Affirm Holdings AFRM

Reference table

1. What they actually sell — and a real disclosure gap

The revenue disclosure is partial, and this needs stating clearly because it limits what can be analysed.

Two-thirds of revenue is not broken out. The undisclosed portion comprises interest income on the loan book, gains on loan sales, and servicing income — the components that carry the credit risk. So the disclosure covers the fee-based two-sided-network revenue and omits the lending revenue. For a company whose principal risk is credit, that is precisely the wrong two-thirds to be missing, and every conclusion in this dive is drawn with that limitation in view.

What can be said from what is disclosed: Merchant Network revenue grew 30.8% and Virtual Card Network 52.8% — the fee-based, capital-light components are growing at least as fast as the whole, which is a positive mix signal. Virtual Card Network revenue has grown 4.6-fold since FY21.

Geography (FY25):

96.3% United States. Affirm is a single-country bet on the American consumer, and the international business — which is growing more slowly than the domestic one — is not yet a diversifier. This is a concentration risk that does not appear in any ratio and it deserves prominence.

2. The turnaround — quarter by quarter

The annual arc:

FY25 was the crossover year. From a $985.3M loss in FY23 to a $52.2M profit two years later, while revenue doubled.

The quarterly record — where the operating leverage is visible:

Read the first and last rows together. Seven quarters ago Affirm lost 19.0 cents on every revenue dollar. In the most recent quarter it earned 19.5 cents. That is a 38.5-percentage-point swing in operating margin in under two years, achieved while revenue grew 48.7% over the same span.

Nine-month fiscal 2026: revenue $3.095B (+31.8% year over year), net income $313.2M, diluted EPS $0.90. Adding the $0.3486 Q4 consensus gives approximately $1.25 for the full year, matching the consensus FY26 estimate of $1.251 almost exactly — a useful confirmation that the estimate series here is on a comparable basis to reported GAAP results, which is not true of several other files in this batch.

And every one of the last six quarters beat consensus, the most recent by 76%.

3. Balance sheet and the lender-accounting warning

This is the section where the standard ratios must be discarded, and we set that out explicitly.

Balance sheet (FY25, 2025-06-30):

• Cash $1.354B + short-term investments $871.4M = $2.226B
• Total current assets $9.759B — overwhelmingly the consumer loan book
• Total debt $7.853B (essentially all long-term: warehouse facilities, securitisations and convertible notes)
• Net debt $5.627B (my computation; the vendor's netDebt of $6.499B nets cash only and excludes $871.4M of short-term investments)
• Shareholders' equity $3.069B; accumulated deficit -$3.057B; goodwill $534.2M
• Total assets $11.155B

Three vendor ratios that are meaningless here and would flag falsely on any screen:

• currentRatioTTM = 717.0. This absurd figure arises because total current liabilities are reported as $180.1M against $9.759B of current assets — the funding debt is classified as long-term while the receivables it funds are current. This is a classification artefact, not 717 times liquidity coverage. Discard it.
• interestCoverageRatioTTM = 1.0003. For a lender, interest expense is a cost of goods sold, not a financing burden — Affirm borrows in order to lend. A coverage ratio near 1.0 here indicates nothing about solvency. Discard it.
• netDebtToEBITDATTM = 5.77. This treats $7.85B of funding debt as corporate leverage. It is not. The debt funds a $9.76B earning asset. The correct framing is a loan-to-asset relationship, not a debt-to-cash-flow one.

The right way to read the balance sheet: Affirm borrows roughly $7.85B and lends roughly $9.76B, earning a spread, and holds $3.07B of equity against it — approximately 3.6 times assets to equity. That is modest leverage for a lender. The risk is not that Affirm cannot service its debt; the risk is that the assets do not perform. And that brings us to the central gap.

THE CREDIT DISCLOSURE GAP. This file contains no delinquency rate, no charge-off rate, no allowance for credit losses, no vintage curve, no average loan size, no term distribution, no approval rate, and no FICO or risk-tier distribution. For a balance-sheet consumer lender, these are the numbers that determine value, and not one of them is available. We can observe that FY25 gross profit was $2.177B on $3.224B of revenue (a 67.5% gross margin, which for a lender embeds provisioning) and that this margin has been stable at 63-68% for four years — weak indirect evidence that credit is not deteriorating. That is the entirety of what this dataset permits us to say about credit quality, and it is why this dive is a Stage-In rather than a Buy.

