Affirm: The Numbers Make Sense Once You File It Under Payment Network

Contents

1. Walk one transaction first
2. Why call it a payment network
3. Where a rate shock actually lands
4. Where the argument breaks
5. A few other things worth remembering
6. Worth reading alongside
7. The one thing to take away

Pride should not be nursed, desire should not be indulged, ambition should not be filled to the brim, pleasure should not be taken to its limit.

—— Book of Rites, Qu Li I (pre-Qin to Western Han; my own translation)

On 22 October 2026, Business Breakdowns EP.249 covered Affirm with Fraser Christie, founder and CIO of Sandstone Global Investment. His claim: Affirm belongs in the American Express file rather than the consumer-lender file, and the numbers back that filing. Christie cited gross merchandise volume above $50 billion growing more than 30% a year while the business is already two to three times its size of a few years ago; roughly 95% of transactions in any given quarter come from existing customers, and cash sales-and-marketing spend has barely moved since 2021. The claim carries a condition: the direct merchant integrations have to hold. Amazon alone is over 20% of the business, and Affirm trails Klarna internationally. Loosen either and the network-effect argument loosens with it.

Walk one transaction first

Christie uses a couch. You are at checkout on a merchant’s site and you see Affirm’s offers — 0% over three months, 20% over six, 25% over twelve. Those are quotes at this point, because Affirm has not pulled your credit yet. You tap through, hand over phone number, name, date of birth and the last four digits of your Social Security number, and Affirm combines the transaction data (which merchant, basket size, SKU details) with the credit pull to finalise terms. Average ticket is around $250. Affirm pays the merchant $245 and keeps $5 — the merchant discount rate, lower for big merchants, higher for small ones. On interest-bearing plans it also earns the interest you pay. On the funding side, about half of originations are sold to third parties and move off balance sheet; the other half stays on balance sheet using third-party facilities.

One detail deserves a pause: the rate is fixed, the schedule is fixed, and the balance runs to zero. A credit card works the other way — Christie’s figure is that half of US cardholders carry a revolving balance, with an average of $7,000. Every Affirm loan is underwritten on its own and designed to be paid off.

Why call it a payment network

Christie deliberately skips the Visa and Mastercard comparison. Those two are rails; the credit sits with the issuers above them. American Express integrates vertically — its own merchant relationships worldwide plus the credit — and it has traded at a discount to Visa and Mastercard for years precisely because of that credit component. Affirm has the same shape: direct integration at checkout, just online rather than at the café down the road, with the credit on its own book.

The network evidence is a set of figures that check each other. More merchants with the Affirm button means more consumers see offers; more Affirm customers makes the button more attractive to the next merchant, who will pick whichever option has the largest customer base. Frequency shows it too: the average user went from two transactions a year at IPO to seven. Customer acquisition cost is negative, because the loan that signs the customer up already makes money. Christie adds an honest footnote — Affirm issued equity to Shopify and Amazon when those deals were signed, so customers from those two channels carry a real cost.

Could something else explain the growth? There is a category of apps growing fast where you apply in-app, the credit lands on a virtual card, and you spend it wherever. No merchant integration, so none of that loop. Christie files those as lending businesses. The line holds up against one number: Affirm’s growth rate has not decayed as it tripled in size. Lending books get harder to grow; networks can get easier.

Where a rate shock actually lands

This was the part I kept thinking about, and it is the backbone of the whole framing.

The intuition is simple: borrow to lend, and rising rates hurt. Christie takes the mechanism apart. Affirm’s rate to the consumer is fixed, average loan life is 11 to 12 months, and a small slice of funding is floating. So a fast hike opens a short gap — the amortising book earns no more, and the new higher-rate loans have not flowed through yet. The book turns quickly: roughly 50% of it is at the new rate within a quarter, 75% within two. Christie cites recent disclosure that a 100 basis point rise is about a $90 million impact against roughly $2.5 billion of gross profit this financial year. Zero to 5% is the scenario that bites; 50 to 100 basis points over several months is not.

The real wound sits elsewhere. If a $500 twelve-month loan at 15% was good credit, the same borrower at 25% may not be worth the risk, so Affirm stops writing that cohort. Growth drops visibly. Credit losses do not blow out in parallel, because the rate on outstanding loans never changes. Losses did rise in 2022, by a different route: consumers were being squeezed on variable-rate debt elsewhere.

Compare a card book — people on minimum payments, balances compounding, the rate jumping from 15% to 25%, and the losses washing out all at once. Christie’s conclusion is that the two businesses take rate damage through different channels, and Affirm’s is shallower and more reversible. In the 2022–23 cycle the ugly print was the December 2022 quarter, which looked existential at the time; the next quarter inflected and the one after that was healthy on both growth and margin.

There is something usable here. When a headline says rising rates hurt lenders, ask two questions: whose rate is fixed, and how fast does the book turn. Fixed to the consumer and turning fast means the damage shows up in growth. Floating to the consumer and turning slowly means it shows up in losses. Those live in different places on the statements and heal at different speeds.

Where the argument breaks

Concentration first. The top five merchants carry a lot of weight and Amazon is over 20%. The episode has a live case: Walmart switched Affirm off for its own OnePay offering, powered by Klarna. The hit was smaller than expected for two reasons — Walmart was never a profitable merchant, given how hard it negotiates, and Affirm has since disclosed that a “large majority” of customers who shopped at Walmart still do, now paying with the Affirm card. That case does not transfer to Amazon or Shopify leaving. Christie names it as a risk, softened by long agreements and close relationships.

