硅谷Fintech观察②｜为什么大家都去打银行：一张费池地图Silicon Valley Fintech Watch ② | Why Everyone Wants a Piece of the Banks: A Map of the Fee Pools

"Why China Has No Stripe" series · Part Two

In brief — A bank isn't one company. It's a stack of fee pools with wildly different margins, all riding on a single charter — payment acquiring, interchange, lending, net interest margin, cross-border FX. Each one is fat enough to sustain a listed company on its own. Over the past decade, the fiercest wave of American fintech startups poured almost entirely into this space. And the result? Most of the stars that charged hardest ended up cut in half, or worse. The ones who quietly made the real money were a different crowd. This piece lays out a map of the fee pools.

Let me start with something that just happened.

In May 2026, an American fintech called Parker shut down without warning. Not long before, it had nearly sold itself — the buyer was Avalara, which does AI tax-compliance, and the deal was worth close to US$90 million. Then, at the last moment, the buyer walked. With its backstop gone, the runway on its books vanished overnight. On May 7, Parker filed for Chapter 7 in the Delaware bankruptcy court.

What's Chapter 7? In plain terms, it's American bankruptcy liquidation — not a restructuring, not a chance to catch its breath, but the company simply closing its doors, selling off its assets, and winding up for good.

Parker was a fintech providing bank accounts and credit cards to small and mid-sized businesses, and it had long told the world it had "raised over US$200 million." But of that US$200 million, US$125 million was actually a lending facility — money to lend out, not the company's own equity. Meanwhile its CEO was still telling the world the company had US$65 million in revenue. The numbers looked great, yet the company was gone in a blink. What broke it wasn't a lack of customers; it was that the acquisition it had already lined up fell through at the eleventh hour — the buyer pulled out, and everything went to zero.

Parker died on the road to "taking on the banks." But it wasn't a one-off. It was just a footnote in the enormous gravitational field of this business.

Last time we covered how the "unbundling" of American brokerage and asset management was the story of the previous decade — and one that was largely finished long ago. Commissions were unbundled by Robinhood, advisory fees by robo-advisors, the promise of beating the market by index funds, and even the "fund" wrapper itself was pried apart by direct indexing. Unbundled all the way down, fees trended toward zero. On that track, you can no longer go after a 0.03% S&P fund, nor can you build another independent, large brokerage in a landscape where the giants ship for free and monopolize the scale profits.

So the smartest money swung its guns around and poured, in lockstep, toward the banks.

This piece covers two things. First, we'll take apart that behemoth called "the bank" and draw you a map of the fee pools — payment acquiring, interchange, lending, net interest margin, cross-border FX — how big each one really is, who's slicing into it, and how well they're doing. Second, we'll offer a judgment more counterintuitive than "everyone's taking on the banks": the star companies attacking the consumer side of banking mostly went on to do quite poorly; the ones who actually crossed the cycle and quietly made the big money were a different crowd — the "people selling shovels."

Numbers are drawn from the most recent 2026 figures wherever possible. Read on.

I. A Bank Isn't One Company — It's a Stack of Fee Pools

We're used to thinking of "the bank" as a single whole. But from a fintech vantage point, a bank is really a stack of mutually independent fee pools with wildly different margins, all riding on a single charter:

• Payment acquiring — every time a merchant collects money, a fee is skimmed off;
• Interchange — every time you swipe a card, the issuing bank skims a "toll";
• Lending and consumer credit — lending you money, earning interest and fees;
• Net interest margin — using your deposits to lend, earning the "borrow short, lend long" spread;
• Cross-border FX — moving your money from one country to another, earning the exchange spread and fees.

Each is fat enough to sustain a listed company, and each is severed enough from the others that it can be sliced off on its own. That's why there are so many players "taking on the banks": they don't need to conquer the whole city — just pry loose one of its bricks.

Here's an analogy that makes it click. If SaaS took one clunky enterprise software suite and broke it into modules you could subscribe to individually, then what fintech does is take a bank and break it into individual "function keys" that can each be billed separately. Stripe pried loose the payments key, Wise the cross-border key, Affirm and SoFi the lending key, Chime the deposits key, Brex and Ramp the corporate-card key — nobody wants to build a bank from scratch; everyone is taking one apart.

