Affirm Holdings, Inc. (NASDAQ: AFRM) — The Inflection Is Real; The Recession Test Isn’t

Independent equity research — for general information only Report date: 2026-06-19 | Price: $73.92 (2026-06-18 close) | Market cap: ~$24.8B | FY ends June 30 | Latest reported quarter: Q3-FY2026 (ended 2026-03-31)

⚡ Claude’s Take

This block is the author’s own subjective opinion and general information only — not investment advice. The analysis that follows takes no position and sets no price target; it discusses valuation only as embedded expectations and scenarios.

Verdict: HOLD / not-a-short at $73.92; accumulate-on-weakness nearer the March-2026 lows (~$42–$55, ≈7–9× P/RLTC). Trim into strength above ~$90. Medium conviction.

Tag: “A real profit, wearing a network’s costume, priced for a credit cycle it has never run.”

Affirm has done something most of its 2021-vintage cohort never did: it grew into genuine GAAP profitability without abandoning growth. FY2025 was its first profitable year (+$52M); the nine months of FY2026 printed +$313M of net income on GMV up 35% and — the single most important quality signal — revenue-less-transaction-costs (RLTC) growing faster than GMV (4.04% → 4.14% of volume). Funding has de-risked dramatically (ABS spreads back under 100bps, funding cost down ~125bps year-on-year), the Affirm Card is a real owned-distribution flywheel (+146% GMV), and the CFPB’s BNPL overhang was lifted in 2025. This is no longer a story stock burning cash; it is a profitable, scaling consumer-finance network. That is why it is the believed name in a cohort (SOFI, Block/Afterpay, PayPal) of de-rated, given-up-on fintech lenders.

But the price already pays for that. At ~$25B the market sits between my bear and base cases — modestly below base — which means you are underwriting two things that are assumptions, not facts: that the ~4.1% RLTC margin survives a real consumer-credit downturn (Affirm has never been a public company through a recession), and that a moat the company calls a “network effect” is durable — when Walmart, a top-five merchant, walked to Klarna in March 2025 and proved these relationships are contractual and poachable. Layer on a 2.7 beta, an anti-low-volatility factor profile, realized drawdowns of −95% and −54%, a compensation scheme tied to GMV/RLTC/adjusted-operating-income (not GAAP earnings, FCF, or per-share value), and uniformly programmatic insider selling (zero open-market buys), and you have a high-quality-but-not-wide-moat business at a fair-to-full price with a fat left tail. I want a recession-discount entry, not today’s price. Framing: a high-beta, fundamentally-improving recovery name — not a crowded momentum long (its momentum loading is negative), not a falling knife (it’s profitable and up 74% off the low), but a levered bet on benign credit and a risk-on regime persisting.

Conviction: medium. Flips bullish if Affirm delivers a full year of GAAP profit with a stable/rising RLTC margin through a measurable consumer-credit softening — the first real cycle test, passed. Flips bearish if RLTC/GMV compresses below ~3.7% for two-plus quarters, a second top-five merchant defects, or on-book net charge-offs break above prior-cycle norms while funding spreads widen.

📈 Stock Price Action — Five-Year Event Map

Factual price history — not a recommendation, not a price target.

The arc. Affirm IPO’d in January 2021 at ~$97, spiked to an all-time high of $168.52 (Nov-4-2021) in the growth-tech euphoria, then collapsed ~95% to an all-time low of $8.91 (Dec-27-2022) in the rate shock. It has since round-tripped most of the way back, year-end close by year-end close: $9.67 (2022) → $49.14 (2023) → $60.90 (2024) → $74.43 (2025). The 52-week range is $42.53 (Mar-27-2026 low) to $92.18 (Sep-19-2025 high); today’s $73.92 sits ~20% below the 52-week high and ~74% above the March-2026 low, and still ~56% below its own all-time relative-strength peak.

Cycle narrative. Affirm is a textbook high-beta round-trip: a 2021 blow-off, a 95% rate-shock collapse to single digits, a multi-year fundamental recovery as the company cut its way to GAAP profitability, then a series of violent ±40–100% swings around earnings prints, the Walmart loss, the Apple win, the Klarna IPO, and the 2026 risk-off/risk-on cycle. The price moves are facts; every attributed cause is an interpretation cross-referenced to earnings dates, 8-K events, and the news record. No price target, no support/resistance — this is the price context the rest of the memo assumes.

1. Executive Summary

Affirm Holdings is the largest US pure-play “buy-now-pay-later” (BNPL) / point-of-sale (POS) installment lender, originating ~$49B of annualized gross merchandise volume (GMV) across 26.8M active consumers and ~515K active merchants. It is a hybrid: part two-sided commerce network earning merchant and interchange fees, part balance-sheet consumer lender earning interest, and part capital-markets machine that sells a large share of its originations forward to securitization and forward-flow investors for gain-on-sale and servicing income. Its signature is consumer-friendly product design — $0 late fees, simple (non-compounding) interest — and a 0%-APR, merchant-subsidized model that aligns it with retailers seeking conversion lift.

The investable fact of 2026 is an inflection: after cumulative losses exceeding $2.6B from FY2021–FY2024, Affirm turned its first GAAP-profitable year in FY2025 (+$52M) and accelerated hard in FY2026 (nine-month GAAP operating income +$270M, net income +$313M, diluted EPS +$0.90). Crucially this is profitable growth — RLTC (Affirm’s revenue-less-transaction-costs gross-margin proxy) grew ~41% over the nine months, faster than GMV’s ~37%, expanding RLTC/GMV to ~4.14%. Funding costs fell ~125bps year-on-year and ABS spreads tightened below 100bps for the first time since 2021. Sales-and-marketing dollars actually fell even as volume grew, helped by the roll-off of non-cash partner-warrant amortization.

The skeptical counterweight is fourfold. First, the moat is narrow and contestable. Management calls it a “network effect”; the evidence does not support that label. Consumer switching costs are near zero (shoppers multi-home across Affirm, Klarna, Afterpay at successive checkouts), merchant relationships are contractual, and Walmart — a top-five account held since 2019 — defected to Klarna in March 2025. The real advantages are scale economics, capital-markets access, and an underwriting-data intangible; genuine but not wide. Second, concentration is acute: the top-five merchants/partners are ~47% of GMV and Amazon alone is ~22%. Third, the business has never run a credit cycle as a public company. Roughly half of revenue is balance-sheet interest income and ~14% is volatile, mark-dependent gain-on-sale; both compress in a downturn precisely as funding spreads widen and GMV slows — a correlated triple-hit. Fourth, the equity is a 2.7-beta, anti-low-volatility regime instrument with realized drawdowns of −95% and −54%, and a management incentive scheme indexed to growth and non-GAAP profit rather than GAAP earnings or per-share value.

At $73.92 / ~$25B, embedded expectations require roughly 20% GMV compounding for the better part of a decade, a durable ~4.1% RLTC margin through a full cycle, and rising conversion of RLTC into GAAP earnings. That is good-but-not-heroic execution — not an obvious bubble, but a price that pays in full for an inflection whose single load-bearing assumption (credit-through-a-cycle) remains untested. This report takes no position and sets no target; the labeled Claude’s Take above is the only exception.

