Correspondent banking is the chain of intermediary banks that carries a cross-border payment from the originator's bank to the beneficiary's bank. A typical wire touches four institutions before it lands: the originator bank, a sending correspondent, a receiving correspondent, and the beneficiary bank. Each hop can add fees, FX conversion, and delay. SWIFT is the messaging layer that carries the instructions (MT103 for customer credit transfers, MT202 and MT202 COV for financial-institution transfers, plus the SWIFT gpi service overlay for tracking and same-day value); SWIFT does not settle the money. Settlement happens on correspondent-bank ledgers or on the underlying RTGS or netting rail that sits under the message. The BIS CPMI cross-border payments programme is the canonical source for the taxonomy and the reform roadmap (BIS CPMI cross-border payments).
​

Cross-border payments are slow for structural reasons, not because banks are indifferent to speed. Four hops, batch cutoffs on each RTGS the chain touches, non-overlapping working hours across time zones, and nostro-vostro pre-funding are the load-bearing sources of latency. The FSB roadmap and the G20 targets on cost, speed, access, and transparency call out exactly these frictions (FSB cross-border payments). Stablecoin rails collapse the chain to a single onchain hop with 24/7 finality, which is why the correspondent model is the sharpest comparison for what stablecoin infrastructure actually replaces.
​

A standard cross-border wire routes through four institutions. The originator's bank debits the sender and issues an MT103 customer credit transfer over SWIFT (SWIFT messaging standards). If the originator's bank does not hold an account in the beneficiary's currency at the beneficiary's bank, it hands the payment to a sending correspondent that does. The sending correspondent, which holds a nostro account in the destination currency, moves the value through its own books and onward to a receiving correspondent in the destination country. The receiving correspondent credits the beneficiary bank's vostro account, and the beneficiary bank finally credits the end customer. The chain can be shorter (two hops when a direct bilateral account exists) or longer (five or more hops on exotic corridors), but the four-hop pattern is the working average that BIS CPMI uses when it publishes reform monitoring data (BIS CPMI).
​

Every hop is an opportunity to charge a fee, apply an FX spread, apply an AML or sanctions hold, or miss a cutoff. The lift, drop, and wire-instruction fields (fields 71A, 71F, 71G on the MT103) exist to specify who pays which leg of the fee chain (SWIFT messaging standards). MT202 and MT202 COV cover the bank-to-bank leg of the payment when a cover payment is used to move the underlying settlement value in parallel with the customer instruction (SWIFT messaging standards).
​

The most common category error is treating SWIFT as a payment rail. SWIFT is a messaging network, not a settlement system. It transports standardized instructions between roughly 11,500 institutions in more than 200 countries and territories (SWIFT). The FIN service carries MT messages (MT103, MT202, MT202 COV) and the ISO 20022 MX service carries the newer XML equivalents; the cross-border payments coexistence period between MT and MX ended in November 2025 per the SWIFT ISO 20022 programme (SWIFT ISO 20022 programme). SWIFT gpi is a service overlay on top of the messaging layer that adds end-to-end tracking, unique end-to-end transaction reference (UETR) identifiers, transparency of fees, and a same-day-value commitment for participating banks (SWIFT gpi). None of these products moves money by themselves. When a payment lands in the beneficiary's account, the value moved on Fedwire, TARGET2, CHAPS, or through a chain of correspondent-bank debits and credits per Bank of England payment system overview (Bank of England).
​

Four structural frictions produce the latency. First, every RTGS the chain touches has an operating window and a cutoff for foreign-currency payments. Fedwire Funds runs on business days on a schedule published by the Federal Reserve Banks (FRBservices operating hours), TARGET2 (now T2) runs on TARGET business days per the ECB (ECB TARGET), and CHAPS runs on Bank of England business days per the Bank of England (Bank of England CHAPS). A payment that arrives after cutoff waits for the next open window.
​