Cash flow:

Free cash flow has doubled for two consecutive years and the company has begun returning capital — a $250M buyback in FY25, roughly 1% of the market capitalisation. The $1.294B of net debt issuance is funding growth in the loan book, not covering a cash deficit (free cash flow was positive $601.7M), which is a crucial distinction from the pattern seen at other loss-making names in this batch. Stock compensation of $321.4M is 10.0% of revenue — meaningful, but roughly half the level seen at the software names here.

Diluted share count: 269.5M (FY21) → 341.0M (FY25) → 348.1M (Q3 FY26). That is +29% over five years, but only +2.1% over the last four quarters — dilution has decelerated sharply and the buyback has begun. This is a materially better shareholder-yield picture than most of this batch.

4. Data integrity

REJECTED — the consensus EBITDA series. Every forecast year shows negative ebitdaAvg: -$903.9M (FY26E), -$1.135B (FY27E), -$1.417B (FY28E), -$1.754B (FY29E), -$3.657B (FY30E) — while the same rows show positive netIncomeAvg of $432.4M, $633.5M, $900.1M, $1.286B and $1.401B. Net income above EBITDA is impossible. The series is corrupt. It also contradicts the vendor's own km_ttm.evToEBITDATTM of +28.45, which is computed from the reported FY25 EBITDA of $712.0M. No EBITDA-based multiple appears in this dive. We would add that "EBITDA" is a conceptually inappropriate metric for a lender in any case, since adding back $425.5M of interest expense removes the principal cost of the business.

REJECTED — the FY30 revenue estimate. Consensus FY30 revenue of $17.002B against FY29's $8.154B is a 108% single-year increase, on 9 analysts, in a series that grows 24-26% in every other year. It is not credible and it is excluded from all analysis. FY26 through FY29 (24, 25, 23 and 11 revenue analysts respectively) are used.

FLAGGED — the ratio distortions in section 3 (currentRatioTTM 717.0, interestCoverageRatioTTM 1.0003, netDebtToEBITDATTM 5.77), all of which are lender-accounting artefacts rather than corruption per se, but all of which would produce false signals on a generic screen.

FLAGGED — the segment disclosure covers only 34.5% of revenue (section 1).

FLAGGED — the profile and technical blocks disagree on the 52-week range: $42.095-$100.00 versus $42.53-$92.18. The high differs by 8%. Percentage-from-high figures depend on which is used (-24.4% versus -18.0%); we use the technical block and note the discrepancy.

CHECKED AND CLEAN — the most recent quarter. inc_q[0] (Q3 FY26, 2026-03-31) passes every test: net income of $102.9M against 348.1M diluted shares gives $0.296 versus the reported $0.30; share counts (337.1M basic / 348.1M diluted) do not duplicate the prior quarter and show a normal basic-to-diluted spread; interest expense of $113.8M is a plausible continuation of $110.0M and $111.7M; the reported EPS matches the earnings-calendar actual of $0.30 exactly. No sign flips, no zeroed share counts, no duplicated fields.

CHECKED AND CLEAN — the estimate EPS series. Nine-month FY26 actual EPS of $0.90 plus the $0.3486 Q4 estimate equals $1.249 against the consensus FY26 figure of $1.251 — a 0.2% difference. The estimate series is on the same basis as reported results, which is worth noting because it is not true of several other files in this batch.

5. Knowledge base — a real lane, including a falsified bear call

Five tagged claims across four tracked sources, and this is one of the more useful lanes in the batch because it contains a dated, gradeable error.

The bear claim — and its falsification:

• compound_and_friends, 2022-12-18 (bearish, conviction 78, entities: AFRM, UPST): "BNPL/lending-platform stocks (Affirm, Upstart) are down ~90% and on their way to zero — you can't reinvent lending; when ABS buyers vanish they're forced to hold deteriorating loans."