But before we walk this map pool by pool, hold on to one line that runs through the whole piece: within this stack of fee pools, the engine that truly drives growth has never been the "deposits" everyone assumes, but "payments." Deposits are "money at rest," sitting on the books waiting to be lent out; payments are "money in motion," and every time it moves, a fee is generated. JPMorgan is JPMorgan because of the payments franchise it clears — more than US$10 trillion a day — not because of how much deposit money is parked on its books. Whoever chokes the "flow of money" holds the fattest, and hardest-to-replace, pool of all.

The map below is drawn around "who sliced which pool, and how well they did it."

II. Interchange: The Biggest Pool, Now Being Pried Loose at the Foundation

Let me first anchor the concept in plain terms — it's the least familiar to Chinese readers and the most consequential in America.

Interchange: every time you swipe a credit card in the U.S., the merchant is skimmed a "toll." That fee is cut into pieces, and the biggest piece goes to the issuing bank — which then uses that money to subsidize your points, cashback, and airport lounges. The whole American premium-card flywheel of "spend more, earn more" runs on this fuel.

In 2026, this fee pool saw its biggest tremor in twenty years: a merchant lawsuit that had dragged on for two decades finally reached a settlement with Visa and Mastercard, preliminarily approved in June by Judge Brian Cogan.

There's a trap here that both the writer and the reader can easily fall into, and we have to clear it up first: these are actually two different settlements. One is the cash settlement compensating for past losses, with a fund of US$5.54 billion, paid to merchants who overpaid on fees in the past; the other is a forward-looking fee-reduction settlement, priced at US$38 billion — but that US$38 billion isn't a lump of cash, it's "the cumulative amount merchants will save on interchange over the next five years." The terms of the latter are what truly move the foundation: credit-card interchange rates cut by 10 basis points over five years, standard consumer-card rates capped at 1.25% over eight years, and — for the first time — merchants allowed to selectively refuse the most expensive premium cards by card tier.

It sounds as if the foundation of the premium-card flywheel has been shaken loose. But the truly interesting part is the twist that follows — the flywheel won't collapse.

Why? Because merchants simply don't dare turn away their best customers at the checkout. In America, nearly nine-tenths of all credit-card spending happens on rewards cards, and this is precisely what gives the card networks the confidence to concede: they're betting that no merchant will shut out the crowd swiping top-tier cards — the highest-spending customers there are — just to save a sliver of fees. A more hidden layer: those 10 basis points that regulators fought so hard to squeeze out very likely never reach the merchant's hands — under the "bundled pricing" models of the likes of Square and Stripe, the saved spread is quietly pocketed by the payment processor, and only the big merchants on "cost-plus" pricing actually benefit. The small merchants, repeatedly "represented" by lobbyists on all sides, turn out to be the silent losers of this settlement.

In fact, this "two-sided market" logic was rubber-stamped by the U.S. Supreme Court back in 2018. That year, Ohio and a group of other state governments jointly sued American Express (Ohio v. American Express), arguing that its rule against merchants "steering customers to lower-fee cards" amounted to monopoly and pushed merchant fees too high. The Supreme Court ultimately ruled 5–4 that Amex hadn't broken the law, on exactly this reasoning: card-swiping is a "two-sided business" — one side the merchant, the other the cardholder — and you can't fixate solely on the fee the merchant is overcharged; you also have to weigh how much of that money subsidizes cardholders' cashback and benefits on the other side. Put plainly, the court effectively conceded that the whole two-sided network effect — "skim from merchants, use it to subsidize cardholders" — is legal, even efficient. So when regulators today try to pry this foundation loose through fee rates, they're prying at a model the Supreme Court itself endorsed — no wonder it's hard going.

Now look at China: this flywheel basically doesn't exist there. China's card-swipe fees are pressed extremely low by regulators, and are almost entirely routed around by the QR codes of Alipay and WeChat Pay — the two of which together account for over 90% of Chinese mobile payments. Once the payment gateway is taken over by two QR codes, the entire chain of "the issuing bank skims a toll, then uses the toll to subsidize points" has no soil to grow in. That's why the word interchange feels self-evident to Americans and nearly alien to Chinese — our generation essentially skipped the golden age of credit cards altogether.

III. Lending, Net Interest Margin, Cross-Border: Three Pools Defined by Cycles and Friction

Start with lending and consumer credit. The loudest number in the media is that Americans are carrying US$1.25 trillion in credit-card bills they can't pay off — which of course makes a great headline. But there's a counterintuitive fact worth putting on the table first: credit cards actually make up only 6.6% of Americans' total debt; the real bulk is mortgages, at 70%. In other words, the "credit-card debt crisis" is terrifying in the narrative but only a supporting player on the balance sheet.

