硅谷Fintech观察⑤｜美国银行 vs Visa：网络这东西，攥在手里就活不了Silicon Valley Fintech Watch ⑤ | Bank of America vs Visa: A Network Dies in a Closed Fist

Silicon Valley Fintech Watch series · Part Five (opening the "How the Giants Began" arc)

In brief. That Visa in your wallet was invented by a bank. In 1958 it dropped sixty thousand cards on one small California city in a single go. Not one recipient had applied for one. First-year delinquency hit 22% and the loss came close to twenty million dollars. Twelve years later, with the business finally making money, that bank handed the whole thing over to a group of other banks, its own competitors among them. The bank was Bank of America. Why would anyone do that? The answer to that question is being acted out on Wall Street all over again in 2026.

Two numbers first.

On 28 July 2026, Bank of America was worth about $446.5 billion. In July 2026, Visa was worth about $705.7 billion.

The son is nearly 60% bigger than the father.

Yes, Bank of America gave birth to Visa. The largest payment network on earth was christened BankAmericard, invented by Bank of America in 1958, surrendered by Bank of America in 1970, and renamed Visa only in 1976. Today, every card Bank of America issues and every swipe it collects on still pays Visa a network fee.

I only recently traced this line properly, and I am a little embarrassed about it. All this time watching fintech and I had never sat down with the most important family tree in the industry. When I laid out the hundred and twenty years from 1904 to 2026, it turned out not to be a company history at all. It is one rule, proven three separate times:

A network you hold in a closed fist cannot live.

So this piece walks those hundred and twenty years honestly, from 1904 all the way to 2027. Yes, part of it is still in the future. The first four pieces were about things happening right now. This one goes backwards, to see how the machinery we live inside actually grew.

1. 1904—1906: the only startup capital was trust

The story starts with a man the respectable banks looked down on.

In 1904, Amadeo Peter Giannini and five partners opened a bank in San Francisco's North Beach and called it Bank of Italy. The name was literal: its customers were the Italian immigrants working the city's docks.

What did an American bank look like then? It served merchants and the wealthy. Working people who needed to borrow went to loan sharks. Giannini did two things that were considered undignified at the time: he lent to small farmers and small businesses, and he went out into the street to find customers. A bank soliciting customers, in 1904, was beneath the profession.

Deposits and loans grew roughly fivefold inside a year.

Then came 18 April 1906, and the San Francisco earthquake.

That day is the bank's real beginning. With the fires closing in on North Beach, Giannini and a handful of staff who had rushed in overnight emptied the vault, hauling the gold, the cash and every ledger out of the city. (There is a much-repeated version in which he hid the gold under a wagonload of orange crates. I could not verify that detail in any authoritative source, so take it as folklore.)

Days later, while every other bank in the city was still waiting for its vault to cool and would not dare open, Giannini set a plank across two barrels on a North Beach street and started taking deposits and making rebuilding loans. A loan closed on a handshake. No collateral, no credit check.

He was the first bank in San Francisco to reopen. For the months when the city needed cash most, one man was very nearly the whole cash flow.

There is something here that I think is the foundation of this entire piece: a bank has never sold money, it sells trust. Money is fungible; a hundred dollars is a hundred dollars wherever it comes from. Whether you will lend when nobody else dares, and whether people believe you will honour what you lent, is the only difference there is. For a few days in April 1906, a plank and a handshake demonstrated that cleanly.

Over the hundred and twenty years since, this bank has changed its name, its headquarters and its owners. But every time it has made serious money, that is still the thing it was selling.

2. 1930—1958: build the network first, then drop the cards

What Giannini did next was copy the model across California.

American banking at the time ran on unit banking: one bank, one office. Giannini built the first statewide branch network in the country. By 1930 he had 292 branches and more than $1 billion in assets. On 1 November 1930, Bank of Italy formally became Bank of America.

By 1945, with 493 branches in California and about $5 billion in assets, it was the largest commercial bank in the world. When Giannini died in 1949, the bank that started from a single counter in North Beach had more than 500 branches and about $6 billion in assets, first in the world by deposits.