This claim has been decisively falsified by results. In the three and a half years since, Affirm has gone from a $985.3M annual loss to four consecutive profitable quarters, $601.7M of free cash flow, a 19.5% operating margin and a $250M buyback. The securitisation market did not vanish; net debt issuance funded growth in every subsequent year. We name this rather than quietly omitting it, because a knowledge base that only records claims that aged well is not a knowledge base. It is also a useful calibration: the specific mechanism the claim identified (funding-market closure forcing balance-sheet retention of deteriorating loans) remains a legitimate risk — it simply did not occur in this cycle.

The bullish lane — recent and high-conviction:

• invest_like_the_best, 2025-12-16 (bullish, conviction 82, entities: Affirm, Max Levchin): "Max Levchin exemplifies the resilient act-two founder — deep technical mastery, elite recruiting, strong leadership — making Affirm a backable compounder."
• invest_like_the_best, 2025-12-16 (bullish, conviction 80, entities: DoorDash, Affirm, Toast): "Durable led the last private rounds of DoorDash, Affirm and Toast and still holds them — persisting and buying more across the curve." — a professional investor disclosing a maintained and increased position, which is revealed preference rather than commentary.

The structural claim:

• business_breakdowns, 2023-05-20 (bullish, conviction 68, entity: Afterpay): "Gen Z and millennials are structurally moving away from credit cards toward BNPL, a win-win-win for consumer, merchant (20-30% higher AOV/conversion), and credit provider." — a category claim rather than a company claim, but it is the demand thesis underneath the whole business, and the quantified merchant benefit (20-30% higher average order value and conversion) is why merchants pay the fee.

One older trading claim:

• marko_papic, 2022-01-20 (bullish, conviction 55, entity: AFRM): "Affirm dropped from 175 to 65 into major technical support; expects a ~20% bounce in the next couple weeks — a trade, not a hold." — explicitly a short-term technical call, self-described as "a trade, not a hold," and carrying essentially no weight for a multi-year view.

Read: Moderate-High conviction. The lane is recent (December 2025), high-conviction (80-82), sourced from an independent investor describing a maintained position, structurally supported by the 2023 category claim, and — importantly — the one bear claim has been tested by three and a half years of results and falsified. That is a stronger evidentiary position than a lane of uncontested bullishness, because the bear case was stated, dated, and beaten.

Search-hygiene note, and a good illustration of the problem. A naive text search on this ticker returns eighteen hits, thirteen of which are false positives generated by the ordinary English verb "affirm" — Darius Dale on whether the Fed will "affirm consensus" (three separate claims), Lex Fridman on treaties that "affirm sovereign states' free choice", a chief executive "reaffirming" a timeframe, Netflix management "reaffirming" revenue guidance, Real Vision on credit spreads "reaffirming the bull market", and several more. Every one of them was read and excluded. The honest count is five, and this is exactly the failure mode the house has flagged before: a ticker whose letters spell a common word will produce a spuriously impressive claim count unless each hit is individually inspected.

6. Valuation — priced in or room?

At $75.63 (market cap $25.328B):

Price-to-free-cash-flow: 32.2x on FY25's $601.7M. Price-to-book: 6.74x. The FY30 estimate is rejected (section 4). Coverage depth here — 22 EPS analysts on FY28 — is the best in this batch alongside Atlassian, and materially better than most.

6a. What today's price assumes (the inversion)

At $75.63 — 42.0x FY27E and 29.3x FY28E — the price embeds the following falsifiable claims:

