Fraud Is the Product Nobody Argues AboutNarrow moat

Visa (V) — moat facet

The only revenue line Visa has that replaces a larger loss rather than imposing a cost — which is why it has no regulatory constituency against it.

Fraud detection is the purest expression of the data advantage. Visa sees the same card attempted at a petrol station in one country and an electronics retailer in another within the hour, and no individual bank does. Selling that judgement back to issuers and acquirers is a service they actively want rather than a fee they resent.

What the fraud model gets to learn from257.5bnTransactions processed, FY2025$3.8bnValue-added services, June quarter ($100m)One card seen across borders and institutions within the hour
Visa sees the same card attempted in two countries in an hour, and no individual bank does — which is what the service sells.

That distinction is worth more than it first appears. Every other revenue line Visa has is a cost imposed on somebody who would prefer to pay less, which is why every other line attracts litigation and legislation. This one is a cost that replaces a larger loss. It has no regulatory constituency against it, no merchant class action attached, and a customer who renews without being argued into it.

It is also the natural home for the AI investment. Visa describes early adoption and integration of AI models in payment systems as long-standing practice1, and fraud is precisely where model quality converts directly into a number a customer will pay for — a basis point of approval-rate improvement on a large issuer's portfolio is worth far more than the service costs.

Moat trajectory: Widening

Model quality converts directly into a price an issuer will pay, and the volume of traffic Visa can learn from grows every year.

The number that tests this moat
Reported
Contracted services revenue not yet recognised
$5.6B at 30 June 2026, about half due within two years

Remaining performance obligations relate mainly to value-added services such as risk tools. Contracted revenue growing shows clients committing to them for years; a shrinking figure would show the services being bought quarter to quarter.

Source: Visa Form 10-Q, Q3 FY2026 ↗
⚠ Threats to the moat
⚠ THREAT 01
The Data Advantage Has a Legal Ceiling
The value of the network's view is a function of how much of the network it is allowed to see at once.
References
  1. ReportedVisa describes early adoption and integration of AI models in payment systems as long-standing practice
    Visa Inc. Form 10-K for the fiscal year ended September 30, 2025 (CIK 1403161), Item 1 Business — the network reaches approximately 12 billion cards, bank accounts and digital wallets and more than 175 million merchant locations across more than 200 countries and territories; nearly five billion payment credentials; clients comprise nearly 14,500 financial institutions; Tap to Pay provisioning is live for more than 1.4 billion Visa credit and debit cards with more than 600 participating issuers; Visa Direct processed more than 12.5 billion transactions for more than 650 partners and can reach approximately 12 billion endpoints through more than 90 domestic payment schemes and more than 60 card and wallet networks; during fiscal 2025, 329 billion payments and cash transactions carried the Visa brand, an average of 901 million a day, of which 258 billion were processed by Visa; the Visa as a Service stack's access layer includes on-demand APIs and an MCP server enabling AI systems to interface with Visa Intelligent Commerce APIs; stablecoins, generative AI and agentic commerce are named as next-generation technologies under investment — FY2025 (year ended 30 September 2025) · publ. 6 November 2025 · source ↗
Sources
Generated September 23, 2026