That branch network is the key to everything that follows: it was the bank's first network-effect asset. Money you deposited in San Francisco could be withdrawn in Fresno; the more branches there were, the more that was worth; the more it was worth, the more people deposited; the more people deposited, the more branches you could afford. Giannini turned that flywheel for forty years.

And it is precisely because it held a statewide net that what happened in 1958 was possible at all.

18 September 1958, Fresno, California. Bank of America mailed a plastic card to roughly sixty thousand people in that city, all at once. Not one recipient had applied. Most had no idea what it was or what it was for. This is the famous Fresno Drop.

One common misunderstanding to clear up first: BankAmericard was not the world's first credit card. Diners Club came earlier, in 1950, but that was a charge card, settled in full at the end of every month. BankAmericard's originality lies elsewhere: it was the first bank-issued revolving-credit card that ever worked at scale. You could pay part of the balance this month and roll the rest into the next, with interest. The business model of every credit card on earth today starts here.

So why choose a method as brutal as an airdrop?

Because payments are a textbook two-sided network, with a textbook deadlock: merchants will not accept the card because there are no cardholders, and consumers will not carry the card because no merchant accepts it. Chicken and egg, both sides waiting for the other to move.

Bank of America's answer was crude enough to look absurd even now: stop waiting, and simply flood one side. Sixty thousand cards into a single city, cardholders conjured out of nothing, and merchants will follow.

The cost was severe. None of these cards had been underwritten. They went to the unemployed, the already indebted, people who plainly could not repay; cards were stolen out of mailboxes and used fraudulently all the time. First-year delinquency hit 22%, against the project lead's optimistic forecast of no more than 4%. The first-year loss came close to $20 million.

In 1958, $20 million was the kind of number that gets an entire management team fired.

Bank of America absorbed it and did not stop. A few years later, BankAmericard started making money.

The significance of this is far larger than "they invented the credit card": it is the first brute-force solution to a two-sided cold-start problem that I know of. Every subsidy war since (ride-hailing paying drivers, delivery apps paying couriers, payment firms paying merchants) has run the same play: lose money flooding one side, then let the network effect take over. Those sixty thousand cards in Fresno are the ancestor of all of it.

3. 1970—1976: handing your own son to your enemies

Up to this point the story is still just "a clever bank pulled off something big." What made me want to write a whole piece is what comes next.

In 1966, Bank of America began licensing BankAmericard to other banks across the country and eventually abroad.

Why do that? Because without it the network could not grow. Why would a merchant outside California install its terminal? Only when the local bank also issued the card, and brought local merchants in with it, could the net actually spread.

Licensing worked quickly, and it drew rivals just as quickly. In August 1966, a group of banks met in Buffalo, New York and formed the Interbank Card Association, which became Master Charge and later Mastercard. Its structure was different from BankAmericard's from day one: it was a consortium a group of banks built together, not one company's product.

By around 1970, the BankAmericard licensing system had become unmanageable. Clearing, authorisation, reconciliation and disputes all landed on Bank of America's own operations, and with legal pressure on top of it, the choice in front of the bank became very clear.

In 1970, Bank of America handed over control and ownership of BankAmericard in its entirety. It went to a new company, National BankAmericard Inc. (NBI), led by Dee Hock.

The structure of that company is worth a look: it issued no stock, and member banks held non-transferable "participation rights." In other words it was owned in common by its issuers. Nobody could buy it and nobody could sell it. Hock later coined a word for this kind of organisation, neither hierarchy nor market: chaordic, from chaos and order.

In 1976, NBI became Visa. Why that word? Because it sounds nearly identical in almost every language. A company that intended from the outset to be global infrastructure picked even its name against that standard.

Now back to the question: why would Bank of America let go of its most valuable invention?

My answer: it had no choice, and it was one of the best decisions it ever made.

The logic is plain enough. A network's value comes from acceptance, and acceptance comes from competitors being willing to join. And no bank will willingly run its own payments business, long term, on a competitor's brand and a competitor's system.

As long as BankAmericard carried the Bank of America name, every other bank was only ever a tenant, always calculating when to move out. They were already moving, in fact, and Master Charge was the new address. The choice Bank of America actually faced in 1970 was not "own Visa or give Visa away." It was "give Visa away or have no Visa at all."

