The world is too loud. Read what matters.

Complex Systems

A wallet's profit engine is not interest on balances, it is money circulating inside the system

The core wallet ledger: money coming in pays a card-network toll, money moving between users inside the system is nearly free, and money going out can be charged again. What a wallet earns is the spread on flow, not interest on idle funds — and PayPal, Cash App and Stripe Link are all betting on that model.

PaymentsWalletsCard networksStripeFintech

The video won't play here. Listen to the audio instead:

It takes a wallet's cost structure apart honestly: the wholesale price of a rail can be under a dime while the platform charges its users tens of cents; Japanese operators go as far as opening accounts at several banks to route around interbank clearing. Even if you never touch payments, it works as a template for taking a business apart.

The argument · tap a timestamp to hear it

1:05

A wallet has to answer how it makes money from day one

The episode opens with a bill of the host's own: $100 paid to his barber over Venmo. Why, for a small transaction that already has perfectly mature payment rails behind it, does PayPal still want to insert itself in the middle and take a cut? From there the scope gets set quickly: this is about the electronic wallet as a business, not crypto private-key wallets. Because a wallet looks like something that holds and releases a balance, it lands immediately in the regulatory zone that is sensitive about anything resembling deposits. So a wallet has to answer the revenue question from the first day of design — if it cannot answer it, the product dies.

— Patrick McKenzie
5:14

Card fraud comes from issuer-side breaches, not scammers building websites

The biggest obstacle to early online commerce was that people would not hand over a card number, and the press kept amplifying the risk of stolen cards. The author points out that the reality runs the other way: the bulk of card fraud comes from mass data breaches at the issuing end, not from scammers setting up parasitic websites — "economies of scale protect criminals too." PayPal's value was standing between the user's card number and the party being paid, capping the risk of any single transaction. That is what gave it the standing to charge a rail fee close to 3%, and it planted the seed for how balances would later be designed.

— Patrick McKenzie
6:19

A wallet that is free to users is not remotely free to the wallet

Putting money into a wallet has to cross the card networks, historically priced at roughly 3%, and the wallet eats that cost. But once the balance sits inside the system, a transfer between users is close to editing one row in a database — too cheap to meter. PayPal charges near list price on both the way in and the way out, so simply substituting internal transfers for external ones adds a layer of gross margin. The wallet that is free to an ordinary user is not free to the wallet at all — and the zero marginal cost of the balance already sitting there is the real dam.

— Patrick McKenzie
9:21

Driving the rail fee down pays better than product innovation

There is no necessary relationship between wholesale and retail. Buying ACH transactions by the million costs around $0.05 apiece, while a payment service provider can quote a merchant close to $0.30; the gap is the tax the platform levies on the plumbing underneath. Japan is more extreme still: an interbank transfer costs roughly 200 yen per item, while a within-bank entry is nearly free — so some products open an account at each of the major banks, let a user's money move inside a single bank, and settle up with daily netting. In finance, pushing the rail fee down converts into wallet profit more directly than product innovation does.

— Patrick McKenzie
12:27

Cashing out is not a user perk, it is a second revenue pipe

Cash App prices the exit too: if you want it fast, you take the more expensive card rail, charged at 1.75% with a 25-cent minimum — the author tested a $50 withdrawal and was quoted 88 cents. Paired with that is issuing prepaid cards with institutions like Sutton Bank, so that money spent outside the system still hands the wallet a share of the swipe fee. The wallet no longer treats cashing out as a favor to the user but as a second revenue pipe, because the partner bank needs the wallet's user base and the wallet needs the bank's card-issuing license.

— Patrick McKenzie
20:53

On co-brand cards the wallet takes a share and eats the losses

The logic of a wallet partnering to issue a credit card is to charge on both sides, but the real battlefield is risk. Wallets typically assume the spending data they have accumulated will let them underwrite more accurately than the issuing bank. The industry has tested this repeatedly: FICO is already strong enough to be very hard to displace, and the extra variables that would materially improve losses tend to sit on the wrong side of the law. What actually prices the wallet is the contract — the bank requires the whole loan portfolio to stay under a loss-rate threshold, or the wallet's revenue share goes to zero and can turn negative. The 10-K language works around the phrase "loss sharing," but in substance the wallet is carrying the issuer's bad debt.

