Pagaya has done the hard part. Six consecutive quarters of GAAP profit, operating costs flat for eighteen months while network volume grew a third, and a genuinely differentiated position inside thirty-six lenders and roughly thirty thousand car dealerships. The stock is optically cheap on almost every headline multiple. It is cheap for identifiable reasons, not because the market has missed something — and on our numbers it is close to fairly priced at $17.62.
Half the headline profit is an accounting posture. Adjusted diluted EPS of $1.07 against GAAP of $0.49. The bridge is $38.0m of credit impairment in the quarter, and $74.4m across the half, on loans Pagaya itself chose to underwrite. For a business whose entire economic function is deciding which consumers to lend to, credit impairment is not a non-recurring item — it is the cost of goods sold.
The economics are deteriorating, and it is visible in one line. Network volume grew 33.5%; gross profit grew 16.4%. Fee revenue less production costs has fallen from 4.96% of volume to 4.16% across four consecutive quarters, and now sits at the floor of management’s own 4–5% target range. Volume is the number the market anchors on. It is the least informative one in the release.
There is no free cash flow. First-half operating cash flow of $117.9m against $6.6m of capex reads as $111m of free cash flow, and several data providers publish a double-digit yield on that basis. In the same period the retained portfolio grew $130.9m net — risk retention that is mandatory, not discretionary. Adjust for it and first-half free cash flow is roughly negative $20m.
On 30 July Pagaya reported adjusted diluted earnings of $1.07 a share and GAAP diluted earnings of $0.49. Both figures are disclosed, both are reconciled in full, and the reconciliation is published prominently rather than buried. Nothing here is hidden. But what sits between the two numbers decides whether this is a cheap compounder or an ordinary specialty finance company priced about right.
| 2Q26 bridge | Amount | What it is |
|---|---|---|
| GAAP net income to PGY holders | $45.3m | Sixth consecutive positive quarter, up 172% year on year |
| Impairment loss on certain investments, net | +$38.0m | Credit losses on retained tranches of Pagaya’s own securitisations |
| Non-recurring expenses | +$3.4m | $4.9m across the half. Recurring non-recurring expenses appear in every period disclosed |
| Share-based compensation and other | +$14.3m | Ordinary add-backs; SBC itself halved year on year to $8.6m |
| Adjusted net income | $101.0m | 2.2x the GAAP figure |
| Adjusted diluted EPS | $1.07 | Against GAAP diluted EPS of $0.49 |
The impairment add-back is the whole argument. Pagaya retains risk-retention pieces and, increasingly, bond tranches of the deals it sponsors — $1,040m of investments in loans and securities at 30 June, roughly half now in bonds against 19% in the first quarter of 2024. When those loans go bad, Pagaya eats the loss. That is not an unusual event visited upon the company from outside; it is the direct consequence of the underwriting decisions that constitute the business. Adding it back to reach “adjusted” earnings is adding back cost of goods sold.
Two things work in the company’s favour here and both are real. The reconciliations are published in full and prominently, so an investor who reads the release cannot be misled by accident. And GAAP operating income of $105.8m, up 87.4%, sits above the credit-mark line and stands entirely on its own without any of this. The bull case does not need the adjusted number. It is stronger without it.
A 224% earnings beat was widely reported on 30 July. It compares the company’s adjusted EPS with a consensus compiled on a GAAP-like basis, and is not a like-for-like comparison. On a GAAP basis the beat was $0.49 against roughly $0.32 — still substantial, and still one of the better prints in the sector this quarter. We flag it because the difference between a 53% beat and a 224% beat is the difference between a good quarter and a re-rating event, and only one of those two things happened.
Every other report on this site discounts unlevered free cash flow to an enterprise value and then bridges to equity with net cash. That is the wrong tool for Pagaya and we have not used it here. Interest income, interest expense and credit marks are all part of this company’s operating model, and its debt largely funds an earning asset portfolio — so subtracting net debt from an unlevered enterprise value would double-count. This model instead discounts free cash flow to equity: GAAP net income, less a retention charge for the mandatory risk-retention portfolio build, discounted at a cost of equity. The output is an equity value directly. The net-debt row in the bridge below is therefore zero by construction, and is shown rather than removed so the arithmetic stays checkable.
The base year is the FY26 revenue guidance midpoint of $1,475m and the share count is 97.2m diluted — the count management itself uses for every EPS figure it quotes, and 16.9% above the basic count of 83.2m. Cash tax is held at 14% across all cases, reflecting an effective rate that has been low and volatile on the Israeli and US structure and on net operating loss carryforwards, and which we assume normalises toward the low-to-mid teens. The NOL slider lets you take a different view of that.
Terminal value is 60% of the total. The net-debt bridge is zero because debt service already sits inside free cash flow to equity.