Neutrality here is not a moral posture. It is the physical precondition for the network growing at all.

Someone will object: hasn't American Express issued cards and run its own network the whole time, and isn't it doing fine? It is doing fine, but it is a different species. Amex runs a closed three-party model, issuing, acquiring and clearing for itself. The upside is that every cent stays home; the cost is that its transaction scale has never reached, and cannot reach, the order of magnitude of Visa and Mastercard. (Acceptance is a separate matter. Amex spent more than a decade closing that gap, and since the end of 2019 roughly 99% of U.S. locations that take credit cards take Amex. But the scale gap cannot be closed, because it is not a question of capability, it is a question of structure.) A closed loop can make money, but a closed loop never becomes infrastructure. That fork in the road happened more than fifty years ago, and there is still no third path.

My summary is one sentence: if you want to own a network, you first have to give up owning it.

4. 1998: the name stayed in San Francisco, the company moved to Charlotte

The timeline has to run on a bit further, or it is impossible to say who this Bank of America actually is today.

After letting Visa go, Bank of America kept doing what it was good at: laying networks, only this time made of branches and balance sheet. In 1983 it picked up Seattle-First National Bank (Seafirst) as it teetered on failure, prying open interstate banking; in 1992 it swallowed Security Pacific, then the largest bank merger in American history.

Then came 1998, the year most easily overlooked and the most consequential.

On 30 September 1998, NationsBank of Charlotte, North Carolina completed its merger with BankAmerica. The new company was called Bank of America Corporation, its chairman and CEO was NationsBank's Hugh McColl, and its headquarters was in Charlotte.

Note the direction of that sentence: legally, the Charlotte bank bought the San Francisco one, and then took the name of the bank it had bought.

NationsBank was itself the product of thirty years of continuous acquisitions by McColl, rolled up out of a North Carolina local bank. So after 1998, what sits under the Bank of America sign is really two entirely different sets of genes: San Francisco left behind the retail bank's instinct for trust, and Charlotte brought an acquisition machine. It was the second one that determined the shape of the company from then on.

The machine's twenty-first-century bill reads like this:

• 2004, FleetBoston Financial
• 2006, MBNA (credit cards)
• 2007, LaSalle Bank and U.S. Trust
• July 2008, Countrywide Financial, about $4.1 billion
• announced September 2008 and completed January 2009, Merrill Lynch, about $50 billion

The last two were bottom-fishing in the crisis, and the most expensive lesson in the company's history.

Merrill is still argued about today: losses surfacing in bulk right after the deal was announced, the bonus scandal, regulatory investigations, the shareholder class action (settled later for $2.43 billion). But with a dozen years of hindsight it turned Bank of America overnight into a top-tier wealth management and investment banking platform. That one nets out positive.

Countrywide was an unmitigated disaster. The purchase price was about $4.1 billion. Mortgage-related settlements and payouts arising from Countrywide and Merrill together came to more than $65 billion, including $16.65 billion in the 2014 settlement with the Department of Justice alone.

What does $65 billion mean here? More than twice the company's entire net income for 2025.

I include this stretch because it sets up an exact contrast with the main thread: in 1970 Bank of America gave up control and received a network that is still alive today; in 2008 it bought complete control twice and received a bill it spent twenty years paying off.

5. 2011: the deal that came out of a bathtub

At the point where the bill hurt most, somebody walked in.

In August 2011, with Bank of America's share price pinned to the floor by Countrywide's litigation, fines and provisions, the common stock traded at a 62% discount to book value. In plain English: the market believed that even if you liquidated this bank's assets, they were not worth what the books said, and forty cents on the dollar would be doing well.

Buffett thought of the deal in the bathtub. That is not a joke, it is what he told CNBC: the idea surfaced in the tub on a Tuesday, and on Wednesday he phoned Bank of America's CEO Brian Moynihan. Moynihan did not call a bank, did not tell his staff, and spoke only to his chairman. From idea to done: twenty-four hours.

The structure is worth looking at, because it is so very Buffett. Berkshire put up $5 billion for cumulative perpetual preferred stock paying 6% a year, and got, thrown in, warrants to buy 700 million common shares at about $7.14 each.

