Money Movement Infrastructure: ACH, SWIFT, SEPA, and How Money Actually Moves

Every time you send a payment, multiple institutions coordinate behind the scenes to move your money. A payroll deposit touches your employer's bank, an ACH operator, the Federal Reserve or The Clearing House, and finally your bank. An international wire can pass through four or five correspondent banks before reaching the recipient. Understanding this infrastructure is essential for anyone building in payments, because the rails you choose determine your speed, cost, and failure modes.

This article breaks down how ACH, SWIFT, and SEPA actually work: the batch windows, the message types, the clearing and settlement layers, and why trillions of dollars sit idle in accounts around the world just to keep the system running.

Messaging vs Settlement: The Core Distinction

The most common misconception about money movement is that sending a payment message means money has moved. It has not. Traditional payment systems separate two functions: messaging (the instruction to move money) and settlement (the actual transfer of value between institutions).

SWIFT is a messaging network. When your bank sends a SWIFT MT103 message, it is telling another bank “please credit this account.” The actual movement of funds happens later, through settlement systems like Fedwire (for US dollars) or TARGET2 (for euros). This separation creates the delays, costs, and reconciliation burdens that define traditional payments.

Key distinction: Messaging says “move the money.” Settlement actually moves it. SWIFT handles the first part. Fedwire, TARGET2, and CHIPS handle the second. The gap between them is where most payment friction lives.

How ACH Works: Batch Processing at Scale

The Automated Clearing House is the backbone of routine US payments: payroll, bill pay, government benefits, person-to-person transfers, and business-to-business transactions. In 2024, the ACH network processed 33.6 billion payments worth $86.2 trillion. By 2025, that grew to 35.2 billion payments worth $93 trillion, averaging 141 million transactions per day.

The Batch Processing Model

ACH is fundamentally a batch system. Payments are not processed individually: they are collected, grouped, and transmitted in files at scheduled intervals. Two operators run the network: FedACH (operated by the Federal Reserve) and EPN (operated by The Clearing House). FedACH runs three same-day processing windows and six future-dated windows per business day.

The lifecycle of an ACH payment follows a specific chain. The originator (the entity initiating the payment) submits a transaction to their Originating Depository Financial Institution (ODFI). The ODFI batches this with other transactions and sends the file to an ACH operator. The operator sorts the transactions by destination and delivers them to each Receiving Depository Financial Institution (RDFI). The RDFI then credits or debits the receiver's account.

Standard ACH takes one to three business days end-to-end. The network only operates on business days: a payment initiated on Friday evening waits until Monday for processing. Federal holidays add additional delays.

Same-Day ACH

Same-Day ACH launched in 2016 with a per-transaction limit of $25,000. NACHA has progressively raised this ceiling: to $100,000 in 2020, $1,000,000 in March 2022, and $10,000,000 scheduled for September 2027. The $1 million increase produced an immediate 30% rise in volume and 118% increase in value within two months.

In 2024, Same-Day ACH processed 1.2 billion payments (up 45.3% year-over-year) valued at $3.2 trillion. By 2025, volume reached 1.4 billion payments worth $3.9 trillion. Even so, Same-Day ACH is not instant: it operates within specific submission windows (the latest at 4:45 PM ET) and still requires batch processing within those windows.

ACH Return Codes and Failure Modes

ACH transactions can fail and be returned, each with a specific code identifying the reason. These returns are a significant operational concern for payment processors, fintechs, and any business running direct debits.

R01 (Insufficient Funds) accounts for roughly 40-50% of all ACH returns. The distinction between a 2-day and 60-day return window matters: R07 and R10 (unauthorized transaction claims) can arrive weeks after settlement, creating chargeback-like risk for businesses pulling funds via ACH debit.

How SWIFT Works: The Global Messaging Layer

SWIFT (Society for Worldwide Interbank Financial Telecommunication) is not a payment system. It is a secure messaging network connecting over 11,500 financial institutions across 200+ countries. In 2024, SWIFT transmitted an average of 53.3 million messages per day, up from 47.6 million in 2023. Roughly every three days, the equivalent of the world's entire GDP passes through the network.