• Revenue compounds at ≥24% through FY29 (consensus: +25.5%, +24.8%, +23.8%), from a base that grew 31.8% over the first nine months of FY26. The price does not require growth to accelerate; it requires roughly three-quarters of the current rate to persist for three years.
• EPS reaches ~$2.58 by FY28 (consensus, 22 analysts) — a 107% increase from the $1.25 FY26 base in two years, which requires operating margin to continue expanding from the 19.5% just delivered.
• Credit losses do not deteriorate materially. This is the assumption doing the most work and it is the one that cannot be verified from this file. Gross margin has been stable at 63-68% for four years, which embeds provisioning; a two-to-three-point deterioration in that margin would consume a large fraction of the operating profit that the entire valuation now rests on.
• Funding costs remain manageable. Interest expense was $425.5M in FY25, 13.2% of revenue, against $7.85B of debt. A material rise in funding spreads compresses the lending margin directly.
• The US consumer holds up. 96.3% of revenue is domestic and the international business is growing more slowly, so there is no geographic hedge.
• The market keeps paying ~30-40x forward earnings for a consumer lender — a multiple normally reserved for asset-light businesses, on a company carrying a $9.76B balance sheet.

The single most fragile assumption: credit performance. Everything else is observable and trending favourably. Credit is neither.

6b. The return bridge (why the multiple moves)

Expected return ≈ EPS growth + multiple drift + shareholder yield.

• EPS growth: consensus +43.9% (FY27E) and +43.6% (FY28E) — the dominant term by a wide margin, and supported by seven quarters of demonstrated operating leverage rather than by forecast alone.
• Multiple drift: our base case assumes COMPRESSION, applying 32x to FY28E against the 42.0x the market pays on FY27E today. The reasons: (i) as a lender's earnings mature, the market applies lending multiples rather than payments multiples, and a 42x forward multiple on a balance-sheet business is a growth-company rating that historically compresses; (ii) the invisible credit variable warrants a permanent discount; (iii) a beta of 3.674 means the multiple is itself highly volatile. The mechanical roll from 60.4x FY26E to 29.3x FY28E at a constant price is not compression — that is arithmetic, and we do not count it. What we assume is that the market pays 32x at the horizon versus 42x today.
• Shareholder yield: approximately +0.9% and improving. The $250M FY25 buyback is roughly 1% of market capitalisation, dilution has slowed to +2.1% over the last four quarters from +29% over five years, and no dividend is paid. This is the second-best shareholder-yield trajectory in the batch, behind only Restaurant Brands, and it is a genuine change of character.

The bridge reads: very strong earnings growth (~44% annually), partly offset by assumed multiple compression, with a small positive yield. Net approximately +10% to base fair value — and note that the earnings term is doing all the work while the multiple term subtracts. That is the correct shape for a thesis: it does not need the market to re-rate anything.

Scenario bands (stated arithmetic, no DCF):

• Bear ~$33 — 18x FY27E EPS of $1.800. A consumer credit cycle turns, loss rates rise two to three points, the operating margin gain reverses, and the market applies a lending multiple to impaired earnings. -56% from spot. Note this is close to the $42.53 52-week low but below it — a genuine credit event would take this stock lower than its recent range, and the -55.1% maximum drawdown already on file shows the capacity for it.
• Base ~$83 — 32x FY28E EPS of $2.584. Revenue compounds at roughly the consensus 25%, operating margin holds near 19-20%, credit performs, and the multiple compresses modestly with earnings maturation. +10% from spot.
• Bull ~$113 — 30x FY29E EPS of $3.772. Growth stays above 25%, operating margin pushes past 22%, international begins to matter, and the buyback scales. +49% from spot. This is above the $106 street high.

Base sits 10% above spot and 7% below the $89.47 street consensus. We are marginally more cautious than a sell-side that is 23-to-1 positive, and the gap is entirely the credit discount.

6c. Variant perception (where we differ, what would surprise)