The 6% preferred was his floor: whatever the stock did, $300 million of interest arrived every year. The 700 million warrant shares were his ceiling: a free call option, not a cent down. Protected below, uncapped above. That is how he has always priced his help at the moment everyone else is panicking.

He exercised in 2017, took roughly $12 billion of paper gain in one go, and became Bank of America's largest shareholder along the way.

What matters about this deal in this piece is not how much he made. It is what he was actually buying.

He was buying the 1906 asset. Not technology, not a network, not a licence, but "this bank's deposits and this bank's trust are still intact, they are just sitting under an overdue bill." He was betting the bill would eventually be paid and the deposit franchise would not disappear. From the first deposit Giannini took across that plank in North Beach, the valuable thing inside this bank has not changed in more than a century.

And there is a detail here that I stared at for a long time after I found it:

In the first quarter of that same year, 2011, Berkshire had already bought Visa and Mastercard.

Which means that in 2011 the father and the son of this family tree entered the same investor's portfolio at the same time. Bought separately, and bought as two completely different companies. The market has never treated them as one thing, even though they share a root.

The end of the story happens in 2026.

Between 17 July 2024 and 30 September 2025 Buffett sold about 465 million Bank of America shares, roughly 45% of the position; together with earlier trims, about 516 million shares in all, close to half the peak holding. The reason is not mysterious: going into 2026, Bank of America's shares carried a 43% premium to book value. The thing that once went for forty cents on the dollar now costs a forty per cent premium. The cheapness was gone, and that is all.

At the end of 2025 Buffett stepped down as CEO and Greg Abel took over. On 15 May 2026 the first 13F of Abel's tenure landed: Visa cleared out, Mastercard cleared out (about 3.99 million shares at an average price of $525.64). Bank of America stayed, still among Berkshire's five largest holdings.

A portfolio held for fifteen years, and what survives is the father while the son is sold.

I am not going to treat that as a prophecy. A 13F is a quarterly snapshot, and a new management team can have a hundred reasons to reposition, most of which have nothing to do with the argument in this piece. But as a footnote to a hundred and twenty years of family history it is genuinely interesting: the most valuable company on this chain and the company that gave birth to it sat in the same investor's account for fifteen years, and neither ever became the other.

6. Two report cards, 2026

History done, now the present. This section is the research layer, so the numbers get dense. Bear with me.

The two side by side:

Bank of America's revenue is nearly three times Visa's, its net income one and a half times, its headcount more than six times. And the market values it at six-tenths of Visa.

Per head it is starker still: each Visa employee generates four times the net income of a Bank of America employee.

Why? Pull one card swipe apart and the answer is sitting there on the table.

When you swipe at a merchant, the total fee the merchant gives up (the industry calls it MDR, the merchant discount rate) breaks broadly into three pieces:

1. Interchange, roughly 1.4%—3.3%, goes to the issuing bank, meaning banks like Bank of America. This piece is usually 70%—80% of the total, by far the largest.
2. The network assessment, about 0.14% on a Visa credit card, goes to Visa, the smallest piece of all.
3. The acquirer markup, roughly 0.1%—1.5%, split among the acquiring bank, the processor and the gateway.

See it? The largest slice on this chain goes to the bank, and Visa takes the smallest one. And yet the market pays two and a half times as much for a dollar of Visa's profit as it does for a dollar of Bank of America's.

Which is the thing this series keeps coming back to: what matters is not how much you take, it is what kind of money you take.

Visa collects a toll. It does not lend, so it carries no credit risk; it does not take deposits, so it carries no run risk and no rate mismatch; its costs barely scale with volume, so every extra transaction is close to pure margin. Inside Bank of America's $30.5 billion of net income sits an entire credit cycle. It looks wonderful when the economy is good, and the moment the economy turns, charge-offs, provisions and regulatory capital all arrive together. The market has never priced those two kinds of profit the same way.

Two more yardsticks.