2. Business Overview

What Affirm does. Affirm operates a POS installment-lending platform that sits at the merchant checkout (online and, increasingly, in-store and in-app) and lets a consumer split a purchase into scheduled payments. It is not primarily a balance-sheet lender by intent, though it is one in practice: loans are originated through two FDIC-insured bank partners — Celtic Bank and Lead Bank — then funded by Affirm through a rotating mix of on-balance-sheet warehouse facilities and securitization trusts, and off-balance-sheet sales to third-party investors (forward-flow whole-loan sales and pass-through arrangements). (Source: FY2025 10-K, Business and “Regulatory Environment” sections.)

The three loan products (share of GMV, Q3-FY26):

• Interest-bearing monthly installments — simple, non-compounding interest, terms up to 60 months. ~70% of GMV (72% FY25, 74% FY24); the dominant and highest-AOV product, and the source of most interest income.
• Pay-in-X (0% short-term) — 1–4 interest-free installments; the classic “Pay-in-4” BNPL product. ~16% of GMV. Most of this volume runs through Shopify and is the fastest-growing segment.
• 0%-APR monthly installments — longer-term, interest-free to the consumer and merchant-subsidized. ~13% of GMV (14% FY25).

Affirm charges $0 in late fees and uses only simple interest — stated in every 10-K since FY2021 (“we do not capitalize on consumer misfortunes through practices such as late fees and deferred or compounding interest”). This is a genuine product and brand differentiator versus revolving credit cards, and — usefully — it insulates Affirm from the late-fee regulatory crackdown that threatens card issuers, because Affirm has no late-fee revenue to lose.

Surrounding the core lending products are the Affirm Card (a physical Visa card that pays by debit upfront or converts eligible purchases to installments afterward — the key direct-to-consumer growth engine), the Affirm Money Account (high-yield savings via Cross River Bank), and the Affirm app/marketplace (24% of FY2025 transactions originated in-app).

How Affirm makes money — revenue decomposition (FY2025, $000s):

• Merchant network revenue (~27%): the merchant discount fee for facilitating and converting a sale; richer on 0%-APR and longer-term loans. The closest thing to a pure platform “take.”
• Card network revenue (~7%): interchange from the Affirm Card and virtual-card rails; the fastest-growing network line (+53% FY25), tracking card adoption.
• Interest income (~50%): interest on loans Affirm holds on balance sheet. This makes Affirm materially a balance-sheet lender, carrying credit and funding risk — not an asset-light network.
• Gain on sales of loans (~12%, +94% FY25, +63% 9mo-FY26): the premium realized when loans are sold forward; the most volatile, mark-dependent, capital-markets-sensitive line. The lowest-quality revenue stream and the one most exposed to reversal if funding appetite softens.
• Servicing income (~4%): fees for servicing loans owned by third parties; the most recurring, capital-light, highest-quality line — but small.

Recurring vs. transactional. This is decidedly not a SaaS-like recurring-revenue model. Roughly two-thirds of revenue (interest income + gain-on-sale) is transactional and credit/capital-markets-exposed; network revenue recurs only insofar as GMV recurs; only servicing income (~4%) is contractually recurring. Affirm’s P&L is a consumer-lending statement with a network-fee overlay.

Key operating KPIs:

A worked example of the funding waterfall clarifies why this matters. When a consumer finances, say, a $1,000 Peloton purchase over 24 months, Affirm’s bank partner (Celtic or Lead) originates the loan; Affirm purchases it, collects a merchant fee up front (higher for a 0% or longer-dated loan), and then must fund the $1,000 of principal. It can (a) hold the loan on a warehouse facility and earn the interest spread (booking interest income, retaining credit risk, consuming equity capital), or (b) sell the loan forward to a securitization trust or a forward-flow buyer, booking a gain-on-sale up front plus an ongoing servicing fee (capital-light, but surrendering the future interest spread and depending on a buyer’s appetite). The same origination thus shows up in different revenue lines depending on a funding decision Affirm makes after the sale — which is precisely why GMV (the volume) is the cleaner growth signal than revenue (the funding-path-dependent monetization), and why RLTC (revenue net of the funding and credit costs of that decision) is the truest measure of platform gross profit. The mix between (a) and (b) is a discretionary lever management pulls toward gain-on-sale when funding markets are friendly — as they have been in FY26.

Verdict. Affirm is a credibly large, increasingly diversified commerce-and-credit platform with a clean consumer value proposition and a genuine engagement flywheel (transactions per consumer rising from 3.9 to 6.7 in three years). But the revenue architecture is that of a lender plus capital-markets desk, not a high-multiple software network — and an investor should resist the “network” framing the income statement does not support.

3. Industry Dynamics

Market structure and size. US BNPL payment volume was roughly $122B in 2025 (+12% YoY), forecast toward ~$184B by 2030 at ~8.5% CAGR (third-party databook estimates; treat as directional). Global BNPL GMV was ~$560B, North America ~56% of it. Affirm’s $36.7B FY25 GMV is therefore ~30% of the US pool — a clear top-two domestic position by volume, alongside Klarna and Block’s Afterpay.

The competitive set (FY25 10-K “Our Competition,” verified externally):

• Pure-play BNPL: Klarna (now public on the NYSE, ticker KLAR; ~111M users / ~790K merchants globally), Afterpay (owned by Block/XYZ), PayPal Pay-in-4 (>$40B TPV, +20%+ in 2025), Zip, Sezzle.
• Legacy card/credit issuers: Synchrony, JPMorgan Chase, Citi, Capital One, American Express, Bread Financial, plus the Visa/Mastercard rails themselves.
• Merchant proprietary plans presented in parallel to Affirm at checkout.

Affirm concedes its competitive exposure plainly in its own 10-K: “many of our competitors are substantially larger… [with the] ability to cross-subsidize… lower-cost funding… greater brand recognition.” This is the crux of the industry problem — Affirm competes for checkout placement against far larger balance sheets (Block’s Cash App ecosystem, PayPal’s installed wallet base, the card networks) that can subsidize distribution.

Regulatory landscape — net-favorable in 2025, but politically variable:

• The CFPB’s BNPL interpretive rule was rescinded. The May-2024 rule (which would have treated BNPL providers as TILA credit-card issuers under Reg Z, imposing dispute-rights, periodic-statement, and refund obligations) was deprioritized for enforcement in May 2025 and the CFPB stated it would not reissue it. A material compliance overhang lifted across the whole BNPL group — it helps Affirm, Klarna, and Afterpay roughly equally. A future administration could revive it (open question).
• Credit-bureau reporting began in 2025: Affirm reports Pay-in-4 to Experian (Apr-2025) and TransUnion (May-2025), and the FICO Score 10 Suite will incorporate BNPL data. Double-edged: it improves industry-wide underwriting and may legitimize BNPL in credit files, but it also democratizes repayment data (eroding Affirm’s proprietary edge) and could surface “loan stacking.”
• No late fees is a structural insulation from the card-issuer late-fee fights.
• Tail risks: APR-cap or rate-cap proposals (severe if enacted — they would compress RLTC directly), interchange/Durbin litigation, and state-AG activism. Management says it is “not hearing about” BNPL-specific rate caps, but this is a regulated activity overseen, in Affirm’s own phrasing, by “51 distinct entities.”