• Our variant perception is that the operating-leverage inflection is more durable than the multiple implies, and the market is still pricing an old story. The stock's twelve-month return is +10.3% against SPY's +19.9% — it has underperformed the index during the four quarters in which it turned profitable and expanded operating margin by twenty percentage points. That disconnect is the opportunity. The market is still trading Affirm as the loss-making 2022 business that a tracked source said was going to zero, while the income statement has become something else entirely.
• We differ from the street by discounting for what cannot be seen. Twenty-three buy ratings and an $89.47 target implicitly assume normal credit. We do not dispute that; we simply refuse to size a position as though we can verify it, because this dataset contains zero credit metrics. Our $83 base is the same forecast as the street's with an uncertainty discount attached.
• We weight the falsified bear claim as evidence, not embarrassment. A structural bear case was made in December 2022 at conviction 78 by a tracked source, was specific about its mechanism, and has been beaten by results. That materially raises our confidence in the franchise's durability — it has already been stress-tested against the most-cited bear argument.
• Positive surprise that would force a repricing: the 2026-08-27 print delivering operating margin above 20% with revenue above the $1.106B consensus, plus any disclosure of credit metrics showing stable or improving loss rates. The second half of that would be worth more than the first, because it would remove the single largest discount in our model. An expanded buyback would be a further positive.
• Negative surprise: gross margin falling below 63% (the four-year floor, and the only proxy for credit deterioration available here); operating margin retreating below 10%; revenue growth below 20%; interest expense rising above 15% of revenue; or any commentary indicating funding-market stress — the precise mechanism the 2022 bear claim identified, which remains a real risk even though it did not occur.

7. Moat and competitive position

Affirm's moat is a two-sided network with underwriting as the differentiator. Merchants integrate Affirm because it demonstrably raises conversion and average order value — the knowledge base quantifies the benefit at 20-30% — and consumers return because approval is instantaneous and terms are transparent. Each side reinforces the other: more merchants make the consumer proposition more useful, more consumers make the merchant proposition more valuable.

Evidence the moat is working: Merchant Network revenue growing 30.8% and Virtual Card Network 52.8% (the fee-based components growing at least as fast as the whole); a 67.5% gross margin sustained for four years; and — the strongest evidence — operating margin expanding 38.5 percentage points in under two years, which does not happen in a commoditised business where pricing is under pressure.

Where the moat is weak. Underwriting quality is the differentiator and it is unverifiable from this file. Barriers to entry are moderate — card networks, large technology platforms and banks can all offer instalment products. And the balance-sheet model means growth consumes capital, which structurally caps the compounding rate relative to a pure payment processor.

The vendor's peer set (Check Point, CyberArk, Figma, Grab, Samsara, Nutanix, PTC, Toast, The Trade Desk, VeriSign) is a market-capitalisation cohort with almost no relevance — only Toast ($19.0B) and Grab ($14.5B) are even loosely comparable, and neither is a balance-sheet lender. No genuine BNPL or consumer-lending comparable appears in this file, which is itself a gap: we cannot benchmark this company's multiple against its actual competitive set.

8. Insider activity — the cleanest tape in the batch

The Form 4 record for 2026-07-01 shows eight transactions and not one open-market sale:

Every transaction is either a routine annual director equity grant (A-Award) or an option exercise with shares withheld for tax (M-Exempt / F-InKind). There is not a single discretionary open-market disposal by any officer or director in the available record, and the President's net holding increased.

This stands in deliberate contrast to three other names in this batch, where chief executives and founders made large open-market sales — including one who disposed of roughly 36% of a personal holding in three weeks, and two who sold a combined $13.1M in two days. Against that comparison, Affirm's record is genuinely clean, and it is a mild positive signal at a moment when the company has just delivered its most profitable quarter. Insider behaviour is weak evidence in isolation; the absence of selling by anyone, during a turnaround that has doubled the stock off its low, is worth a sentence.

9. Verdict, kill-criteria and flip conditions

Stage-In. Affirm has completed a turnaround that the share price has not yet fully credited. Across the first nine months of fiscal 2026 it grew revenue 31.8% to $3.095B and earned $313.2M, against a $517.8M full-year loss two years earlier. Quarterly operating margin has swung 38.5 percentage points in under two years, to +19.5%. Free cash flow was $601.7M in FY25, a $250M buyback has begun, dilution has slowed to 2.1% over four quarters, and every one of the last six quarters beat consensus. Consensus from 22 analysts models EPS to $2.584 by FY28, putting the shares at 29.3x — and the stock has underperformed the S&P 500 over the last twelve months while all of this happened. The knowledge base carries a recent bullish lane at conviction 80-82, and its one bear claim has been falsified by results. The insider tape is the cleanest in this batch.

