Payment Infrastructure Companies

Why a separate infrastructure layer exists at all

The companies grouped here do not sell payment acceptance to merchants. They sell to the companies that do, or to banks, or to software platforms that want to become payment companies without becoming payment companies. Twenty years ago most of this was one vertically integrated business: a processor built its own gateway, wrote its own risk rules, ran its own billing and reconciled its own settlement files. That model unbundled, and this category is the result.

Unbundling happened for the ordinary reason: each layer turned out to be a hard, specialized problem that rewarded focus. Writing a subscription billing engine that handles proration, plan migrations, dunning and revenue recognition is a different discipline from authorizing a card. Deciding in under a second whether a checkout is a fraudster is a different discipline again, and one that gets better with data volume across many merchants rather than within one. Once those functions could be bought as components, the economics of building them in-house stopped working for almost everyone.

The commercially important consequence is that the merchant-facing layer became the thin part of the stack. Acceptance is competitive, price-transparent and increasingly commoditized. The layers above and below it are where switching costs live, and switching costs are where margin lives.

The rules layer: card networks and clearing

At the base sit the institutions that write the rules everyone else operates under. Visa and Mastercard are card networks: they operate the switching systems that carry authorization, clearing and settlement messages, they write the operating rules that acquirers and issuers agree to follow, and they set the default interchange rates that determine the largest single component of what a merchant pays. Neither issues cards, lends money, nor holds a merchant account. They are frequently described as processors in the trade press, and they are not.

That distinction has a commercial edge. Because interchange is set by the network rather than by the acquirer, it is the one part of a merchant's cost that no provider can discount. A processor competing on price is competing on its own markup only. And because the networks also write the rules, they set the chargeback framework, the monitoring thresholds that decide when a merchant is placed into a remediation program, and the data standards that determine whether a business qualifies for lower commercial-card interchange.

The Clearing House sits alongside them on the bank side rather than the card side: a bank-owned utility operating the RTP instant payments network, CHIPS for large-value dollar transfers, the private-sector EPN automated clearing house network and image check exchange. It is infrastructure in the most literal sense, invisible to merchants and indispensable to everything they do. FIS occupies the other end of the same layer, selling core banking systems, issuer processing and debit network operation to financial institutions; it completed its exit from merchant acquiring in January 2026 when it sold the last of its Worldpay stake, which is a clean illustration of a company choosing which side of this line to stand on.

Bank rails and payment operations

Not every payment is a card payment, and for large-value or recurring business flows the card is usually the wrong instrument. Bank rails — automated clearing house transfers, direct debit schemes, wires, and the newer instant networks — cost less per transaction, carry different dispute rights, and settle on different timetables. They are also considerably harder to work with, which is why a category of company exists to make them tractable.

Dwolla provides a single API over ACH, RTP and FedNow with the transfers themselves performed by partner financial institutions. GoCardless specializes in pulling money from customers' bank accounts across the domestic direct debit schemes — Bacs, SEPA, ACH, BECS, PAD — plus open-banking-initiated instant payments, which is a different problem from card acceptance because the merchant is initiating a pull against a mandate rather than presenting a credential. Modern Treasury approaches it from the operations side, giving businesses one API and a ledger over their own bank accounts for ACH, wires, RTP, FedNow, checks and stablecoins.

The reason this category grew is unglamorous: reconciliation. A business moving money over bank rails at scale spends far more engineering effort on matching returns, tracking failures, handling exceptions and keeping an accurate ledger than it does on initiating payments. The rails are free-ish; the operations are not. Companies that underestimate this build a payments integration in a sprint and then maintain a reconciliation problem forever.

Embedded payments and PayFac-as-a-service, where the growth went

If one shift explains why this half of the industry has grown faster than the merchant-facing half, it is this one. Vertical software companies — restaurant systems, salon software, property management, medical practice management — discovered that payments could earn them more per customer than their software subscription did, because they already owned the merchant relationship and the workflow the payment sits inside.

Becoming a payment facilitator in full is a serious undertaking: network registration, sponsorship, underwriting capability, risk management, compliance obligations and real balance-sheet exposure to sub-merchant losses. PayFac-as-a-service exists to sell the economics without all of the obligations. Finix operates as a full-stack acquirer-processor with direct connections to the card networks rather than a reseller relationship with a legacy processor. Rainforest lets vertical software platforms embed payments under its own facilitator registration and competes explicitly on portfolio migration and contractual data portability. Moov offers card acceptance, ACH, instant payments, stored-value wallets and virtual card issuing through one API. Payrix, founded in 2015, has been independent, then owned by FIS, then folded into the Worldpay for Platforms brand, and is now part of Global Payments — a useful reminder of how quickly ownership in this category changes.

Growth here is also self-reinforcing. Every software platform that embeds payments removes a merchant from the addressable market of the direct acquirers, which is why several of the large processors in the merchant-facing category now sell platform products of their own.

The layers above the transaction: orchestration, billing, B2B and fraud

Four more categories sit above the authorization, each solving a problem that no processor solves.

Orchestration and payment intelligence. A merchant running more than one processor needs something to decide where each transaction goes, to retry intelligently when one declines, and to hold the card credentials independently so the processors can be swapped. Spreedly is a PCI Level 1 vault that stores payment credentials on the merchant's behalf and routes transactions across more than a hundred payment services without ever touching the money. Pagos reads data from a merchant's existing processors and gateways to monitor and benchmark performance, and processes, routes and holds nothing at all. Orchestration is genuinely valuable at multi-processor scale and pure overhead below it; the mistake is buying it before there is anything to orchestrate.

