The MoatWide moat

The usual explanation of Visa's moat is the network effect, and the usual explanation is right but shallow. Yes, cardholders want the card merchants take and merchants take the card cardholders carry, and yes, that deadlock is very hard to break from nothing. But plenty of two-sided networks get built anyway, and several have been built in payments within living memory. The network effect explains why Visa is hard to replace. It does not explain why Visa earns a 60% operating margin while doing it, and the difference between those two questions is the whole company.

The person who chooses the network is not the person who pays for it. A cardholder decides which card to pull out and pays nothing for the privilege — frequently they are paid, in points funded by the very fee the merchant is objecting to. The merchant pays and has no say in which card arrives at the terminal. The issuing bank receives the largest share of that fee as interchange and therefore wants the arrangement preserved exactly as it is. Visa keeps a small slice and, crucially, writes the rules that allocate the rest.

In an ordinary market a supplier that raised prices would lose customers to a cheaper rival. Here the customer who would switch cannot, and the customer who could switch is being paid not to. That is not a competitive outcome and Visa does not pretend otherwise; it is a structure, and the structure is what regulators have been attacking for forty years while competitors have made almost no impression at all.

It also explains the strangest fact in this industry. Visa and Mastercard have run the same model against each other for fifty years and have never once competed by cutting the merchant fee, because neither collects most of it and neither would win a single cardholder by trying. The competition is real and it runs backwards: the way to win volume is to pay issuers more, which is what the $15,751 million incentive line is¹, and why it grows faster than revenue.

The third layer is arithmetic rather than structure. The network costs roughly the same to run whether it carries 200 billion transactions or 260 billion, so almost every incremental transaction is margin. Visa processed 257.5 billion in fiscal 2025, up 10%, and the operating margin rose with the count. A competitor starting today would have to fund the same fixed cost against a fraction of the volume and would then have to persuade fourteen thousand banks to switch — which returns to the incentive line, this time as the entry price rather than the retention cost.

The fourth layer is newer and less discussed. Having spent decades watching the traffic, Visa now sells what it learned from it: fraud scoring, tokenisation, authentication, dispute handling, consulting. That business reached $3.8 billion in the June 2026 quarter and grew 36%² — faster than the network beneath it, priced on what it saves the customer rather than on a share of the transaction, and bought willingly rather than imposed. It is rated below the other three facets here precisely because it is the one part of Visa that faces ordinary competitors doing ordinary competition.

What holds the first three together is that none can be attacked with a better product. A rival network offering lower fees, faster settlement and better technology would still face a cardholder with no reason to care and a bank with every reason to decline. This is why the serious threats to Visa have never come from companies. They come from governments, which can rewrite the rules the whole structure rests on, and which have already done so twice — capping interchange outright in Europe and mandating debit-routing choice in the United States. Both appear as root threats rather than here, because they are risks to the arrangement rather than weaknesses in it.

The number that tests all of it is return on invested capital: 38.5% against a cost of capital near 8%³. If the network's power is genuinely Visa's, that spread persists. If the toll is being legislated away a basis point at a time, it narrows first — and it will narrow before volume does, because volume is the last thing to go.