Capital-cycle read (Marathon lens). BNPL is mid-cycle and post-shakeout, but with capital and competition re-entering at the top. The 2021 mania (Klarna’s $45B private mark) corrected violently to $6.7B in 2022; the sector consolidated; and it has since re-rated — Klarna IPO’d on the NYSE in September 2025 at a $15.1B valuation, and Apple exited its in-house Apple Pay Later product. For incumbents this is the more attractive phase of the cycle (irrational capacity was wrung out), but the Klarna IPO, the proprietary BNPL pushes by card issuers, and Walmart’s poaching of share are exactly the signs of capital and rivalry returning. And Affirm’s own ~35% GMV growth is precisely the asset-growth that Marathon’s framework flags for return-mean-reversion scrutiny.

Why the product economics are structurally weaker than a credit card’s is worth making explicit, because it frames the whole industry. A revolving credit card earns a merchant interchange fee plus recurring interest on a perpetually-revolving balance plus, historically, late fees — and the relationship is durable because the card sits in the consumer’s wallet and the balance compounds. A BNPL loan earns a merchant fee plus (for interest-bearing products) finite simple interest on a self-amortizing balance that disappears in months, with no late fees and no automatic re-engagement — the consumer must affirmatively choose Affirm again at the next checkout. BNPL trades the card’s durability and compounding for better consumer economics and merchant conversion lift. That is a better product for the consumer and a structurally worse annuity for the lender — which is exactly why the Affirm Card (a re-engagement mechanism that pulls Affirm back into the wallet) is strategically central, and why the bull case leans so heavily on it.

Verdict: structurally mixed, leaning unattractive at the product level. Consumer installment credit is fundamentally a commodity with low switching costs, price-taking dynamics, funding-cost sensitivity to rates, and credit cyclicality. Barriers to entry are low for the product but meaningful for scale + data + capital-markets access. A few scaled players can earn acceptable returns; none enjoys real pricing power. This is an average-to-poor industry in which operational excellence and scale — not structure — separate the winners.

4. Competitive Position

Greenwald verdict: a narrow, contestable moat — economies of scale + capital-markets access + an underwriting-data intangible — not the durable two-sided network effect management claims. Real enough to drive improving unit economics at scale; not wide enough to prevent share loss to a larger, better-funded rival.

1. The claimed network effect — pressure-tested, and it largely fails. Affirm asserts “strong network effects” (FY25 10-K p.8). A genuine network effect requires that each additional consumer makes the platform more valuable to every other consumer (and vice versa for merchants). In BNPL this does not hold:

• Consumers face near-zero switching costs and multi-home — a shopper uses Affirm at one checkout, Klarna at the next, Afterpay at a third.
• Merchants integrate multiple BNPL providers simultaneously and present them side-by-side at checkout (Affirm’s own 10-K acknowledges this).
• The decisive disconfirming evidence: if a network effect were real, a top-five merchant could not be poached. Walmart was. In March 2025, Walmart’s majority-owned OnePay fintech selected Klarna to underwrite installment financing, displacing Affirm, which had held the account since 2019. The stock fell up to ~14% on the news.

This is a distribution and scale advantage, not Greenwald demand-side captivity. The “network” framing should be retired.

2. Economies of scale + capital-markets access — the strongest, genuinely real leg. Affirm’s $36.7B GMV gives it fixed-cost leverage over a large underwriting-and-technology platform (technology-and-data opex was roughly flat as a share of revenue while GMV grew 38%), and the financial proof of scale economics is RLTC/GMV rising from 3.34% (FY23) to 4.14% (9mo-FY26). The funding machine is the clearest barrier: $7.8B held off-balance-sheet, multiple warehouse and ABS programs, and ABS priced below 100bps of spread — terms a sub-scale entrant simply cannot access. A new BNPL competitor can copy the checkout button; it cannot replicate the funding curve.

3. The underwriting-data intangible — real but eroding at the margin. Ten-plus years of transaction-level, item-level data feed a continuously-learning risk model that benefits from scale; the financial proof is that Affirm has navigated multiple rate cycles while holding ~3–4% RLTC/GMV and building (not releasing) reserves. The erosion risk: FICO Score 10 plus bureau reporting will democratize some BNPL repayment data industry-wide, shrinking the proprietary edge over time (magnitude is an open question).

4. The 0%-APR merchant-subsidized model — a product advantage, not a barrier. Affirm earns higher merchant fees on 0% loans and aligns with merchants who want conversion lift, but Klarna and Afterpay offer the same structure. Differentiator, not moat.

5. Anchor integrations — high value, low durability, and the central risk. Amazon (~22% of GMV), Shopify, Apple Pay (iOS 18), and Peloton are powerful distribution; the flip side is that top-five merchants/partners are ~47% of GMV (up from ~42% in FY23). The Apple win (Apple discontinued its in-house Apple Pay Later in 2024 and routed financing to Affirm) partly offsets the Walmart loss — but both events confirm the same truth: these are renewable commercial contracts, the opposite of switching-cost lock-in.

Direct comparison. Versus Klarna: larger globally, now public, just won Walmart; Affirm is more US-centric and more weighted to higher-AOV interest-bearing loans (hence a higher revenue/GMV take rate), and arguably more disciplined on credit — but neither has a defensible moat against the other. Versus Afterpay (Block) and PayPal: both enjoy structurally lower-cost distribution through existing consumer ecosystems, which is exactly the cross-subsidization Affirm names as a threat.

A worked comparison with Klarna sharpens the point. Affirm and Klarna are close substitutes at the merchant level, but the unit economics differ in instructive ways. Affirm skews to higher-AOV, longer-dated, interest-bearing loans (hence its ~8.8% revenue/GMV take rate, well above the low-single-digit take of a pure Pay-in-4 book), which means more interest income, more balance-sheet capital at risk, and more credit-cycle exposure — but also more revenue per dollar of GMV. Klarna skews more to short-dated Pay-in-4 and a broader global merchant footprint. Neither structure confers a defensive advantage over the other: when Walmart’s OnePay ran a competitive process, it chose Klarna’s underwriting on terms, and Affirm could not invoke any switching cost to retain a six-year relationship. The lesson is that in BNPL, share is won and lost on price, approval rates, and funding cost every contract cycle — the defining property of a commodity-credit market, not a moated network. Affirm’s genuine edge over a sub-scale entrant (the funding curve, the loss-discipline data) is real; its edge over an equally-scaled, equally-funded rival like Klarna is execution, not structure.

Verdict: a good business, not a great one. Affirm’s edge is execution and scale, not a wide structural moat. The advantage is real enough to compound improving unit economics, but it does not prevent a larger rival from taking an anchor account — as one already did. If the “moat” can be tied to a financial outcome that would deteriorate without it, the honest version is: the funding-cost advantage and the data-driven loss discipline are the moat; the consumer “network” is not.

5. Growth History and Forward Opportunities

Historical trajectory. Affirm has compounded GMV at roughly a 45% CAGR (FY21 $8.3B → FY25 $36.7B) and revenue at ~39% ($870M → $3,224M), while active consumers grew ~34% (7.1M → 23.0M) and engagement deepened (transactions per consumer 3.9 → 5.8). The most recent quarter (Q3-FY26) showed GMV +35%, revenue +33%, and RLTC +~41% over nine months — decelerating modestly off a larger base but still 30%+, with RLTC outgrowing GMV, the highest-quality combination available.