The new players slicing into this pool are "buy now, pay later" (BNPL).

What's BNPL? It's that "4 interest-free installments" button on the checkout page — essentially splitting a purchase into a few installments of short-term credit. Think of it as the American version of Huabei and the various installment plans.

Affirm and Klarna are the poster children of this track. But you have to remember the underlying nature of this business: at heart it's lending, and it eats from the plate of the credit cycle. In 2021, Klarna was still losing about US$700 million with roughly US$500 million in bad debt (SEK 4.6 billion); and its valuation duly slid from US$45.6 billion that year all the way down to US$15 billion when it listed in 2025. Lending can grow fast, but it has never been a business that sails easily through the cycle — a point that will keep recurring in Section IV.

Here's a character Chinese readers know better: DiDi. In Latin America, DiDi has long been more than a ride-hailing app — it has laid out an entire fintech stack across the value chain: lending in Mexico (DiDi Préstamos, with over 20 million loans disbursed cumulatively, of which about 70% of users had never borrowed a cent before), issuing credit cards (DiDi Card, with Mexico its first card-issuing market worldwide, breaking a million cards in a single year), and adding BNPL (called DiDi Paga Después, i.e. "buy now, pay in installments"); in Brazil, its 99Pay arm plugged straight into the local instant-payment system, PIX. A company that started in ride-hailing went overseas and built lending, card issuance, BNPL, and payments end to end — which is exactly why BNPL has never been a business that can stand on its own; it's just one link in the full chain of consumer finance.

Now look at net interest margin — in plain terms, the spread a bank earns from "borrowing short and lending long": paying depositors low, charging borrowers high, and pocketing the difference. Its defining trait is that it's highly cyclical: in Q1 2026, the industry-wide net interest margin in the U.S. slipped 8 basis points quarter-on-quarter to 3.31%, while over the same period industry net profit was actually still rising, reaching US$80.5 billion. Profit rising while the spread shrinks — that very divergence is the fingerprint of the cycle. A business that lives on net interest margin has its lifeline held by interest rates, not by itself. Interestingly, that year consumer-confidence surveys fell to a decades-low trough, yet bank executives saw actual spending and borrowing that was quite robust — PNC's CEO flatly said he "can't reconcile those headline consumer-confidence surveys with what we see in our business every day; they're almost completely at odds." The real temperature of a fee pool is hidden in transaction data, not in questionnaires.

Now look at the most fragmented — and most expensive — piece on the map: cross-border FX.

Why is this pool so expensive? It's worth spelling out in plain terms (because it's precisely the pain point that the next piece's "shovel-sellers" set out to route around): a small or mid-sized bank isn't qualified to plug directly into another country's clearing system, so it can only open a "correspondent account" at a large bank — and as a result, every wire is charged a fixed fee of US$5 to US$50, millions of dollars of funds are forced to sit idle in non-interest-bearing overseas accounts, and a single system false alarm means another US$20 to US$100 "investigation fee." Look further up, and even SWIFT often gets blamed unfairly: the root of the slowness and cost isn't the SWIFT network itself, but the cutoff times of banks and various central banks.

Precisely because the friction of the traditional rails is absurdly thick, the players who came specifically to slice off this cut are living well. Wise (think of it as the player that made cross-border remittance both cheap and transparent — akin to the role LianLian and PingPong play in China's cross-border collection and payment on the individual and SMB side) posted 2026-fiscal-year underlying income of £1.61 billion, up about 18% year-on-year, serving roughly 18.9 million cross-border customers, driving its cross-border take rate all the way down to 52 basis points while still holding a 26% pre-tax profit margin — and in May 2026 it dual-listed on Nasdaq (with London remaining its primary listing), at a market cap of about US$12 billion. Its founder built the company years ago because of two people living in London — one holding pounds to pay off a mortgage in Estonia, the other holding krona to cover living costs in London — both fed up with the exchange-rate spread the banks kept hidden.

IV. The Twist: Those Attacking the Consumer Side Mostly Suffered; the Shovel-Sellers Kept Ascending

Now to the pivot of the whole piece. Lay two kinds of companies from the same window of time side by side, and the contrast is jarring enough to read almost like a fable.