Visa's fiscal 2025 payments volume was about $14 trillion, total volume about $17 trillion, across roughly 257.5 billion transactions. Dividing net revenue by payments volume puts Visa's overall take at about 29 basis points: for every $10,000 of transactions it clears, Visa keeps around $29. (A note on basis: this is an aggregate derived from net revenue over payments volume, not the actual rate on any single transaction.)

For comparison, Adyen's take rate in Part Three was about 17 basis points. An acquiring infrastructure company and a global card network, taking cuts of the same order of magnitude. The shovel seller and the rail layer both earn a layer this thin, and both earn it with extraordinary stability.

Now Bank of America's card business: 2025 credit card spend of about $377.8 billion, year-end card balances of about $106 billion, about 3.53 million new accounts opened in the year.

$377.8 billion sounds large, but it is 2.7% of Visa's payments volume.

This is the fundamental difference between an issuer and a network: the issuer earns a thick layer of spread and interchange on its own $377.8 billion; the network earns a thin layer of toll on the whole world's $14 trillion. The thick layer carries risk, eats the cycle and supports 213,000 people. The thin layer has to manage none of it, and only needs the road to stay busy.

Worth mentioning: that toll is now being eyed. The Credit Card Competition Act in the U.S. Congress in 2026 essentially wants to take the choice of which network a transaction routes over away from the issuer and hand it to the merchant, forcing Visa and Mastercard to compete on the same card; Trump publicly backed it, calling the status quo extortion run wild. On a separate track, the interchange settlement Visa and Mastercard reached with merchants in November 2025 commits to cutting interchange by about 0.1 percentage points over five years, landing between the end of 2026 and early 2027, though large retailers including Walmart have objected outright and a judge has yet to approve it.

Neither is settled, but the direction is clear: the toll business is too good for regulators and merchants to keep watching it forever.

7. Devil's advocate: was letting Visa go actually a catastrophe?

As usual, let the other side speak in full first.

The bear case, stated seriously: this was the worst trade in Bank of America's history.

First, it simply looks terrible on paper. Had it not let go in 1970, it would today be holding a company worth $705.7 billion, nearly 60% larger than itself. It handed away a company more valuable than it is.

Second, what it handed away was the best business of the lot. No capital consumption, no credit risk, a 60% operating margin, able to ride through cycles. Those are precisely the scarcest properties in banking, and all of them went to a consortium that included its own competitors.

Third, and most ironic of all, it now pays Visa a network fee. Its own child charges it a toll.

All of that is correct. But back to the judgment, and my answer is the same one: the thing it would have been holding could never have grown to $705.7 billion.

The reason has already been said once, so let me nail it down again. A network's value comes from acceptance, and acceptance comes from competitors being willing to join. As long as that card carried the Bank of America name, other banks were only ever tenants packed and ready to move, and they were already moving, since Master Charge had been founded in 1966. History even supplies a ready-made control group for the counterfactual: Amex has issued cards and run its own network for more than fifty years, and has never reached Visa's order of magnitude. A closed loop can make money, but a closed loop never becomes infrastructure.

Somebody will push further: couldn't it have owned Visa and issued cards, taking both ends? No. That is exactly the one thing the other banks at the 1970 table would never have agreed to. Neutrality is not a posture, it is the precondition for the net spreading at all.

And this has to be said clearly: Bank of America did not give it away for nothing. What it surrendered was only the thinnest layer on the chain, the toll. The thickest layer, the issuer's interchange and spread, 70% to 80% of a swipe's total fee, it kept, and has kept all the way to 2026, where $377.8 billion of annual card spend is still generating revenue for it.

What it lost was not the fee pool. It lost only the right to charge a toll on top of the fee pool. And with that right it bought the existence of the fee pool itself.

Without Visa there is no trillion-dollar fee pool today, and its own $377.8 billion would have nowhere to go.

8. The same play, running for the third time in 2026

I was ready to wrap up here, and then found that the rule has not stopped. It is being staged on Wall Street right now, and staged almost identically.

The first time was NBI in 1970, covered above.

The second time is Zelle. A transfer network owned jointly by seven large American banks, which moved $1.2 trillion in 2025. Structurally it is cast from the same mould as NBI in 1970: no single bank owns it, and every bank uses it. Because if Zelle carried the JPMorgan name, the other six would not accept it for a day.