Organic vs. partner-driven. Growth is a blend, heavily levered to a few anchors. FY25 GMV growth was attributed to “top-five merchants and platform partners… direct-to-consumer products, including Affirm Card, and… the active merchant base.” Amazon (~22%) and the Apple/Shopify integrations are partner-driven; the genuinely owned, organic engine is the Affirm Card.

Affirm Card — the key forward vector. Card metrics are the most important growth story because they break the anchor-concentration dependency: 4.4M active cardholders (+~121–130% YoY); card GMV +146%; ~20% of active consumers now hold a card; management’s medium-term target is ~7.5M cards / $10B GMV. Management calls it “by far the fastest-growing and most profitable product,” grown with no performance marketing (Affirm re-engages known users rather than buying new ones), and card users skew slightly higher credit quality, are “least lossy,” and transact ~40% more frequently. The card is the single strongest counter to the no-moat/concentration bear case because it is owned distribution that lifts transactions-per-consumer (5.8 → 6.7) and diversifies away from merchant contracts.

Other forward opportunities. International is the large optionality: Canada live via Shopify (Apr-2025), the UK in beta (FCA pre-authorization; Wayfair launch Feb-2026), with Australia and Western Europe planned — but these are early and immaterial to FY26, a multi-year call option rather than a current earnings driver (and a small near-term RLTC drag as new markets ramp). B2B financing (Intuit/QuickBooks), enabling banks and regionals to issue BNPL-enabled debit (via Fiserv and FIS — a “half-a-billion US debit cards” TAM, in management’s framing), processor “default payment method” placements, the Affirm Money Account, and a pending US industrial-bank (ILC) charter application round out the surface area. Management is explicit that the charter is about regulatory certainty, “years” away, with “no model impact near-term,” and that the newest experiments (rent payments, agentic commerce) should be put at “nothing in your model.”

Verdict: high-quality, profitable, compounding growth — the strongest part of the story. This is not GMV-at-any-cost. RLTC/GMV is rising, the company reached GAAP profitability, S&M dollars fell year-on-year even as volume grew, and engagement is compounding. The quality concern is durability and concentration (Amazon 22%, top-five 47%, Walmart already lost), not the economics of the growth itself.

6. Financial Quality

The multi-year P&L backbone ($000s; FY figures from 10-K MD&A; FY21–22 reconciled to filing presentation; 9-month from Q3-FY26 10-Q):

The inflection is real and accelerating. Affirm turned its first-ever full-year GAAP net profit in FY2025 (+$52.2M, a $570M swing from FY24), and its first full-year GAAP operating profit is arriving in FY26: nine-month operating income +$269.7M (versus −$145.3M a year earlier) and net income +$313.2M. Q3-FY26 alone delivered operating income +$88.4M, net income +$102.9M, and diluted EPS $0.30.

Decomposing the FY26 inflection (9mo-FY26 vs. 9mo-FY25; the +$415M operating-income swing): revenue grew +$747M (+31.8%) — interest income +$291M, network revenue +$249M, gain-on-sale +$167M (+63%), servicing +$40M — while total operating expense grew only +$332M (+13.3%). The standouts: sales-and-marketing fell −$105M (−29.7%) to $250M, and funding costs grew only +$16M (+5.0%) against a $1.5B larger loan book, confirming per-unit funding-cost compression (~125bps year-on-year, per management). Provision for credit losses grew +$113M (+24.7%) with the book — building reserves, not releasing them.

An honest caveat on the “operating leverage.” A meaningful chunk of the S&M decline is the roll-off of non-cash commercial-agreement warrant expense (the Amazon/Shopify partner warrants), which fell from $230.1M to $142.7M over the nine months — an ~$87M tailwind that is mechanical decay of a non-operational charge, not pure efficiency. The underlying efficiency gain is real but smaller than the headline.

Unit economics and take rate. Revenue take rate (revenue/GMV) is a stable-to-rising ~8.8% in FY25. The cleaner gross-margin metric is RLTC = revenue less transaction costs (transaction costs = loss on loan-purchase commitment + provision + funding costs + processing/servicing): ~$1.48B FY25 (4.04% of GMV), rising to ~$1.50B and 4.14% of GMV in 9mo-FY26. RLTC grew faster than GMV — the single most important quality signal in the whole report. Affirm’s headline non-GAAP profit metric, Adjusted Operating Income, was $778M in FY25 (~24% of revenue).

Balance sheet and funding (3/31/2026 vs. 6/30/2025, $000s):

Liquidity is solid ($1.72B cash + $0.76B AFS securities), and equity grew +$714M over nine months (net income plus option/RSU settlement into paid-in capital). The structural point: equity is only ~4% of the total platform portfolio — an extremely thin capital base for a consumer-credit book, and the reason Affirm is structurally dependent on third-party funding appetite. The ~$7.3B of funding debt is collateralized by the loan book, not corporate leverage in the conventional sense.

Credit performance — the heart of the matter. The allowance rate rose to 5.98% (3/31/26) from 5.65% (6/30/25). Management attributes the increase to seasonality and elevated tax-season prepayments (good loans paying off early, leaving the remaining book higher-loss on a smaller base). The critical read is more favorable than the excuse implies: provisioning ($573M over nine months) ran ahead of net charge-offs ($441M) — Affirm is building reserves, not releasing them, which means the GAAP profit is not flattered by reserve releases. On-book 30+ day delinquencies are ~$236M / ~2.74% of amortized cost, with non-accrual UPB just $6.0M. That is a genuinely positive quality-of-earnings signal.

But there is a structural blind spot. Affirm sells a large share of originations to forward-flow buyers and securitizations before those loans would charge off (over the nine months it sold $14.1B of held-for-investment loans plus $2.1B held-for-sale). Selling performing and early-stage loans removes their eventual losses from the reported on-book charge-off and delinquency series. The blended, platform-wide net loss rate (on-book + sold) is not cleanly disclosed — so the on-balance-sheet credit metrics likely understate the platform’s true gross loss experience. The reported delinquency trend is a real but incomplete read on underwriting quality.

A through-cycle credit sensitivity makes the central risk concrete. RLTC is revenue less transaction costs, and provision for credit losses is the largest swing component of those transaction costs. In FY26, RLTC runs ~4.1% of GMV with provisioning that tracks net charge-offs in the ~5–6% range of the on-book balance. The historical playbook for a consumer-credit downturn is a 50–100%+ surge in loss rates over 12–18 months. A rough illustration: if blended net losses on the funded book rose by ~200bps of the loan balance in a recession (well within the range of past consumer-credit cycles), and Affirm could not fully reprice or tighten quickly enough to offset, RLTC/GMV could compress from ~4.1% toward ~3.0–3.4% — a ~20–30% hit to the gross-profit margin on top of any GMV slowdown. Simultaneously, gain-on-sale (~14% of revenue) would compress as forward-flow buyers demand wider spreads or pull back, and funding costs on the warehouse book would rise. This is the correlated triple-hit, and it is why a ~4%-equity balance sheet that has never run a recession as a public company warrants a wider margin of safety than a stable-margin software business. The point is not that this will happen on any timetable — it is that the FY26 margin embeds a benign-credit assumption the price treats as structural.