Message Types and ISO 20022

SWIFT messages follow a structured format identified by “MT” codes. The most important for payments are MT103 (single customer credit transfer, the standard international wire transfer), MT202 (bank-to-bank transfers for settling interbank obligations), and MT760 (standby letters of credit and bank guarantees).

The network completed its migration to ISO 20022 messaging in November 2025. The legacy MT format for cross-border financial institution payments is now subject to automatic conversion to MX (ISO 20022) format with additional processing charges. This migration enables richer, more structured payment data: where an MT103 could carry limited remittance information, the equivalent pacs.008 message supports detailed invoice references, regulatory reporting fields, and structured address data.

Correspondent Banking: The Chain of Intermediaries

SWIFT messages instruct payments. But the actual settlement of cross-border payments relies on correspondent banking: a network of bilateral relationships where banks hold accounts at each other. These are called nostro and vostro accounts.

When Bank A in the United States needs to pay Bank B in Japan, and they have no direct relationship, the payment flows through intermediaries. Bank A sends funds to its nostro account at a correspondent bank that has a relationship with a Japanese bank, which in turn credits Bank B's account. Each intermediary deducts fees and adds processing time.

A single cross-border payment can involve three to five banks in the chain. Each bank needs to reconcile the payment against its own ledger, check compliance requirements, and manage foreign exchange conversion. This is why international wires historically took three to five business days and cost $25-50 in fees per leg.

SWIFT gpi improvement: SWIFT's Global Payments Innovation (gpi) initiative, now used by 4,450+ financial institutions handling over $530 billion in daily value, has significantly improved transparency. Nearly 60% of gpi payments are credited to end beneficiaries within 30 minutes, and 90% of cross-border payments reach the destination bank within one hour. But gpi improves visibility into the existing correspondent chain: it does not eliminate the chain itself.

The Cost of International Payments

According to the World Bank's Remittance Prices Worldwide database, the global average cost of sending remittances was 6.65% in Q2 2024. Banks are the most expensive channel at 12.66% on average. Digital remittance providers average 4.96%, but even this far exceeds the UN Sustainable Development Goal target of 3% by 2030.

Costs vary dramatically by corridor. Sub-Saharan Africa is the most expensive receiving region at 7.73%, while South Asia is the cheapest at 5.16%. These costs fall disproportionately on the people who can least afford them: migrant workers sending money home to support families.

How SEPA Works: Europe's Single Payment Area

The Single Euro Payments Area encompasses 41 countries: the 27 EU member states, the three EEA countries (Iceland, Liechtenstein, Norway), and 11 additional territories including Switzerland, the UK, and several Crown dependencies. SEPA's goal is making euro payments across borders as simple as domestic transfers.

SEPA Credit Transfers and Direct Debits

SEPA operates two core payment instruments. The SEPA Credit Transfer (SCT) enables one-time or recurring euro payments across the zone, using IBANs for account identification. Standard SCT payments settle within one business day. The SEPA Direct Debit (SDD) allows creditors to pull funds from debtor accounts with prior authorization, following mandate-based rules for consumer and business-to-business transactions.

SEPA Instant: SCT Inst

SEPA Instant Credit Transfers (SCT Inst) launched in November 2017, making the eurozone one of the first large economic regions with a standardized instant payment scheme. The 2025 rulebook mandates execution within 5 seconds, a hard processing limit of 7 seconds, and final confirmation by 9 seconds. The system operates 24/7/365.

The European Union adopted the Instant Payments Regulation on March 13, 2024, mandating that any payment service provider offering standard credit transfers must also offer instant transfers at no additional charge. Eurozone PSPs were required to receive instant transfers by January 9, 2025, and must be able to send them by October 9, 2025. Daily instant payment volumes increased 72% in 2024 compared to the prior year.

TARGET2 and Settlement

Behind SEPA sits TARGET2 (now T2 after its March 2023 migration): the Eurosystem's real-time gross settlement system. More than 1,700 banks connect directly, with over 55,000 banks reachable through the network. In 2024, T2 traffic reached its highest level since the introduction of the euro: nearly 108 million transactions.