First, the consumer side — those stars aiming directly at you and me, out to "remake a bank."

Their growth numbers are often gorgeous, but glance for a second at unit economics and valuation, and the story changes:

• Chime, the benchmark American neobank, listed in June 2025 at a market cap of about US$11.6 billion — which sounds substantial, yet that's already more than a 50% cut from its US$25 billion peak on the private markets in 2021; after listing, the stock kept sliding, at one point down to around US$6.6 billion, and now back to about US$8 billion. It did turn its first-ever GAAP profit in Q1 2026 with revenue up 25%, but the market is still only willing to hand it a very low valuation — effectively writing "thin-margin, highly cyclical consumer side" straight onto the price tag.
• Klarna, the European poster child of "buy now, pay later" (BNPL — in plain terms, Huabei-style installment spending), saw its valuation fall from US$45.6 billion in 2021 to about US$15 billion when it listed in September 2025 — two-thirds of its value evaporated in four years.
• Colder still is the price of exits. The door wasn't closed — Chime and Klarna both went public — but they squeezed out at prices cut in half and then in half again. Per CB Insights, in Q1 2026 there were only 199 global fintech M&A deals, a six-quarter low; large exits above US$1 billion fell from 14 in the year-earlier quarter to 6.
• The unit economics are more startling still: one licensed neobank (Varo) grew its account count by about half in a year, yet the deposits per account came to just US$52 — which all but confirms its customers are either living paycheck to paycheck or treating it as a dispensable side account; and Robinhood's premium card offering "unlimited 3% cashback on all spending," on the card-swipe itself, runs at nearly negative margin — per Flagship Advisory Partners' breakdown, each cardholder earns it no more than US$140 a year (and mostly from interest). The real recovery comes from membership fees, sticky assets, and the US$656 million a year in margin interest that Gold users pay. Put plainly, this is a "lose money to make noise" business — the cashback is the bait, and the money is made elsewhere.

Lay out the whole consumer-side table and it gets clearer: Chime around US$8 billion, SoFi around US$22.2 billion, Affirm around US$26 billion, Block (Cash App's parent) around US$44.3 billion — apart from Robinhood, hyped up to around US$101.5 billion by the frenzy around crypto and prediction markets (which is itself a signal of the cycle), most are below their own private-market peaks, or fall far short of the top-tier infrastructure.

(For reference: these neobanks = pure digital banks, with no branches, acquiring customers through an app and retaining them with small features like early payday deposit. The closest Chinese analogues are the C-end retail arms of WeBank and MYbank.)

The shared predicament of everyone at this table: growth bought by burning cash, profit propped up by cross-subsidy, valuation rising and falling with market sentiment.

Now look at the shovel-selling side — the companies that don't face consumers directly but provide the payment infrastructure for everyone.

Over the same stretch of time, they've been ascending to godhood:

• Stripe is the most dramatic reversal. In 2023, in that Shixiang report it was still being written off as in a "death spiral," its internal valuation at one point knocked down to US$50 billion and even that seen as steep; by February 2026, its latest round of employee equity transactions set the valuation at US$159 billion, up 74% from US$91.5 billion a year earlier. In 2025 it processed US$1.9 trillion in transaction volume, up 34% year-on-year, equivalent to 1.6% of global GDP, and was "solidly profitable"; even Microsoft routed "a substantial portion" of its volume to Stripe, and its client list includes Nvidia and Amazon. It also made its largest acquisition ever — buying stablecoin company Bridge for US$1.1 billion. That old "darkest before dawn" bet paid off.
• Adyen, Europe's payment-acquiring giant, posted full-year 2025 net revenue of €2.36 billion, up 21% year-on-year (at constant currency), with an EBITDA margin as high as 53%, targeting over 55% by 2028 — that's a software-company-level margin, not one a financial company should have.
• One contrast is especially glaring (the next piece will dig into it): both starting from the "corporate spend" track, the one that chose to build infrastructure, Ramp, is valued at US$44 billion, while the one that pioneered the category yet chose to be absorbed, Brex, was acquired by Capital One in 2026 for US$5.15 billion — the former is 8.5 times the latter's exit price.