ROE/ROIC: not a meaningful quality signal here. With equity at ~4% of the portfolio, any reported return-on-equity reflects extreme leverage on a loan book, not franchise economics. The right lenses are RLTC/GMV, the through-cycle net charge-off rate, and RLTC-to-net-income conversion — not ROE.

Verdict: economics genuinely improve with scale — but the quality is thinner than the headline and untested by a downturn. RLTC outgrowing GMV, reserves building ahead of charge-offs, and falling funding costs are real, high-quality signals. They are offset by a ~14%-of-revenue, +63% gain-on-sale line that is mark-and-timing-sensitive; SBC and rolling-off warrant amortization that flatter the margin trajectory; and a credit series that excludes the sold book. The economics are improving and real; the durability is the open question.

7. Capital Allocation

Frame it correctly: Affirm has no traditional free cash flow. It is a balance-sheet lender that consumes capital to grow its loan book (nine-month FY26 loan originations/purchases ran ~$33B, funded through financing and investing flows). Aggregator “free cash flow” figures (ROIC shows ~$1.3B TTM) are misleading because they do not net the cash absorbed by loan-book growth. Capital allocation here means funding strategy, dilution management, and M&A — not buybacks or dividends.

No dividend; no genuine buyback. The FY25 10-K confirms no equity repurchases. The one “repurchase” (a $250M purchase of 3.5M shares in December 2024) was a convertible-arbitrage / anti-dilution mechanic executed alongside the 2029 notes offering — bought from holders of the 2026 convertibles — not a value-driven buyback.

Convertible-note management is the real activity. Affirm issued 2029 notes (0.75% coupon, $920M, conversion price $101.02) in late 2024 and used the proceeds plus cash to repurchase ~$1.1B of its 0%-coupon 2026 notes for ~$1.0B (booking an $82.4M extinguishment gain in FY24, with smaller amounts since). Swapping 0% paper for 0.75% paper is a modest increase in cash interest cost, undertaken to push out the maturity wall and reset the conversion strike far higher ($101 vs. the old ~$23) given the stock’s recovery — sensible liability management. About $249M of 2026 notes still mature in November 2026.

M&A is in the past and a non-event today. Legacy deals — PayBright (Canada, ~$264M, FY21), Returnly (~$300M, FY21), and the smaller Butter (FY22) — left goodwill flat at ~$534M with no impairments through the 2022–23 fintech bust (a mild positive). Cash spent on acquisitions has been negligible since FY22. M&A neither compounded nor destroyed value; it is not part of the current thesis.

The most important capital-allocation insight: distribution was bought with the share count. Affirm granted equity warrants to its anchor partners (Amazon, Shopify) to win and retain distribution. The P&L cost is the commercial-agreement warrant expense ($230M → $143M over the nine months, declining as the commercial-agreement asset runs off to $41M from ~$105M two years earlier). In substance, Affirm paid for its largest distribution channels in dilutive equity — a real, quantifiable cost of customer acquisition that also flattered the S&M optics as it amortizes off. Combined with $500M of annual stock-based compensation, the share count is the place where the cost of growth actually lands.

Funding choice as allocation. The lever between retaining loans (interest income, capital-intensive) and selling forward (gain-on-sale, capital-light) is the core allocation decision. Management has tilted toward more gain-on-sale (+63%) in a favorable funding market — capital-efficient, but revenue-quality-diluting, and dependent on third-party appetite persisting.

Verdict: competent liability management, no value-destructive M&A, and no buyback to misjudge — but the alignment is imperfect. Capital allocation is sound on the debt side and clean on M&A. The blemish is that growth has been financed substantially with dilutive equity (warrants + SBC), and management’s incentives (below) reward the volume and non-GAAP profit that drive that dilution without a direct per-share or GAAP-earnings consequence.

8. Changes and Headwinds — Last Two Years

The GAAP-profitability inflection (FY25→FY26) — covered above — is the dominant positive change: a structural shift from a cash-burning growth story to a profitable, scaling platform, accompanied by falling funding costs and tightening ABS spreads.

The Walmart loss (Mar-2025) is the dominant negative. Walmart’s OnePay fintech selected Klarna for installment financing, displacing Affirm after six years. This is almost certainly the “top-three merchant we sunset in Q1” that management cited as a multi-point GMV growth headwind into FY26 (without naming it on the calls). It is the single clearest evidence that anchor-merchant relationships are contestable, and a recurring risk, not a one-off.

The Apple win (2024) partly offsets. Apple discontinued its in-house Apple Pay Later and partnered with Affirm; from iOS 18 (Sept-2024), US Apple Pay users can finance through Affirm — a major distribution channel, though warrant-linked like the others.

Other material developments:

• Klarna IPO (Sept-2025, NYSE, $15.1B): a better-capitalized, now-public direct rival, and a re-rating event for the BNPL group.
• CFPB BNPL interpretive rule rescinded (2025): a regulatory overhang lifted (a tailwind), though revivable under a future administration.
• Credit-bureau reporting launched (2025): Affirm began reporting Pay-in-4 to Experian and TransUnion; FICO Score 10 will incorporate BNPL — industry-legitimizing but edge-eroding.
• Affirm Card scaling (+146% GMV) and international launches (Canada live, UK beta).
• Governance/board: director Keith Rabois resigned (eff. 6/30/25); the company reincorporated from Delaware to Nevada (a management-friendlier jurisdiction on takeover/liability).
• Compensation reset: a new January-2026 PSU grant to CEO Max Levchin (333,667 units, vesting on FY26–28 RLTC growth and Adjusted-Operating-Income growth) as the FY21 IPO Value Creation Award lapses largely unearned (below).
• Pending ILC industrial-bank charter application (regulatory-certainty play, years out, no near-term model impact).

Verdict: the changes net positive on fundamentals (the inflection, funding de-risking, the regulatory tailwind, the Apple win) but introduce two durable structural concerns — proof that anchors are poachable (Walmart) and a better-capitalized public Klarna. The thesis is stronger on profitability and weaker on competitive durability than it was two years ago.

9. Risk Analysis

Catastrophic-loss assessment. A total loss is unlikely in the near term given $1.7B cash, a thinly-but-positively-capitalized balance sheet, and access to deep funding markets — but the realistic severe scenario is a correlated triple-hit: a consumer-credit downturn simultaneously (a) raises charge-offs and provisions, compressing RLTC; (b) widens ABS/forward-flow spreads, raising funding cost and shrinking gain-on-sale; and © slows GMV. With equity at ~4% of the portfolio and the business never having been public through a recession, the left tail is genuinely fat — the realized −95% (2022) and −54% (2025–26) drawdowns are evidence, not theory.

10. Valuation

No price target and no recommendation in this section — embedded expectations and scenarios only.

Rebuild the enterprise value at the live price, and net funding debt against the loan book. This is the single most important valuation mechanic for Affirm. At $73.92, ~335M basic shares give a market cap of ~$24.8B (~$25.7B fully diluted on ~348M shares). Of the ~$9.1B of total debt, the large majority (~$6.8B warehouse + securitization) is funding debt secured by — and over-collateralized by — the $8.06B loan book; it should be netted against the loans exactly as one would not add a bank’s deposits to its EV. Only the ~$2.3B of convertibles is true corporate debt. The naïve gross EV (~$31.4B, and any EV/EBITDA built on it — ROIC’s ~21× would be ~28–30× at the live price) is the wrong number; it double-counts matched funding. The right figure: with funding debt offset by loans (net −$1.3B) and convertibles minus cash roughly netting, corporate EV ≈ market cap ≈ ~$24.6B. EV/EBITDA is not the right lens for Affirm.