For instant payments specifically, the ECB operates TIPS (TARGET Instant Payment Settlement), which saw a five-fold increase in settled transaction volume in 2024, reaching approximately 1.35 billion transactions. EBA Clearing's RT1 system also processes instant payments, hitting a record of 95.5 million monthly transactions in September 2024.

Fedwire and CHIPS: Where US Dollars Actually Settle

While ACH handles high-volume, lower-value payments, the US has two systems for large-value settlement: Fedwire and CHIPS.

Fedwire Funds Service

Fedwire is the Federal Reserve's real-time gross settlement (RTGS) system. In 2024, it processed an average of 836,322 transactions per day with a daily average value of approximately $4.51 trillion. The average transfer is $5.4 million, though the median is around $18,000: a small number of very large transfers dominate the value.

Fedwire settles in real time: each transaction is final and irrevocable when processed. It currently operates 22 hours per day on business days, with planned expansion to include Sundays and weekday holidays in 2028-2029.

CHIPS: Netting for Efficiency

The Clearing House Interbank Payments System processes approximately 565,000 payment orders per day worth $1.8 trillion. Unlike Fedwire's real-time gross settlement, CHIPS uses a patented multilateral netting algorithm that achieved a 29:1 ratio in 2024: every $1 in intraday funding supported $29 in settled payment value. This saves participant banks an estimated $5+ billion annually in liquidity costs.

Clearing Houses: The Hidden Plumbing

Between messaging and settlement sits clearing: the process of matching, validating, and netting payment obligations between institutions. A clearing house acts as the central counterparty, becoming the buyer to every seller and the seller to every buyer. This multilateral netting reduces the number of actual settlement transactions required.

Without netting, if 50 banks all owe each other money, you would need up to 2,450 individual settlement transactions. With multilateral netting, the clearing house calculates each bank's single net position: one payment in or out. This is how CHIPS achieves its 29:1 efficiency ratio.

In the US, The Clearing House operates both CHIPS and the RTP Network. In Europe, EBA Clearing runs EURO1, STEP2, and RT1, connecting over 4,800 financial institutions. These organizations also collaborate with SWIFT on the IXB pilot for immediate cross-border payments, linking US and European instant payment systems.

Why Money Movement Is Still Batch-Based

Given that real-time payment technology exists, why does the majority of money still move in batches? Three factors explain the persistence.

First, economics. Batch processing is cheaper per transaction. Grouping thousands of payments into a single file and processing them together reduces the overhead for banks and operators. For routine, predictable payments like payroll and recurring bills, real-time processing adds cost without meaningful benefit to the end user.

Second, legacy infrastructure. Many core banking systems were designed decades ago around batch processing. They cannot handle continuous processing, immediate posting, or API-first interactions without significant middleware layers or full ledger redesigns. The cost and risk of replacing these systems while maintaining 24/7 operations is enormous.

Third, risk management. Batch processing provides a window for fraud checks, compliance screening, and reconciliation. Real-time payments compress this window to seconds, requiring fundamentally different approaches to transaction monitoring and anti-money laundering compliance.

Real-time alternatives are gaining traction. FedNow, launched in 2023, saw a 1,200% year-over-year increase in transaction volume by Q1 2025, with 1,400+ financial institutions participating (95%+ of which are community banks and credit unions). The RTP Network processed 98 million transactions worth $80 billion in Q4 2024 alone. But these remain a fraction of overall payment volume: instant payments are projected to constitute just 22% of the global payments mix by 2028.

The Trapped Liquidity Problem

The correspondent banking model requires banks to pre-fund accounts at partner institutions around the world. These nostro accounts must hold enough liquidity to cover expected payment flows: if your customers regularly send $10 million per day to Japan, you need at least that much sitting in your Japanese correspondent's account.

Industry estimates put the total capital trapped in prefunded cross-border accounts at roughly $27 trillion globally. A top-30 global bank may hold $10-25 billion in nostro balances across its correspondent relationships. In a 5% interest rate environment, $1 billion held idle in a foreign settlement account represents roughly $50 million in annual opportunity cost.