International is second, and here the logic turns on itself. Network effects mean merchants pick the button with the biggest base — globally, a merchant like Airbnb or eBay wants one button for every region, and today only Klarna offers that, so they went with Klarna. Affirm is being pulled into Canada, the UK and Australia by existing US merchants such as Shopify and Costco. The direct growth contribution is modest; what it buys is the ability to go head-to-head for cross-border merchants later. Some investors think those markets are saturated and not worth entering. Christie’s position is that the US fight requires it anyway.

Third is who owns the multiple. Market cap is around $25 billion fully diluted at roughly 20 times adjusted earnings with stock-based comp stripped out. Christie’s framing has a picture to it: growth investors show up when things go well, bank investors show up when they do not, and a Wall Street Journal piece about private credit funds gating is enough to move the stock. Growth investors have not done the work on credit; bank investors will not pay for founder-led technical culture. So the argument over 10 times versus 20 times earnings keeps running.

Last is agentic commerce. Payments and consumer finance sold off on the idea that agents will comparison-shop credit offers, which they will. Christie’s counter rests on two things: Affirm’s offers are generated at checkout and personalised per transaction and per user, and the 0% plans are merchant-funded — buy that $250 couch and you may get six or twelve months at zero, which no card or credit app can match. An agent fills in the details faster than you would, so friction falls rather than rises. That is a projection, not a result, so I would park it under “to be verified” and watch whether core GMV excluding the card holds up over the next few quarters.

That 0% mechanic has an angle I had not considered. Merchants see subsidised interest as expensive; Affirm’s pitch is that the alternative is a 20% storewide sale where every full-price customer also gets the discount and gross margin falls across the board. A 0% plan is a discount delivered one customer at a time. A fashion business can open a season with 0% offers to move volume and only reach 20 or 30% off at the end.

A few other things worth remembering

• Max Levchin started from his own ruined credit. He funded student-era startups on a credit card and paid late; after the PayPal IPO he flew to LA to buy a convertible to impress his then girlfriend, now wife, and the dealer declined the loan. He sat there waiting for a wire to clear while his face was on magazine covers. Friends kept the joke running for years, arranging with restaurant staff to pretend his card had been declined.
• His framework for picking problems is HVF — hard, valuable, fun — which was also the name of his incubator, where Affirm began. The hard end here means unsecured lending at size over long terms with no late fees, no compounding balances, no deferred interest.
• Starting at the hard end shows up in both the cost and the payoff. Levchin’s line is that lending someone $500 over twelve months gives them twelve chances to decide not to pay you. Affirm’s monthly installment cohorts run loss rates around 3.5%; short-term low-ticket BNPL is closer to 1%. Klarna started on the easy loans and is now working up; Affirm went the other way and is more than twice Klarna’s size in the US.
• The profit mix runs against intuition. Roughly 70% of GMV is interest-bearing, where merchant discount rates are small and the economics come from interest. Pay in X runs 1 to 3% gross margin on GMV; interest-bearing runs 4 to 6%; blended, about 4%. Which is why customers using the Affirm card out in the world, skewed even further toward interest-bearing plans, is good news for unit economics.
• Only 75 of the top 250 US merchants have Affirm integrated. Christie’s explanation for the pace is education — plenty of merchants treat card acceptance as a cost to grind down rather than a revenue lever.
• The discipline has hard numbers behind it: G&A flat in dollars for five years, stock-based comp flat for six, headcount flat for years. Levchin’s salary is under $40,000 with no cash bonus against a couple of billion dollars of stock; the two presidents have been there eight to ten years and each hold over $100 million.
• The most human detail: Christie met Levchin a few weeks before recording, and Levchin mentioned spending a recent weekend using AI to build an app controlling every TV and stereo in his house, because he was tired of digging through couches for remotes. Somewhere between 70 and 80% of the company’s code is already deployed by AI.

Worth reading alongside

• Business Breakdowns EP.249, “Affirm: The Internet’s American Express”, 22 October 2026, hosted by Avidan with Fraser Christie.
• Affirm’s recent analyst day and investor forum disclosures, which carry the gross margin bands by loan type, the rate sensitivity figure, and the card share of volume.
• The Visa episode in the same series with Alex Rampell, useful for contrasting rails-separate-from-credit against rails-integrated-with-credit.
• The long-standing valuation gap between American Express and Visa or Mastercard is public material, and it is the cleanest way to see how the market discounts a credit component.

The one thing to take away

One idea: whether a burden drags you under depends on whether it has an end date, not on how big it is. That is the design centre of the whole Affirm product — every loan underwritten on its own, rate fixed, balance walking to zero, so when the rate environment turns the existing book does not compound on you. All that breaks is whether new business gets written. A revolving card balance inverts it: the number can be modest and still have no end.

Here is something I tried. List everything that leaves your account every month — subscriptions, installments, premiums, the class you committed to, the favours you owe. One per line. Then write a date next to each: when does this stop. The ones with a date are fine. Circle the ones without, because those are the revolving balances on your own book. The first time I did it I circled four, and for two of them I could not remember why I had started.

Raw Markdown

This article is an educational discussion of investment method. It is not advice to buy or sell any individual security, offers no target prices, and does not analyze any current holding. Investing carries risk; make your own decisions or consult a qualified professional.