— Patrick McKenzie
24:50

Users care about the phone, not that piece of plastic

Apple's central argument to the banks is that users care about the phone, not the plastic rectangle of a credit card; and if the banks will not pay, Apple can back or build a new two-sided payment network and bring its users across. The reported figures are that Apple takes 15 basis points on every Apple Pay transaction while Google is said to take nothing — which the author uses as an occasion to knock how badly both giants execute on non-core products. The next layer is the conclusion that the checkout page itself can behave like ad inventory: once the item and the price are fixed, whose payment method gets listed first becomes an auction. Stripe has already replaced the traditional static display with ordering that adapts to user preference, and even lets merchants tune the options.

— Patrick McKenzie
31:13

Rail fees come down in the default settings, not at the negotiating table

Extend Link into a settlement-level negotiating layer: if a user stores their own low-cost payment method in Link, the next checkout can bypass the card networks and run off the bank account, and the saved rail fee gets redistributed between merchant and user. No single company can negotiate this sitting at one table; it is the aggregate of decisions by millions of users and merchants, which ultimately means the volume it represents on the other side of the payments negotiating table approaches 1% of global GDP. Only a total that size earns the right to reprice the rail fee.

— Patrick McKenzie

In their own words · checked verbatim

Organized crime does not outscale capitalism. The threat is when it gets to piggyback illegitimately on capitalism.

Patrick McKenzie5:14

The real prize is having vastly better economics on the transactions.

Patrick McKenzie8:19

Without asking me for further details or telling me how it would go out, Cash App quoted 88 cents, about 1.75%, sent it to Capital One, and delivered it in almost instant fashion as promised.

Patrick McKenzie13:34

FICO scores are unreasonably effective. Many, many, many teams have thought, I bet I can get better loss rates if I supplement FICO scores with another data source, and just about the only data sources for which that is actually true are illegal to use.

Patrick McKenzie20:53

This smells like an ads business.

Patrick McKenzie26:04

The number in the annual letter this year was 92%, by the way.

Patrick McKenzie29:13

But why shouldn't the Internet get the deal the Internet couldn't negotiate if the Internet were capable of negotiating on its own behalf?

Patrick McKenzie32:15

Figures

PayPal's common list price on an incoming transactionabout 2.9% plus 30 cents6:19
Interest income on customer balances disclosed in PayPal's 2023 annual reportclose to $500 million8:19
Cost of a single ACH debit when bought at scaleabout $0.059:21
Japanese interbank transfer pricingabout 200 yen per transfer, with within-bank transfers close to free10:22
2023 revenue of the relevant Cash App segmentabout $498 million12:27
Cash App instant cash-out rate0.5% to 1.75%, plus a 25-cent minimum fee13:34
Rail fee shown in the author's test $50 withdrawal88 cents, about 1.75%14:37
Per-transaction share of Apple Pay volumeApple reportedly 15 basis points, Google reportedly 024:50
Repeat-card share that recurs in Stripe's annual report92%29:13
Volume on the side pushing for cheaper payment railsroughly equivalent to 1% of global GDP32:15

Glossary

ACH
The US interbank batch transfer network; debits and credits over it cost very little, with wholesale pricing as low as $0.05 per item.
interchange
The slice of a card transaction's fee that goes to the issuing bank, usually a percentage of the amount; bypassing it is the point of the wallet model.
Durbin-exempt institution
A small bank exempted under US rules from the cap on debit interchange, often the entity actually issuing a wallet's co-brand card.
bps / bips
Basis points, a pricing unit in finance: 1 basis point equals 0.01%, so 15 bps is 0.15%.

How to listen

Who it's for

Product and strategy people in payments, wallets and acquiring — especially anyone trying to understand how PayPal, Stripe and Cash App actually compete.

Skip

14:49 to 18:09 is the Mercury and Granola sponsor read; skip it.