Use the lender-and-network lenses instead:

Embedded-expectations / reverse-DCF. Projecting GMV → RLTC → normalized net income over seven years, with an exit P/E and an ~11% discount rate:

At ~$24.8B the market sits between bear and base, modestly below base. To justify today’s price you must believe Affirm (1) compounds GMV ~18–20%+ for most of a decade (notably, the market is not fully extrapolating the current 35% — it implicitly fades it toward ~20%, which is reasonable); (2) holds or expands an ~4.1% RLTC margin through a full credit cycle (the genuinely unproven part); and (3) converts a rising share of RLTC into GAAP net income as SBC and provisioning leverage. The risk is not that 20% GMV growth is too optimistic — it is that the 4.1% RLTC margin and benign credit have never been stress-tested in a downturn. The market is underwriting “the inflection is real and durable,” which is partly proven (FY26 GAAP profitability is a fact) and partly an extrapolation of benign credit (an assumption).

Comp context. Versus the de-rated peers on file, Affirm trades at a clear premium on gross-profit/earnings: PayPal at ~7.7× earnings and Block at ~3× EV/gross-profit are the value end of BNPL; SOFI’s growth-adjusted multiple (~26–28× forward P/E, PEG ~0.66 on ~9% ROTCE) is closer. Affirm is the highest-quality-network and highest-beta name in the set, priced accordingly. Klarna (KLAR, ~$15.1B IPO cap) is the direct global read-across; Sezzle the speculative small-cap end.

A sensitivity on the load-bearing variable. Because RLTC margin is the assumption the price most depends on, it is worth isolating. Holding base-case 20% GMV CAGR and a 22× exit multiple, varying only the terminal RLTC/GMV and the RLTC-to-net-income conversion moves the equity value materially: at a recession-impaired ~3.5% RLTC/GMV and 35% conversion, the model lands near the bear value (~$10–14B, roughly $30–42/share-equivalent); at the base ~4.1% and 42% conversion it lands at ~$32B (~$73.92 area, today’s price implied); at a sustained ~4.4% and 50% conversion it approaches the bull (~$60B+). In other words, the gap between “the stock is expensive” and “the stock is cheap” is almost entirely a function of whether the RLTC margin holds through a cycle — not the GMV growth rate, which the market already sensibly fades from 35% toward 20%. That single variable, untestable until the next downturn, is the whole debate.

Percentile caveat (mandatory). Affirm’s own-history valuation percentiles span only ~2021–2026 — entirely post-IPO, dominated by a 2021 bubble peak and a 2022–23 near-zero trough, with the denominator a loss for most of that window and no full credit cycle. The “34th-percentile P/E” reads cheap but is close to meaningless. Treat all own-history percentiles here as low-signal: Affirm has never traded through a consumer-credit recession as a public company.

11. Variant Perception

Consensus. The market (corroborated by a sell-side read that Affirm, Klarna, and SoFi are priced for a profitability inflection / positive cash-flow return on equity from 2026) believes Affirm has graduated from a cash-burning growth story to a structurally profitable fintech network — GMV compounding ~20–35%, RLTC durable near 4%, GAAP profitability now real and scaling, funding de-risked. The Walmart loss is “shrugged off”; Apple Pay and the Affirm Card (+146%) are the next legs. Affirm is the believed name in a cohort of given-up-on peers.

The strongest bull case. A genuinely scaling, now-profitable platform (26.8M consumers, 6.7 transactions each) with real structural advantages — underwriting-data depth, deep merchant integrations, and a funding machine pricing ABS below 100bps that a sub-scale rival cannot match. GAAP profitability is a fact, not a promise, and RLTC/GMV is rising. If GMV compounds 20%+ and credit stays benign, the base/bull range ($32–86B) is in play and today’s ~$25B is cheap. The Affirm Card breaks the concentration dependency with owned, no-marketing-cost distribution.

The strongest bear case. Affirm is a balance-sheet consumer lender in a network’s costume that has never been public through a recession. The 4.1% RLTC margin and benign charge-offs are a function of a strong-jobs, low-default environment; a credit cycle compresses RLTC and widens funding spreads and slows GMV at once — a correlated triple-hit on a ~4%-equity book. Merchant concentration (Amazon ~22%, top-five ~47%) is an existential single-customer risk, and Walmart already left. Competition is intensifying (Klarna public and capitalized, Block/Afterpay, PayPal, Apple itself, card issuers) against no durable consumer-side moat. The 2.7 beta and realized −95%/−54% drawdowns prove the equity is a regime instrument, not a compounder. And the reported on-book credit metrics flatter the truth because the sold book’s later losses sit off-statement.

The 3–5 assumptions that matter most:

1. Credit performance through a cycle — does RLTC/GMV hold ~4% when charge-offs spike? (Unproven; this is the whole thesis.)
2. Funding-market access — do ABS/forward-flow spreads stay tight in stress? (Cyclical, and correlated with #1.)
3. GMV durability ~20%+ — without Walmart, against Klarna/Apple/Block. (Partly proven via Affirm Card and Apple Pay.)
4. Merchant concentration — does Amazon re-bid favorably, or follow Walmart out? (Binary; low-probability/high-impact.)
5. RLTC→GAAP conversion — does SBC/opex leverage continue, or does dilution eat the inflection?

Falsification tests. The bull case is falsified if RLTC/GMV compresses below ~3.7% for two-plus quarters, or delinquencies/charge-offs break above prior-cycle norms, or a second top-five merchant defects, or GMV growth decelerates below ~15%. The bear case is falsified if Affirm posts a full year of GAAP profitability with a stable/rising RLTC margin and benign credit through a measurable consumer-credit softening — the first real-world stress test passed — while funding spreads hold.

Factor-positioning read (the tape). Affirm carries a market beta of ~2.7–2.8, a strongly anti-low-volatility loading (−1.56), and — tellingly — a negative momentum loading (−0.66) despite the +74% bounce off the March-2026 low, because the trailing 12-month window still contains the −54% drawdown. Its factor-similar cohort is the ARK-innovation / high-beta-consumer-cyclical basket (Block, Sea, Carvana, the ARK ETFs, the cruise lines, Zillow), and ~half its variance is explained by these systematic factors (R² 0.54). This is therefore not a crowded momentum long (momentum is out of favor on the longer look-back) and not a falling knife (it is profitable and rallying) — it is a high-beta, anti-low-vol, risk-on recovery name: a levered bet on benign credit and a persisting risk-on regime, with a proven fat left tail. The factor profile is evidence that consensus is offsides in both directions episodically — the stock overshoots up in risk-on and down in risk-off — which is exactly why an entry discount, not the spot price, is what a disciplined buyer should demand.