Three reasons the position is staged rather than completed:

1. Credit risk is unobservable. This file contains zero delinquency, charge-off, loss-rate or vintage data for a lender with a $9.76B receivable book funded by $7.85B of debt. The single most important variable cannot be checked.

2. 96.3% of revenue is United States — an undiversified bet on one country's consumer, with the international business growing more slowly.

3. Beta 3.674, the highest in this batch, with a maximum drawdown on file of -55.1%. Position sizing has to respect that.

Staged entry — build a 1.5-2.5% position in three tranches:

1. Tranche 1 — now (~1/3 of target), at ~$76. Above a rising 50-DMA and 200-DMA, RSI 34.8, four profitable quarters, clean insider tape, 10% to base fair value. You are buying the completed turnaround, not the next print.

2. Tranche 2 — the reaction to the 2026-08-27 print. Add on operating margin at or above 19.5% with revenue above $1.106B — a full profitable fiscal year confirmed. If gross margin falls below 63%, do not add, regardless of the headline.

3. Tranche 3 — either a pullback to the 200-DMA (~$66) that holds, or the first disclosure of credit metrics showing stable loss rates. The latter would justify removing the uncertainty discount entirely.

Pre-registered KILL criteria (any one converts Stage-In → Avoid, and stops further tranches):

• Gross margin below 63% — the four-year floor, and the only proxy available in this dataset for credit deterioration. This is the most important single trigger in the list.
• Operating margin retreating below 10% for two consecutive quarters, reversing the leverage that the whole valuation rests on.
• Revenue growth below 20%.
• Interest expense above 15% of revenue (FY25: 13.2%) — funding-cost pressure compressing the lending spread.
• Any indication of securitisation or warehouse-funding stress — the precise mechanism identified in the falsified 2022 bear claim, which remains a genuine risk even though it did not occur in that cycle.
• The 200-DMA (~$66) decisively lost on company-specific news rather than market beta.

Pre-registered FLIP TO HIGH-CONVICTION (upsize toward 3-4%):

• Disclosure of credit metrics showing stable or improving loss rates — the single change that would most reduce the discount applied here, and the reason this is Stage-In rather than Buy.
• Operating margin sustaining above 22% for two consecutive quarters.
• An expanded buyback beyond the initial $250M, taking the shareholder yield decisively positive.
• International revenue exceeding 8% of the total, beginning to diversify the 96.3% US concentration.

Where AFRM fits in the Synthos Framework Portfolio. The fintech / consumer-credit sleeve at 1.5-2.5%, sized below a core weight for one specific and stated reason: the dominant risk variable is invisible in the available data. It is complementary to, not overlapping with, the payments-infrastructure and software positions elsewhere in the portfolio — this is a balance-sheet lender, and it should be sized as one. Logged as a tracked Synthos call (Stage-In) as of 2026-08-04 at $75.63.

Single biggest risk: credit deterioration that we cannot see coming. Affirm holds a $9.76B consumer loan book funded by $7.85B of debt, with 96.3% concentration in a single country, and this dataset provides no delinquency, charge-off or vintage information whatsoever. The operating leverage that has taken margin from -19.0% to +19.5% would be substantially consumed by a two-to-three-point rise in loss rates, and the first place it would appear is a gross margin that has been stable at 63-68% for four years.

Most fragile assumption in the price: that the market continues to pay ~30-40x forward earnings for a balance-sheet consumer lender. That is a payments-company multiple applied to a business that funds its growth with borrowed money. It is defensible while credit performs and operating leverage compounds; it has no defence at all if either stops.