The problem is compounding. Correspondent banking relationships declined approximately 25% globally between 2011 and 2020, driven by mounting compliance costs and increased regulatory scrutiny. When banks exit correspondent relationships (a trend called “de-risking”), remaining correspondents become more concentrated. This means more capital trapped in fewer, larger nostro accounts: and more single points of failure in the payment chain.

When local banks lose correspondent access entirely, the effects ripple through the real economy. Research from UNCTAD shows that loss of correspondent banking relationships correlates with decreased exports and lower revenues for the affected country's businesses.

How Real-Time Payment Systems Compare Globally

The push toward real-time payments is a global phenomenon, but adoption varies dramatically. India's UPI leads the world with 89.5 billion real-time transactions in 2022, accounting for 46% of all global real-time payments. Brazil's PIX has seen explosive adoption since its 2020 launch. The UK's Faster Payments has operated since 2008.

These systems solve the domestic speed problem. But cross-border payments remain fundamentally constrained by the correspondent banking model. Real-time domestic rails cannot settle a payment between two countries with different currencies, different regulations, and different banking systems without intermediaries. The IXB pilot linking US and European instant payment systems is promising, but it still requires correspondent banks to bridge the gap.

Where Stablecoins Change the Architecture

Traditional payment rails separate messaging and settlement because they were built on systems designed to batch-process paper checks. Every subsequent layer of technology (SWIFT messaging, ACH automation, real-time payment overlays) has been added on top of this foundational separation. The result is infrastructure where sending a payment instruction and actually moving value are two distinct events with different timelines, different systems, and different failure modes.

Stablecoin payment rails collapse this distinction entirely. When you send a stablecoin, the message is the settlement. There is no batch window, no correspondent chain, no clearing house netting obligations at end of day. The transfer of the token is the transfer of value: messaging and settlement happen in the same atomic operation.

This architectural difference eliminates several categories of friction. There are no return codes because transfers are not provisional: either the sender has the balance or they do not. There is no trapped liquidity because there are no nostro accounts to pre-fund. There is no reconciliation burden because the shared ledger is the source of truth for all participants.

Atomic settlement matters: In traditional rails, a payment can be “in flight” for days: sent but not settled, visible but not final. This creates float, counterparty risk, and reconciliation overhead. When settlement is atomic, there is no in-flight state. The payment is either complete or it has not happened.

Spark: Settlement Without the Intermediary Chain

On Spark, sending a stablecoin like USDB is settling it. The transfer completes in seconds, operates 24/7/365 (no business day restrictions), and reaches finality without passing through correspondent banks, clearing houses, or batch processing queues. A payment from a user in the US to a recipient in Southeast Asia follows the exact same path and takes the same time as a payment between two people in the same city.

For cross-border B2B payments, this eliminates the correspondent banking chain entirely. There are no intermediary fees, no multi-day settlement windows, and no capital trapped in nostro accounts. For remittance corridors where traditional rails charge 6-12%, the cost reduction is dramatic.

The comparison is not theoretical. Traditional rails require three to five business days and multiple intermediaries for cross-border settlement. Stablecoin rails on Spark settle in seconds with no intermediaries. The gap between these two approaches is not incremental: it is a different architecture.

What This Means for Builders

If you are building a payments product, the choice of rails determines your user experience ceiling. ACH gives you low cost but multi-day settlement and return risk. Wires give you finality but high cost and limited hours. FedNow and RTP give you domestic speed but no cross-border capability. SWIFT gives you global reach but correspondent banking overhead.

Stablecoin rails on infrastructure like Spark offer a different set of tradeoffs: global reach with instant settlement and no intermediary chain, but with the requirement to manage on-ramp and off-ramp connections to the traditional banking system. The right approach depends on your corridors, your users, and your regulatory environment.

For developers looking to integrate stablecoin payments, the Spark SDK documentation provides the technical foundation for building on these rails. General Bread is one example of a Spark-powered wallet already enabling instant stablecoin transfers. For a deeper comparison of traditional and stablecoin payment infrastructure, see our research on stablecoin payment rails vs traditional systems.

This article is for educational purposes only. It does not constitute financial or investment advice. Bitcoin and Layer 2 protocols involve technical and financial risk. Always do your own research and understand the tradeoffs before using any protocol.