12. Fact vs. Interpretation Table

13. Open Questions

1. Blended platform-wide net loss rate (on-book + sold). The sold portfolio’s later losses are not in the on-book delinquency/charge-off series; the true gross loss rate is not cleanly disclosed. This is the single most important number that is missing.
2. Gain-on-sale sustainability. How much of the +63% reflects a favorable spread/forward-flow environment that normalizes? What happens to this ~14%-of-revenue line if securitization spreads widen?
3. Discount-rate sensitivity of retained-interest and gain-on-sale marks in a rate-shock — not quantified in the materials reviewed.
4. Amazon contract economics and renewal (warrant terms, exclusivity, expiry) — Amazon is ~22% of GMV; the durability of this single relationship is under-disclosed.
5. ILC bank-charter status and timeline — filed, “years” away, approval uncertain; the strategic and funding-cost implications if granted are material but unmodeled.
6. Fate of the six unearned Value Creation Award tranches at the FY26 lapse, and whether the new PSU design fully replaces the lost incentive.
7. International unit economics (UK/Canada) as they scale — RLTC drag magnitude and path to contribution.

14. What Must Be True

For the bull case (the inflection is durable and the platform compounds):

• GMV compounds ~18–20%+ for the better part of a decade, with the Affirm Card and Apple Pay offsetting anchor concentration and the Walmart loss.
• RLTC/GMV holds or expands near ~4.1% through a measurable consumer-credit softening, while funding spreads stay tight.
• RLTC converts into a rising GAAP net-income margin as SBC and provisioning leverage, without share count eroding the per-share result.
• Falsification test: the bull case breaks if, in the next consumer-credit downturn, RLTC/GMV compresses below ~3.7% for two-plus quarters, or a second top-five merchant defects, or GMV growth falls below ~15% — any one would show the model is more cyclical and less defensible than the price assumes.

For the bear case (a contestable, cyclical lender priced as a network):

• The next downturn delivers the correlated triple-hit (charge-offs up, funding spreads wider, GMV slower), compressing RLTC and turning the thin-equity book into a capital constraint.
• Competition (Klarna, Block, PayPal, Apple, card issuers) compresses take rates and steals anchor accounts, and the on-book credit series proves to have flattered the true loss rate.
• Falsification test: the bear case breaks if Affirm posts a full year of GAAP profitability with a stable/rising RLTC margin and benign credit through an actual consumer-credit softening — the first real cycle test, passed — at which point the “never-tested” discount this report demands is no longer warranted.

The two falsification tests share one clock: the first real consumer-credit downturn Affirm experiences as a public company. Everything else is prologue.

15. Source Appendix

See the accompanying AFRM_source_appendix.md (Appendix B in the combined report) for the full source list. Primary sources include: Affirm Holdings FY2021–FY2025 Forms 10-K (CIK 0001820953; FY ends June 30); the Q1–Q3 FY2026 Forms 10-Q (latest afrm-20260331, filed 2026-05-07); Forms 8-K (Walmart/OnePay 2025-03-17, Shopify global agreement Feb-2025, board/comp changes, the 2026-01-13 Levchin PSU grant); the DEF 14A (filed 2025-10-24); the Form 4 corpus; Affirm earnings-call transcripts (Q2-FY26 and Q3-FY26); third-party BNPL market data; and public same-sector peer data (SOFI, Block/XYZ, PayPal, Capital One) used for cross-read and comp context. Quantitative figures cross-checked against public market-data and factor-model sources, each reconciled to the underlying filing.

APPENDIX A — Standard Diligence Questionnaire — Affirm Holdings, Inc. (NASDAQ: AFRM)

Supplemental diligence questionnaire. Fact / Interpretation / Assumption labels applied where they matter. As-of 2026-06-19.

General

What thoughtful questions have other investors asked about this company? The recurring institutional debates: (1) Is the GAAP profitability real or engineered? — answered largely yes, it survives full SBC expensing in FY26, but the trajectory is flattered by rolling-off partner-warrant amortization and a rising SBC-capitalization rate (Interpretation). (2) What is the true through-cycle loss rate? — the central unanswered question, because the on-book delinquency series excludes the large sold/forward-flow book (Open Question). (3) Is the “network effect” a real moat? — pressure-tested and largely failed; Walmart’s defection to Klarna is the disconfirming fact (Interpretation). (4) Can RLTC/GMV hold ~4% as rates and competition normalize? — the load-bearing assumption (Assumption). (5) How dependent is the model on benign capital markets? — very: equity is ~4% of the platform portfolio and ~$7.8B of loans sit off-balance-sheet (Fact).

Cyclicality & Earnings Nature

Are earnings at a cyclical high or low? Interpretation: at an early-cycle high for credit quality (benign consumer, low defaults, tight funding spreads), but at the low end of the company’s eventual margin-maturity curve (operating leverage still ramping). The danger is that the credit tailwind and the margin ramp are being conflated — the former is cyclical and will reverse.

Driven by external environment or internal actions? Both. Internal: genuine operating leverage, S&M discipline, funding-cost execution (~125bps YoY improvement), Affirm Card flywheel. External: tightening ABS spreads (<100bps, first time since 2021), benign consumer credit, falling benchmark rates. Roughly half the FY26 margin improvement is durable internal efficiency; the rest is a friendly external funding/credit environment (Interpretation).

How stable are revenues? Low stability. ~50% interest income (book-dependent), ~14% gain-on-sale (mark/timing/spread-dependent and lumpy), ~35% network revenue (GMV-dependent), only ~4% truly recurring servicing. This is a transactional, cyclically-geared revenue base, not a contracted one (Fact).

Outlook for products/services; how big is the market? US BNPL ~$122B (2025) → ~$184B (2030E), ~8.5% CAGR; global ~$560B. Affirm has ~30% US share and is expanding into the UK and Canada. The market is growing, domestic-led with international optionality, but competitive and commoditizing (Fact/Interpretation).

Business Quality & Competitive Moat

Is the industry getting more or less competitive? More. Klarna is now public and capitalized (NYSE, Sept-2025, $15.1B); Block/Afterpay, PayPal, Apple, and card issuers all compete; merchants multi-home. Capital is re-entering the space (Marathon late-cycle signal) (Interpretation).

How profitable is the business (ROIC/ROE)? ROE/ROIC are not meaningful — equity is ~4% of the portfolio, so reported returns reflect leverage on a loan book, not franchise economics. The honest profitability metrics are RLTC/GMV (~4.14%, rising) and Adjusted Operating Income (~24% of revenue; $778M FY25), plus newly-positive GAAP net income (Fact).

How profitable is the industry; barriers to entry? Industry profitability is thin and competed; barriers are low for the product (any wallet can add a BNPL button) but meaningful for scale + data + capital-markets access. Affirm’s funding curve and underwriting data are the real barriers; the consumer relationship is not (Interpretation).

Can the business be easily understood? Moderately. The checkout product is simple; the funding model (warehouse + securitization + forward-flow sales, on- vs. off-balance-sheet, gain-on-sale accounting, fair-value marks) is genuinely complex and is where the QoE risks hide (Interpretation).

Can it be undermined by foreign low-cost labor? No — not a labor-cost-exposed business. The analogous threat is capital arbitrage (a better-funded rival underpricing), which is live (Klarna).

Do brands matter? Yes, at the margin. Affirm’s “$0 late fees / simple interest / no consumer-misfortune” brand is a genuine differentiator versus revolving credit and a regulatory-goodwill asset. But it does not lock in consumers or merchants (Interpretation).