Second, non-overlapping working hours across time zones compound the cutoffs. A payment leaving Singapore for New York can hit a receiving-side cutoff before the sending-side even opens, adding a business day per the BIS CPMI operating-hours analysis (BIS CPMI operating hours report). Third, nostro pre-funding traps working capital: the sending correspondent must hold a balance in the destination currency at the receiving correspondent before it can pay out, and the FSB reform roadmap flags nostro pre-funding as one of the largest reservoirs of trapped liquidity in the system (FSB cross-border payments). Fourth, AML, sanctions, and beneficiary-verification screening at each hop can pause a payment for hours or days; the BIS CPMI programme calls this out as a distinct source of end-to-end delay (BIS CPMI).
​

The G20 endorsed the FSB roadmap in 2020 and set explicit end-2027 targets for cross-border payments: for the wholesale segment, 75 percent of payments credited to end beneficiaries within one hour, with the remainder within one business day per the FSB targets note (FSB Targets for Addressing the Four Challenges of Cross-Border Payments). Progress is monitored publicly by the FSB (FSB cross-border payments), and gpi participation already delivers same-day value for a large share of MT103 traffic per SWIFT (SWIFT gpi). The reform work runs alongside, not against, the correspondent model.
​

A stablecoin transfer is a single onchain state transition. There is no sending correspondent, no receiving correspondent, and no nostro-vostro account structure between them. Value moves peer-to-peer between wallet addresses, and settlement is final on inclusion under the finality guarantees of the underlying chain per the Ethereum consensus documentation (ethereum.org consensus docs). Public chains run 24 hours a day, 7 days a week, so the operating-window and time-zone problems disappear at the settlement layer.
​

USDC and USDT are the two largest dollar-denominated stablecoins. USDC is issued by Circle Internet Financial and reserved against short-duration US Treasury bills and cash held with regulated custodians per Circle's monthly reserve reports (Circle transparency). USDT is issued by Tether Holdings and reports its reserve composition on a quarterly attestation basis (Tether transparency). Institutional cross-border volume in stablecoins now runs through payment orchestration providers rather than through correspondent chains: Circle publishes end-to-end payments products and reserve mechanics for institutional USDC use (Circle Payments Network). Costs at the network layer are not zero on any major stablecoin corridor; Tron network fees for USDT transfers are set by the Tron energy and bandwidth model per the Tron developer documentation (Tron resource model), and Ethereum-layer transfers of USDC pay ETH gas priced by the base fee and priority fee mechanics per the EIP-1559 specification (EIP-1559). The correspondent-chain fee stack is what collapses, not the network fee itself.
​

Correspondent banking is not a dead technology. It is the default settlement mechanism for regulated fiat corridors in jurisdictions that have not yet licensed stablecoin issuance or that restrict access to onchain payments. In the United States the GENIUS Act, signed into law on July 18 2025, establishes the first federal framework for payment stablecoin issuers and takes full effect on the earlier of eighteen months after enactment or one hundred twenty days after the primary federal regulators finalize implementing regulations per the House Financial Services Committee summary (House Financial Services Committee). Corridors that do not yet have equivalent frameworks will continue to route wholesale fiat through the correspondent chain, and the FSB has been explicit that stablecoin rails are additive to, not a replacement for, the reform work on correspondent banking (FSB cross-border payments).
​

The other case for correspondent banking is central-bank money settlement for the largest institutional payments. Fedwire Funds settles USD large-value payments in central-bank money on the books of the Federal Reserve per the Federal Reserve Banks operational page (FRBservices Wires), and no stablecoin rail settles in central-bank money today. Treasury operations that need central-bank-money finality use RTGS; treasury operations that need 24/7 access, single-hop settlement, and predictable end-to-end fees increasingly use stablecoin rails. The pattern is convergence at the orchestration layer, not substitution at the settlement layer.
​

For the taxonomy that sits underneath every wholesale payment system, see Net vs Gross Settlement: RTGS and DNS Explained. For the market picture on cross-border stablecoin flows, see how stablecoin cross-border payments compare to correspondent banking and SWIFT's blockchain settlement pilot.
​

For the treasurer's view of the pre-funded balances that make correspondent chains expensive, see Nostro and Vostro Accounts: The Real Cost of Correspondent Banking.

For the three US wire settlement rails (Fedwire RTGS, CHIPS net with intraday finality, and CLS PvP for FX) compared side by side, see US Wire Settlement Rails Compared: CHIPS, Fedwire, and CLS in 2026.