The third time is right in front of us. On 5 June 2026, JPMorgan, Citi, Bank of America and Wells Fargo, together with BMO, BNY, HSBC, PNC, Truist, TD and others, jointly announced a shared tokenized-deposit network: turning commercial bank deposits into on-chain tokens for round-the-clock interbank transfers, serving multinational corporate clients first, targeting launch in the first half of 2027. The operator is The Clearing House, the clearing company these banks already owned together.

Why the sudden partnership? Because stablecoins arrived at the door. Stablecoins in circulation are around $263 billion, and B2B stablecoin payments alone came to about $226 billion in 2025. By Bank of America's own estimate, as much as $6 trillion of bank deposits could be pulled away by stablecoins.

Forbes put it with charming bluntness: these banks are dusting off the old Zelle playbook.

To which I would add: the Zelle playbook was itself the NBI playbook from 1970. The same bank, Bank of America, wrote that script fifty-six years ago and is a co-signer of it today.

Will it work this time? We can test it against the two conditions history supplies.

There is no shortage of cautionary tales: the trade finance consortium we.trade went bankrupt in 2022, Marco Polo in 2023, Contour shut down in 2023, and the bank stablecoin alliance USDF has been dormant since 2024. Their common failure has been summarised well: when the parties do not feel the same pain, and the operator is an unfamiliar face, the alliance falls apart.

Why did 1970 work? Both conditions were met. Every issuer faced the same existential threat (Master Charge existed, and not partnering meant being eaten). And there was a neutral operator everyone could accept (NBI, which everyone owned a piece of and nobody could own outright).

In 2026 both conditions are present again. The shared existential threat is stablecoins draining deposits. The ready-made neutral operator is The Clearing House, not an unfamiliar face but old infrastructure these banks have shared for years.

So I lean optimistic on this one. Not because of the technology, but because the organisational structure it is copying was proven once fifty-six years ago, and the thing that came out of it is worth $705.7 billion today.

One boundary has to be added, though. What is being built is a tokenized deposit, not a stablecoin. The difference: a tokenized deposit is an on-chain claim on a bank deposit and stays inside the banking system, whereas a stablecoin moves dollars outside it. That the banks chose the former and not the latter is itself a statement of intent: they do not want to join the new world, they want to move the new world onto their own turf. Whether they pull it off is something we can reconcile in two years.

9. Closing: three sentences, and something personal

First, the through-line of these hundred and twenty years is not "a bank got big," it is one rule proven three times. NBI in 1970, Zelle later, the tokenized-deposit network in 2026. Every time, a group of mutually hostile banks discovered they could not do it alone, and handed the asset to a neutral body none of them could own outright. A network you hold in a closed fist cannot live.

Second, the one who takes the most on a chain is not necessarily the one worth the most. The issuer takes 70%—80% of a swipe's fees and Visa takes 0.14%; and the market pays 35.1 times for a dollar of Visa's profit against 14.6 times for Bank of America's. The difference is not how much you take, it is what kind of money you take: one rides the cycle, carries the risk and supports 213,000 people; the other only needs the road to stay busy.

Third, control and network effects sit on a seesaw. In 1970 Bank of America let go of control and got a network that is still alive; in 2008 it paid $4.1 billion for complete control of Countrywide and got a bill north of $65 billion. Those two trades ran in opposite directions, and they ended in opposite places.

Now the personal part.

When Giannini died in 1949, his estate was $489,000, about the level of a comfortable middle-class household at the time, while the bank in his hands was the largest in the world. In 1926 the board voted to pay him 5% of net profits with a $100,000 floor. He refused, and told the board to give the money away. In January 1928, $1.5 million of it went to the University of California. A line of his has survived: "The hardest job I ever had was keeping myself from becoming a millionaire."

At first I took that for a moral anecdote about a businessman, unrelated to the industrial rule running through the piece. The longer I sat with it, the more they looked like the same thing:

A founder who refused to clench the company's money into his own fortune, and a bank that refused to clench a network into its own property. The first is why the bank survived 1906. The second is why Visa is alive today.

Some things only grow if you open your hand.

