Visa's Real Moat Isn't Technology. It's a 1970 Ownership Trick Nobody Can Copy.

In 1958, in Fresno, California, a bank executive named Joseph Williams mailed unsolicited credit cards with a $300 limit to thousands of strangers and waited to be celebrated as a visionary. Instead, delinquency ran above 20%, fraud was rampant, and Bank of America lost nearly $9 million in just over a year. Williams resigned.⁴ The product that became Visa did not begin as a moat. It began as a near-catastrophe — and the company hid the moment it finally turned a profit, in May 1961 — a fact kept close to its chest to ward off potential competitors.⁶⁶

The official story of Visa's moat is the network: 234 billion transactions processed on its own rails in fiscal 2024¹, and by fiscal 2025, nearly 5 billion payment credentials available at more than 175 million merchant locations worldwide.² That's the standard answer, and it's wrong — not factually, but causally. Those numbers are the output of the moat, not the moat itself. The real defense was poured into the foundations in 1970, and it isn't made of fiber or code. It's made of paperwork.

The trick was making your rivals own the road with you

Dee Hock did not invent the credit card. He joined the BankAmericard network in 1968 — a decade after Fresno — to run a licensed rollout in the Pacific Northwest, and he inherited a mess: a sprawl of squabbling bank licensees with no shared rules and no trust. His genius wasn't a feature. It was a corporate structure. In June 1970 he led the negotiation that pried the program out of Bank of America's hands and reincorporated it as National BankAmericard Inc. — an independent company owned not by one bank but by the issuing banks collectively.⁵ Hock renamed it Visa in 1976, choosing a word he believed was instantly recognizable across many languages and denoted universal acceptance.⁵

Read that ownership change slowly, because it is the whole game. Every bank that issued the card was now simultaneously a competitor fighting for the same cardholders and a co-owner of the shared rails they all needed to compete. That's not a network effect in the textbook sense of 'more users make it more useful.' It's something stronger and stranger: a coordination lock. To leave Visa, a bank wouldn't just be switching a vendor. It would be walking away from an asset it partly owned, dissolving contracts with thousands of peers, and rebuilding fraud rules, settlement timing, and dispute arbitration from scratch — all while its customers expected the card to keep working in more than 200 countries and territories the next morning¹⁰.

And Visa itself takes none of the risk that makes the cards valuable. Its filings are blunt about it: Visa is not a financial institution, does not issue cards, does not extend credit, and does not set the rates or fees for account holders — the banks do all of that.³ Visa simply owns and runs the coordination layer, and charges a toll for access, authorization, clearing, and settlement. On $17 trillion of total volume in fiscal 2025, that toll came to $40 billion of net revenue.² The banks compete furiously on the surface; underneath, they are all paying rent to the shared road they were invited to co-own — and can't afford to leave.

Why a faster rail isn't enough to win

This is where the fintech pitch usually goes wrong. The standard challenger thesis is technological: card rails are old and slow, so a cleaner, cheaper, real-time alternative should win on merit. But the rails were never the hard part. Building a faster authorization system is an engineering problem, and engineering problems get solved every quarter. Re-coordinating thousands of mutually suspicious institutions into a single set of fraud-liability rules, settlement guarantees, and arbitration procedures — and getting them to trust it with their customers' money — is a problem of governance and incentives. That took Hock years to solve under conditions a startup can never recreate: he had Bank of America's existing licensees to herd, not a blank field. A new entrant has to manufacture the trust Visa was handed by history.

Isn't this just rent — and isn't China proof it can be beaten?

The honest counter has two sharp edges. The first: a position this entrenched looks like extractable rent, and rent invites attack. It has — Visa carried a litigation accrual of about $1.5 billion as of September 30, 2024, the bulk of it tied to long-running merchant interchange litigation, with more than $3 billion sitting in a U.S. litigation escrow.⁹ A moat that generates billions in legal liabilities is a moat under genuine, sustained assault. The second edge is the more interesting one: Visa is not the largest card network in the world by every measure. By debit-card purchase transactions and overall purchase volume, China UnionPay ranks first globally — but that lead is almost entirely a creature of its protected domestic market, where it holds roughly 95% of card spend.¹¹ UnionPay didn't out-coordinate Visa; it was handed a closed home court. Outside China, UnionPay accounts for less than 1% of global purchase volume, with Visa representing the largest share — and Visa's multi-decade coordination across borders is the asset no domestic champion can shortcut.¹¹

So the moat is real, but it is not a force field. It is defended rent — a coordination lock that holds because re-coordinating the world's banks is harder than building a faster payment app, and because outside the two markets where a government simply legislated a rival into existence, no one has done it. The lesson isn't that Visa owns the best technology. The 1958 launch was a fiasco; the technology has been copied a hundred times. The lesson is that in 1970, one restructuring turned a chaotic licensee network into a club its own members could never afford to quit — and that the most durable thing a company can build isn't a better product, but a structure in which its rivals are quietly invested in its survival.