Provenance & disclosures

• Traceability: 5 tagged KB claims across 4 tracked sources (breadth 4, net conviction positive-moderate). Bullish lane: invest_like_the_best 2025-12-16 (conviction 82, the founder as a backable act-two operator) and 2025-12-16 (conviction 80, a professional investor disclosing a maintained and increased position); business_breakdowns 2023-05-20 (conviction 68, the structural shift from credit cards to instalment payment, quantifying a 20-30% merchant benefit). Bear lane: compound_and_friends 2022-12-18 (conviction 78, "on their way to zero") — a dated, specific claim that four consecutive profitable quarters have DECISIVELY FALSIFIED, named here rather than omitted. One older self-described trading call (marko_papic 2022-01-20, conviction 55, "a trade, not a hold") carries minimal weight. Thirteen additional text matches were excluded as false positives generated by the ordinary English verb "affirm" — including three separate claims about whether the Federal Reserve would "affirm consensus," management "reaffirming" guidance at two unrelated companies, and a geopolitical claim about treaties that "affirm sovereign states' free choice." Each was read individually before exclusion.
• Data as-of: fundamentals 2026-03-31 (FY26 Q3; fiscal year ends 30 June) · estimates 2026-08-04 · prices 2026-08-04 ($75.63, 50-DMA $74.23, 200-DMA $66.12, RSI 34.8, beta 3.674) · KB claims 2026-08-04. Sole market-data source: the Synthos vendor file for this ticker.
• Data REJECTED: (1) the entire consensus EBITDA series — every forecast year is negative (to -$3.657B in FY30) while the same rows show positive net income of up to $1.401B; net income cannot exceed EBITDA, and the series also contradicts the vendor's own evToEBITDATTM of +28.45 computed from reported FY25 EBITDA of $712.0M. No EBITDA-based multiple appears in this dive. (We would add that EBITDA is conceptually inappropriate for a lender in any case, since adding back $425.5M of interest expense removes the principal cost of the business.) (2) The FY30 revenue estimate of $17.002B — a 108% single-year jump from FY29 on 9 analysts, in a series growing 24-26% in every other year. FY26-FY29 estimates are used.
• Vendor ratios that must NOT be read naively (lender-accounting artefacts, not corruption): currentRatioTTM 717.0 (funding debt classified long-term against a current receivable book); interestCoverageRatioTTM 1.0003 (interest is cost of goods sold for a lender, not a financing burden); netDebtToEBITDATTM 5.77 (treats $7.85B of funding debt as corporate leverage — it funds a $9.76B earning asset). All three would flag falsely on a generic screen. Net debt recomputed by hand as $5.627B (total debt $7.853B − cash $1.354B − short-term investments $871.4M); the vendor's $6.499B nets cash only.
• THE CENTRAL DATA GAP: this file contains no delinquency rate, no charge-off rate, no allowance for credit losses, no vintage curve, no average loan size, no term distribution, no approval rate, and no risk-tier distribution. For a balance-sheet consumer lender these are the determining metrics. The only available proxy is the gross margin (63-68%, stable for four years), which embeds provisioning. This gap is the explicit reason the verdict is Stage-In rather than Buy, and the reason the position is sized below a core weight.
• Further data gaps noted: segment disclosure covers only 34.5% of FY25 revenue (Merchant Network and Virtual Card Network), with the interest-income, loan-sale and servicing components — the ones carrying credit risk — not broken out; the profile and technical blocks disagree on the 52-week high ($100.00 versus $92.18); and the vendor's peer set contains no genuine BNPL or consumer-lending comparable, so the multiple cannot be benchmarked against the actual competitive set.
• Data VALIDATED: the most recent quarter (inc_q[0], Q3 FY26) passes the standard corruption screen — EPS reconciles against net income and share count, share counts do not duplicate the prior quarter and show a normal basic-to-diluted spread, interest expense is a plausible continuation, and the reported EPS matches the earnings-calendar actual exactly. The estimate EPS series is on the same basis as reported GAAP results — nine-month FY26 actual EPS of $0.90 plus the $0.3486 Q4 estimate equals $1.249 against a consensus FY26 of $1.251, a 0.2% difference. This is not true of several other files in this batch and is worth noting.
• Fair-value caveat: the $33 / $83 / $113 anchors are scenario multiples applied to consensus EPS, not a discounted cash-flow model. The base case assumes multiple compression (32x at the horizon versus 42.0x FY27E today) and all of the expected return comes from earnings growth, none from re-rating. The bear case explicitly contemplates a consumer credit cycle and sits below the 52-week low, because a genuine credit event would take this security outside its recent range.
• Not investment advice. Independent research, educational and informational only, never personalised.
• Version: 2026-08-04-full.