostro accounts into stablecoin float.

"The $10T prefunding number is overstated - modern banks already use intraday liquidity facilities and bilateral netting."

The argument: McKinsey's $10T figure aggregates global nostro balances without distinguishing genuinely idle capital from operationally active liquidity. Major banks already manage nostro positions actively through intraday credit lines from correspondents, bilateral netting agreements, and treasury sweeping. The real trapped portion that could be eliminated by multilateral netting may be closer to $3-5T - still meaningful, but a fraction of the headline number. Using the headline number to estimate a $640B annual dividend may overstate the addressable inefficiency. Valid critique of the headline number, but the policy conclusion holds. Even at $3-5T addressable, the recoverable annual capital cost is $240-400B - still the largest capital efficiency opportunity in cross-border payments. The piece does not depend on the precise McKinsey figure; it depends on the directional finding that nostro prefunding is the dominant trapped-capital category and that multilateral netting compresses it materially. Both are corroborated by CLS Bank's 96% efficiency benchmark and BIS multilateral platform research. The order of magnitude of the dividend, not its exact size, is what motivates central bank investment in Project Agora and Project Nexus.

"Stablecoin transit does not free working capital - it just moves the lock from nostro to stablecoin float."

The argument: an operator that runs stablecoin payment infrastructure must hold USDC or USDT float across multiple chains to serve corridors. The float requirement scales with payment volume just like nostro prefunding. The capital is not freed; it is relocated. The headline "$1.5M/year saved per $10M nostro" overstates the benefit because it ignores the offsetting stablecoin float requirement. Partially valid. Stablecoin float is real, but it is materially smaller than nostro prefunding for two reasons. First, stablecoin float can be redeployed within minutes, not days, so the same dollar serves more transactions per unit time. Second, multilateral netting on a clearing layer compounds with this turnover: net positions can be settled with much smaller float than gross volume. CLS demonstrates this for FX with 96% efficiency at $5T daily FX settlement. The full dividend requires both faster transit and multilateral netting, as the piece argues.

Evidence And Sources

This raw HTML export preserves source visibility for crawler and contractor review. Indexing decision: index, follow.