Behind that fallen Parker from the opening lies an entire front of SMB (small and mid-sized business) banking, and the divergence of fortunes on this front is even sharper than on the consumer side. All providing accounts and corporate cards to small businesses: Parker went straight into liquidation and out; Brex, which pioneered the category, was absorbed by Capital One for US$5.15 billion in early 2026, becoming part of a big bank; Mercury took a third path — in late 2025 it simply applied for its own bank charter, aiming to go from "borrowing someone else's bank charter" to "being the bank itself," and now has over 300,000 customers and processed US$248 billion in transaction volume in 2025.

The most interesting is Slash: it started out issuing cards to "sneaker resellers" (a crowd with big cash flow who, being young and unincorporated, couldn't even get a decent bank card), then Kanye West's antisemitic remarks crashed the entire sneaker market, and Slash's revenue evaporated by 80% overnight; the two just-dropped-out founders wrenched the wheel and pivoted — going instead to build vertical banks for niche industries like marketing agencies, crypto companies, and even air-conditioning installers, reaching US$250 million in annualized revenue in 2025, and in April 2026 raising US$100 million at a valuation of US$1.4 billion. Look at that: on the same SMB-banking track, some liquidate, some get absorbed by a big bank, some simply go become a bank themselves, some survive by a single pivot — those who can "quietly survive, or even thrive," either marry into a big bank, or find an extremely narrow, extremely deep vertical no one else can be bothered to touch.

In fact, this fault line was already drawn crisply back in 2022, when the market was at its most panicked. That year Coatue broke down the declines of listed fintech companies by sector: consumer-side SMB payments fell 55%, insurtech 49%, consumer finance 42%; while enterprise payment infrastructure fell only 23% and bank core software only 7% — at the time, just 13 enterprise-grade infrastructure companies together held up US$953 billion in market cap, 5.6 times the entire bank-software sector. Three years on, this structure not only hasn't reversed, it's been borne out even more thoroughly by reality: on one side Chime listed at a 50% cut and Klarna at a 50% cut, while on the other Stripe went from being written off all the way back to US$159 billion. The fault line didn't heal; it became a chasm.

The divergence of fortunes between these two sides is no accident — it's priced by the market with a crystal-clear valuation logic. A Q1 2026 tally puts it bluntly: enterprise-grade infrastructure fintechs carry valuation multiples averaging 17x, the highest tier; while payments/transfers command only 7.7x. Why? Because the former has "software economics" (recurring revenue, plus customers that expand), whereas the latter's growth relies on piling on more transaction volume, bearing more risk and cost — so the market values infrastructure like "software" and values payments and consumer like "financial services." Another data point from the same period corroborates the same direction: of the largest B2B payment fundraises, enterprise-grade infrastructure took 60%.

This is exactly the continuation of last piece's key: in this industry, valuation has never tracked how much money you "handle," but how much you can "make" from it — and at what multiple you get priced. The consumer side that handles trillions yet drives fees to zero is worth far less than the infrastructure side that handles transactions yet squeezes out software-level margins.

V. A Devil's Advocate: Is the Consumer Side Really "Suffering"? Maybe It's Exactly the Thing to Attack

As is my custom, in this section I'll argue the other side in earnest — maxing out the case for the consumer side before returning to my own judgment.

The bull case, argued seriously: attacking the consumer side of banking is actually a pretty solid story.

First, it really can be profitable. Chime turned GAAP-profitable after listing, with revenue up 25% and over 10 million members. Second, it has the world's strongest counterexample — Nubank: this Brazilian digital bank is now valued at about US$62 billion, with over 131 million users, covering roughly 62% of Brazilian adults, the largest private financial institution in Brazil. It proves a pure digital bank can grow into a beast that swallows a traditional big bank. Third, the consumer side reaches the customer directly and comes with its own distribution, unlike infrastructure, which has to read the faces of big clients — the moment Cash App launched stablecoins, behind it sat a user base of nearly 60 million. Fourth, even the "super-app" story holds up: Robinhood is now valued at as much as around US$101.5 billion, far above the vast majority of consumer fintechs, and it's laundering its image from "gamblers' carnival" into "the younger generation's gateway to wealth." Fifth, that scary "great deposit exodus" never happened either — Coinbase paid stablecoin holders over 4% yields for nearly three years, the promised deposit outflow simply never materialized, and over the past two years total U.S. bank deposits have actually risen.

Let me add one analogy for the other side: the Brazilian version of Google is still Google, but the Brazilian version of "JPMorgan" is Nubank — in emerging markets, homegrown C-end digital banks aren't impossible to build; they've slapped the original version flat on the beach. So betting on the consumer side isn't foolish.