Nature of competition; switching costs? Competition is on checkout placement, merchant terms, approval rates, and funding cost. Consumer switching costs are near zero (multi-homing); merchant switching costs are contractual, not structural (Walmart left). The Affirm Card raises consumer engagement but not lock-in (Fact/Interpretation).

Financial Condition & Balance Sheet

Assets not fully recognized on the balance sheet? The underwriting-data intangible and partner-distribution relationships are not capitalized at economic value. Off-balance-sheet: ~$7.8B of loans held by third-party investors/securitizations (with only ~$8.7M recourse repurchase liability) (Fact).

Off-balance-sheet liabilities? The forward-flow/securitization structures carry limited recourse (~$8.7M) but create a funding-access dependency rather than a contingent liability. Convertible notes (~$1.13B net) are on-balance-sheet (Fact).

How conservative is the accounting? Mixed. Conservative on reserves (provisioning runs ahead of charge-offs — reserves building, not releasing) and on holding loans-held-for-investment at amortized cost. Less conservative where it matters most: gain-on-sale recognizes future value up front and is mark-dependent, and the on-book credit series excludes the sold portfolio (Interpretation).

How CapEx-hungry? Light on physical CapEx; capital-hungry in a different sense — it consumes balance-sheet capital to grow the loan book (~$33B of originations funded over nine months). Notable: ~36% of SBC is capitalized into internally-developed software, pulling expense off the P&L (Fact).

Capital Allocation & Management

How much FCF does the business generate; how is it used? No traditional FCF — the business funds a loan book. Aggregator “FCF” figures are misleading. Capital “allocation” = funding strategy, dilution management, and (historically) M&A (Interpretation).

Significant recent acquisitions? None recently. Legacy: PayBright (~$264M), Returnly (~$300M), Butter — all FY21–22, goodwill flat at ~$534M, no impairments (Fact).

Buying back shares? No genuine buyback. The $250M December-2024 “repurchase” was a convertible-anti-dilution mechanic, not a value buyback (Fact).

Issuing large amounts of stock to insiders? Yes — meaningfully. ~$500M annual SBC (~15.5% of revenue, 36% capitalized) plus dilutive Amazon/Shopify partner warrants. Distribution was substantially bought with the share count (Fact/Interpretation).

Compensation policy of directors/management? CEO Max Levchin takes ~$0 cash (FY25 base ~$39K, no bonus); economics are all equity (founder shares + the FY21 Value Creation Award, of which only 4 of 10 price-hurdle tranches were earned before it lapses in FY26). NEO cash-incentive metrics: GMV 50% / Revenue 25% / Adjusted Operating Income 25%; the new Jan-2026 Levchin PSU: RLTC 50% / Adjusted Operating Income 50%. Alignment gap: comp is indexed to growth and non-GAAP profit, not GAAP earnings, FCF, or per-share value (Fact/Interpretation).

Motivations of management? Founder-led, mission-framed (“no late fees, don’t screw the customer”), with Levchin’s wealth tied to the equity. The offsetting concern is that the incentive metrics reward GMV and non-GAAP profit — the very levers that drive the lower-quality gain-on-sale revenue and the dilution (Interpretation).

Valuation & Market Data

ADR / MLP / K-1? No — a US C-corp common stock, NASDAQ-listed, reincorporated in Nevada (from Delaware) in 2026. No K-1 (Fact).

Dividend policy? None; no dividend, retains all capital to fund growth (Fact).

How profitable is the business? Newly GAAP-profitable (FY25 +$52M, 9mo-FY26 +$313M); ~24%-of-revenue Adjusted Operating Income; ~4.14% RLTC/GMV. ROE/ROIC not meaningful (Fact).

Is net income diverging from cash from operations? OCF (~$935M, 9mo) >> net income (~$313M) — structurally normal for a lender (provisions, SBC, warrant expense, D&A are non-cash add-backs). But “operating cash flow” overstates discretionary cash because loan originations are funded through investing/financing flows; the business consumed ~$1.36B of net investing cash to grow the book (Fact/Interpretation).

Risks & Downside

What factors would cause the stock to decline? A consumer-credit downturn (charge-offs up, RLTC compressed), funding-spread widening, loss of an anchor merchant (Amazon), an APR/rate-cap regulation, a high-beta risk-off regime, or simple multiple compression from ~12× P/RLTC / ~65× trailing P/E (Interpretation).

Risk of a catastrophic loss? The realistic severe scenario is a correlated triple-hit (credit + funding + GMV) on a ~4%-equity book — the realized −95% (2022) and −54% (2025–26) drawdowns show the left tail is fat (Fact/Interpretation).

Chance of a total loss? Low near-term — $1.7B cash, positive equity, deep funding access. But the business has never run a recession as a public company, so the tail is genuinely uncertain (Interpretation).

Recent News & Events

Has the business environment changed recently? Yes: GAAP-profitability inflection (FY25→FY26); funding de-risking (ABS <100bps); Walmart lost to Klarna (Mar-2025); Apple Pay win (iOS 18, 2024); Klarna IPO (Sept-2025); CFPB BNPL rule rescinded (2025); credit-bureau reporting began (2025) (Fact).

Significant acquisitions? None recent (Fact).

Change in accounting policies? No material change; continued amortized-cost LHFI + fair-value retained interests + gain-on-sale recognition (Fact).

Recent changes — new markets, facilities, management? UK (beta) and Canada launches; pending ILC bank-charter application; director Keith Rabois resigned (6/30/25); Nevada reincorporation; new Levchin PSU grant (Jan-2026) (Fact).

APPENDIX B — Source Appendix — Affirm Holdings, Inc. (NASDAQ: AFRM)

As-of 2026-06-19. Primary sources prioritized over secondary. CIK 0001820953; fiscal year ends June 30.

Primary — SEC filings (Affirm Holdings, Inc.; SEC EDGAR)

Primary — Earnings-call transcripts (public earnings-call transcripts)

Secondary — market, industry, regulatory

• US/global BNPL market size and growth estimates (third-party databook; ~$122B US 2025 → ~$184B 2030E; ~$560B global).
• Klarna NYSE IPO (KLAR), September 2025 (~$15.1B valuation); Apple discontinuation of Apple Pay Later and Affirm partnership (iOS 18, 2024).
• CFPB BNPL interpretive-rule rescission (2025); credit-card late-fee rule vacated (April 2025); credit-bureau reporting (Experian Apr-2025, TransUnion May-2025) and FICO Score 10 BNPL inclusion (2025).
• Walmart/OnePay–Klarna selection (March 2025), trade and financial press.
• UBS HOLT note on payments/fintech-lender market-implied profitability expectations (May 2026).

Quantitative data services (third-party; reconciled to filings — filing governs)

Prior same-sector reports (internal cross-read context)

• SOFI (2026-06-11) — fintech lender; valuation method, ROTCE test, mark-to-model QoE.
• Block / XYZ (2026-06-14) — owns Afterpay; BNPL industry framing, EV/gross-profit method, un-seasoned-credit risk.
• PayPal (2026-06-12) — Pay-in-4 BNPL; capital-cycle take-rate compression.
• Capital One (2026-06-12) — consumer-credit-cycle framing, NCO normalization, CFPB/late-fee regulatory map.

Note on method: every non-obvious quantitative figure in the memo traces to an SEC filing or the company’s reported KPIs; third-party data services were used as cross-checks and reconciled to the filing, which governs in any discrepancy.