That rule is not confined to payments. Whatever it is you are holding, a process, a dataset, an interface you will not hand over, the tighter you grip it the lower the ceiling it can reach. Everybody understands this when it is explained. Very few people in a hundred years manage it at the moment letting go is actually required.

Back to China. In Part Four we discussed why China cannot grow neutral infrastructure of the Stripe kind: the soil is missing. The history in this piece adds a more concrete footnote to that judgment. Visa was not grown, it was forced out of a bank's open hand. It needed a group of commercial banks of comparable size, none able to swallow the others; it needed them to face a common threat at the same moment; and it needed a neutral operator all of them would accept. China has UnionPay, but UnionPay is a state team, not a bank-owned consortium: it is designated neutrality, not negotiated neutrality. Those two kinds of neutrality grow different things.

Which is better is not something this piece will rule on. But at the very least, what happened at that negotiating table in 1970 deserves a careful look from anyone in the infrastructure business.

Tell me in the comments what you would like the next one to be about.

(Sources and basis. Bank of America's history from 1904 to 1998 is compiled from public materials including the OCC, Encyclopaedia Britannica, the Giannini Foundation and the American Bankers Association; the 1998 NationsBank merger date is per the SEC 8-K and contemporaneous reporting. The 60,000 cards of the BankAmericard Fresno Drop, the 22% first-year delinquency rate, the first-year loss of close to $20 million, the 1966 licensing, the founding of NBI in 1970 and the renaming to Visa in 1976 are all per public histories. Bank of America's 2025 calendar-year revenue of about $113 billion, net income of $30.5 billion, year-end headcount of about 213,000, credit card spend of $377.8 billion, card balances of $106 billion and about 3.53 million new accounts are per its 2025 annual report and 10-K; the market capitalisation of about $446.5 billion is on the 2026-07-28 close. Visa's fiscal 2025 (ended 30 September) net revenue of about $40 billion, GAAP net income of $20.1 billion, payments volume of about $14 trillion, total volume of about $17 trillion, about 257.5 billion transactions processed and headcount of about 34,100 are per its filings and 10-K; the market capitalisation of about $705.7 billion is on the 2026-07-28 close. Net margin, net income per employee, the market-cap-to-net-income multiple and the 29 basis point take rate are all derived by the author from the public data above and are not company-disclosed; the two run on different fiscal years and the market caps are taken on different dates, so cross-comparison is indicative only. The three-part swipe breakdown (interchange 1.4%—3.3%, Visa network fee about 0.14%, acquirer markup 0.1%—1.5%) reflects prevailing industry ranges, not the actual rate on any single transaction. Mortgage-related settlements and payouts tied to Countrywide and Merrill exceeding $65 billion in total, the $16.65 billion Department of Justice settlement in 2014 and the $2.43 billion Merrill shareholder class action settlement are per public reporting. Berkshire's August 2011 $5 billion preferred stock (6% annual dividend) and warrants for 700 million shares at about $7.14 are per Bank of America's 8-K of the period and contemporaneous reporting; the bathtub origin and the twenty-four-hour close are per Buffett's own interviews and public reporting. The roughly $12 billion paper gain on exercise in 2017, the 62% discount to book value at purchase in 2011, the 43% premium in early 2026, the sale of about 465 million shares (roughly 45%) between 2024-07-17 and 2025-09-30 and about 516 million cumulatively, together with Berkshire's first purchases of Visa and Mastercard in the first quarter of 2011 and the exits from both in the first-quarter 13F filed 15 May 2026 (Mastercard about 3.99 million shares at an average price of $525.64) while Bank of America remained a top-five holding, are compiled from 13F filings and public reporting; a 13F is a quarterly snapshot and does not represent real-time positions. The tokenized-deposit network of 5 June 2026, the first-half 2027 target launch, The Clearing House as operator, stablecoins in circulation of about $263 billion, B2B stablecoin payments of about $226 billion in 2025, Zelle's $1.2 trillion in 2025, the $6 trillion deposit exposure figure, and the failures of we.trade, Marco Polo, Contour and USDF are per public reporting from June and July 2026. Listed company market capitalisations move with the market.)