But back to my judgment: this bull case slams into three walls — unit economics, valuation multiples, and the exit environment.

Take apart the two sexiest counterexamples first. Nubank's success is in large part Brazil's high-interest-rate dividend, not tech magic — in a country where the benchmark rate is routinely in the double digits, feasting on the deposit-lending spread alone is enough to grow a beast; move that model to a low-rate mature market and it fails instantly. Robinhood's high market cap is just as fragile: its profit still depends heavily on trading — especially on the extreme-cycle species that is crypto — and the moment the market chills and crypto drops, the profit shows its true form; and its new bet, prediction markets, has the regulatory sword of "is this a financial instrument or gambling in disguise?" hanging over its head.

Now the three walls. Unit economics — Varo's US$52 in deposits per person, Robinhood Gold cards earning under US$140 per account a year (Flagship's breakdown) — this kind of business is inherently propped up by cross-subsidy. Valuation multiples — the market only grants the consumer side the low "financial services" multiple (7.7x), reserving the high "software" multiple (17x) for infrastructure. The exit environment — Klarna's two-thirds of value evaporating in four years, Chime cut in half on listing, large exits above US$1 billion falling from 14 to 6 within a year — capital has already voted with its feet.

And what says it best is the choice of the consumer-side winners themselves: Chime, SoFi, and Mercury are all scrambling to get bank charters, because only by becoming a bank can they keep for themselves the spread they'd otherwise hand to a partner bank; the "second-wave neobanks" that can't reach this step can only wait to be absorbed by a super-app — which is precisely the opening of the story I'll tell in Part Four.

VI. Closing: After Unbundling, Who Crossed the Cycle?

Three sentences to close out this second piece.

First, the bank's fee pools are big and scattered, so everyone comes to attack them. Payments, interchange, lending, net interest margin, cross-border FX — each is a billion-dollar wedge in the U.S.; the worst legacy experiences leave the most openings. This was the most surging startup wave of the past decade.

Second, but the star companies attacking the consumer side of banking mostly went on to do quite poorly. Chime cut in half on listing, Klarna's two-thirds evaporated in four years, large exits falling from 14 to 6 within a year — they're mostly cyclical, thin-margin, cash-burning businesses, and a winner's fate is either to become a bank or to be absorbed by a super-app.

Third, the ones who actually crossed the cycle and quietly made the big money were a different crowd — the "people selling shovels." Stripe went in three years from being written off at US$50 billion to a solidly profitable US$159 billion, Adyen guards a software-level 53% margin, and the market — with a valuation gap of 17x versus 7.7x — has clearly priced "enterprise-grade infrastructure" and "consumer payments" as two different things.

So who exactly are these "people selling shovels," what makes their moats so deep, and how will the new rail of stablecoins rewrite the battle?

That's the subject of my next piece — "The People Selling Shovels: Stripe / Adyen / Airwallex."

Finally, a bonus: a physical exam of the American wallet — understand it, and you understand where the demand for BNPL and neobanks grows from.

(Data in this piece uses the most recent figures available as of 2026: Parker's Chapter 7 filing on May 7, 2026, its would-be acquirer Avalara and the roughly US$90 million deal, and the US$200 million in funding including a US$125 million lending facility come from public reporting; Ohio v. American Express is the U.S. Supreme Court's 2018 5–4 ruling; DiDi's Latin America lending / card-issuance / BNPL figures come from public reporting; Brex's US$5.15 billion acquisition by Capital One, Mercury's bank-charter application and US$248 billion in transaction volume, and Slash's US$250 million annualized revenue and US$1.4 billion valuation are all public disclosures; Stripe's US$159 billion valuation and US$1.9 trillion transaction volume come from its 2025 annual letter and February 2026 employee equity transactions; Adyen figures come from its full-year 2025 shareholder letter; Chime and Klarna market caps and IPO figures come from public-market disclosures; the two Visa/Mastercard settlement amounts are, respectively, the US$5.54 billion cash fund and the US$38 billion five-year cumulative fee reduction; the 199 fintech M&A deals and the 14→6 drop in US$1B+ exits come from CB Insights' State of Fintech Q1 2026; the Robinhood Gold card unit economics (US$140 / US$656 million) come from Flagship Advisory Partners' breakdown. Listed-company market caps are as of June 2026 and will fluctuate with the trading day.)