Subscription billing and merchant of record. Chargebee and Recurly manage the billing relationship — plans, proration, invoicing, dunning, churn recovery, tax and revenue recognition — on top of a merchant's own gateways, without processing the payment. Paddle does something categorically different: as a merchant of record it buys and resells the vendor's product, so Paddle is the legal seller, the name on the cardholder's statement, and the party registered for sales tax and VAT worldwide. For a small software company selling into dozens of tax jurisdictions, that transfer of liability is often the entire reason to buy.

B2B, accounts payable and accounts receivable. When one business pays another, the authorization is trivial and everything around it is not. BILL runs approval workflow and money movement for small and mid-sized businesses and their accounting firms. Tipalti handles global mass payouts to suppliers, creators and affiliates along with the tax documentation that comes with them. Melio lets a business pay a vendor by card even when the vendor does not accept cards, delivering an ACH transfer or a check instead. Corpay, renamed from FLEETCOR in 2024, combines commercial cards, cross-border payments and closed-loop fuel and lodging networks.

Fraud and risk decisioning. Forter returns an instant approve-or-decline on e-commerce transactions, signups and logins, and takes on chargeback liability for fraud on the transactions it approves. That liability shift is the product. A fraud tool that only scores transactions leaves the merchant carrying the loss and the decision; a tool that guarantees its own decisions has aligned itself with the outcome.

Vertical infrastructure: healthcare, billers, freight, stablecoin and BNPL

The remaining groups exist because certain industries have payment problems that general-purpose processing cannot reach.

Healthcare. Most of a US medical bill is not a payment problem at all; it is an eligibility, claims and remittance problem, and the patient's share is only knowable after the insurer has adjudicated. Waystar sits between providers and payers handling eligibility checks, claim submission, remittance, denial recovery and patient billing at very large scale. Cedar takes over the patient-facing part — communication, coverage, payment — on top of the provider's existing billing system rather than replacing it. Neither is competing with a merchant acquirer; both are addressing the reason the acquirer alone was never sufficient.

Bill payment and EBPP. Utilities, municipalities, tax offices and insurers collect recurring non-discretionary bills, where presentment, channel coverage and posting results back into the billing system matter more than checkout conversion. InvoiceCloud replaces the biller's own payment portal as a hosted service. Paymentus runs the bill-payment layer for billers and connects them to consumer channels including PayPal and Walmart.

Freight audit and payment. Cass Information Systems audits, rates, pays and reports on transportation and facility invoices for large shippers, and owns Cass Commercial Bank, through which it disburses the money. It is a category where the provider's economics often depend on float as much as on fees, which is worth understanding before comparing quoted service rates.

Stablecoin and crypto. BVNK gives businesses one API and a set of licensed entities for collecting, converting, holding and paying out between stablecoins and fiat, so a business can use stablecoin rails without holding crypto risk directly. It is the newest group here and the one with the least settled regulation, which is a reason to read the licensing carefully rather than a reason to dismiss it.

Buy now, pay later. Affirm underwrites instalment loans at the point of sale, originated through partner banks and funded through securitizations and loan sales, and charges no late fees. Klarna is a licensed bank that funds most of its lending from its own retail deposits and offers pay-now, pay-later and longer financing across 26 markets. For a merchant, BNPL is not a cheaper way to accept payment; it is a conversion and basket-size product bought at a materially higher cost per transaction, and it should be evaluated on incremental sales rather than on fees.

How to evaluate an infrastructure provider, and when not to buy one

Infrastructure is sold on the promise that it removes work. The honest test is whether the work actually disappears or simply moves. Several questions separate the two.

1. Who holds the licence, and who holds the risk? In embedded payments, bank rails and stablecoin especially, the important question is which regulated entity is performing the regulated activity and which party absorbs losses.
2. What happens to the data on exit? Stored card credentials, sub-merchant portfolios and ledger history are the real lock-in. A provider that contractually commits to portability is making a meaningfully different offer from one that does not.
3. Does the provider touch the money? Spreedly and Pagos deliberately do not; facilitators and rails providers necessarily do. This changes the counterparty risk, the settlement timetable and the regulatory posture entirely.
4. How is it priced against the value it claims? Fraud and orchestration products are often priced as a share of what they protect or save. That is defensible, but it requires a baseline measured before deployment, not after.
5. Who owns it this year? Consolidation in this category is relentless, brands get retired, and roadmaps change with owners. A five-year integration decision deserves a look at the ownership history.

And the cases where buying is wrong: a merchant that simply needs to take card payments does not need orchestration, a ledger API or a PayFac platform, and buying one adds an integration without adding revenue. A business with one processor has nothing to orchestrate. A company with a handful of subscription plans and one tax jurisdiction does not need a merchant of record; it needs a billing page. A software platform with a small customer base will not earn back the engineering cost of embedding payments, because the economics only work once the payment volume flowing through the platform is large enough to make a share of it meaningful.

The clearest way to think about this half of the industry is that it sells leverage rather than acceptance. That leverage is worth a great deal to a company already operating at the scale where it applies, and worth nothing at all to a company that is not there yet.

Frequently asked questions