Reading the inputs. Terminal pre-tax margin of 15.0% at a 14% tax rate is a 12.9% net margin, against 11.7% delivered in the June quarter and roughly 11.4% implied by the FY26 guide. The risk-retention charge of 2.60% of revenue is 20% of that net income, the share we assume the mandatory portfolio build absorbs; the bear case runs it at 30% and the bull at 15%. First-half experience was worse than any of them — a $131m net build against $70m of net income, a draw above 100% — but that period included a deliberate shift of the portfolio mix toward bonds, so we assume it normalises. NOL shield is set to zero in all three presets, which assumes the carryforwards are exhausted by the terminal year; move it if you disagree. The widget is a five-year constant-growth form, so the growth rates shown — 2.4% / 8.4% / 14.8% — are the constant-rate equivalents of the underlying 4% / 14% / 22% three-year rates fading to terminal over four, six and eight years. They produce the same present value; they are not the near-term growth forecast. The probability-weighted case at $19.67 cannot be loaded as a preset — the model supports three — so it is set out below instead.
| Scenario | Value per share | vs $17.62 | Implied mkt cap | Implied P/E, FY26E |
|---|---|---|---|---|
| Bear | $3.95 | −78% | $384m | 2.3x |
| Base | $16.57 | −6% | $1,611m | 9.6x |
| Bull | $41.59 | +136% | $4,044m | 24.1x |
| Probability-weighted (25/50/25) | $19.67 | +12% | $1,913m | 11.4x |
$3.95 is not a solvency forecast. It is what falls out of two adverse assumptions compounding at once: net margin halving to 5.5% and cost of equity rising to 17.5% simultaneously. If only one of those happens the number lands closer to $8–$11, which is roughly where the stock actually traded between February and April this year. Treat $8–$11 as the realistic stress zone and $4 as the tail.
| Cost of equity ↓ / Terminal g → | 2.0% | 2.5% | 3.0% | 3.5% | 4.0% |
|---|---|---|---|---|---|
| 13.0% | $18.38 | $19.17 | $20.04 | $21.01 | $22.08 |
| 14.0% | $16.78 | $17.43 | $18.15 | $18.93 | $19.79 |
| 15.0% | $15.42 | $15.97 | $16.57 | $17.22 | $17.92 |
| 16.0% | $14.26 | $14.73 | $15.24 | $15.78 | $16.37 |
| 17.0% | $13.26 | $13.67 | $14.10 | $14.56 | $15.06 |
Note how flat this surface is. Across the entire plausible range of discount rate and terminal growth, value lands between $13 and $22. Terminal assumptions are not what drives this stock, which is unusual and worth knowing before spending an afternoon arguing about the equity risk premium.
| Revenue growth ↓ / Terminal net margin → | 9% | 11% | 13% | 15% | 17% |
|---|---|---|---|---|---|
| 6% | $10.60 | $12.32 | $14.05 | $15.77 | $17.50 |
| 10% | $11.50 | $13.38 | $15.27 | $17.16 | $19.04 |
| 14% | $12.45 | $14.51 | $16.57 | $18.63 | $20.69 |
| 18% | $13.46 | $15.71 | $17.95 | $20.20 | $22.45 |
| 22% | $14.54 | $16.98 | $19.43 | $21.87 | $24.32 |
Value is roughly twice as sensitive to the terminal net margin as to the growth rate. Moving growth from 6% to 22% at a 13% margin adds $5.38 a share; moving margin from 9% to 17% at 14% growth adds $8.24. In plain terms: Pagaya’s equity value is a bet on take-rate and credit costs, not on volume. Which is exactly why the FRLPC percentage is the most important line in the release, and why the 33% volume headline is the least informative one.
What you are holding, in the terms those two grids describe. Pagaya’s thirty-day realised volatility was running at approximately 65% annualised at the time of writing and its twelve-month figure at approximately 78%, and the shares have ranged from $10.40 to $44.99 over twelve months. Consumer credit is a cyclical, leveraged industry with concentrated partner and vertical exposure; a substantial portion of this issuer’s assets are carried at fair value using the company’s own loss assumptions; and the equity value is most sensitive to the cost of equity and the terminal margin, which is exactly what the grids above are showing you. The share price used throughout is the $17.62 quote as at 31 July 2026 — the shares have traded in the sessions since, and every multiple in this report scales directly with it.
Holding the base case’s 15% cost of equity, 3% terminal growth, 13% terminal net margin and 20% retention charge, today’s $17.62 implies a 17.1% revenue compound growth rate over six years, fading to 3% thereafter. Against FY26 guided revenue growth of roughly 13.5% and second-quarter actual of 18.6%, that is demanding but not absurd. The market is asking Pagaya to sustain approximately its current revenue growth rate for six years while holding a 13% net margin. That is a fair description of what you are underwriting if you buy the shares here.
| Measure | On basic (83.2m) | On diluted (97.2m) | Basis |
|---|---|---|---|
| Market capitalisation | $1,466m | $1,714m | At $17.62 |
| Enterprise value | $2,091m | $2,339m | Plus $875m debt, less $249m unrestricted cash |
| P/E — LTM GAAP | 11.5x | 13.5x | LTM net income of $127m |
| P/E — FY26E guide midpoint | 8.8x | 10.2x | Guided net income of roughly $168m |
| P/E — on 4Q26 exit run-rate | 7.3x | 8.6x | Company-implied, above $200m annualised |
| EV / EBITDA — LTM adjusted | 4.9x | 5.5x | EBITDA of $423m, which excludes credit marks |
| EV / Revenue — LTM | 1.50x | 1.68x | LTM revenue of $1,390m |
| Price / Book | 2.47x | 2.88x | Shareholders’ equity of $594m |
| Price / free cash flow | n.m. | n.m. | Free cash flow after portfolio build was roughly −$20m in 1H26 |
| Return on equity, LTM | ~21% | — | Company cites 22% |
On headline multiples Pagaya is the cheapest profitable name in its group by a wide margin. Upstart, growing faster on revenue and far less profitable, trades at something like five times Pagaya’s earnings multiple. That gap is the entire investment debate — and the discount is not a market error. It is a stack of specific, defensible haircuts: an earnings-quality haircut for the impairment add-back, a balance-sheet haircut for $1.04bn of Level 3 assets at 1.75x book equity, a cost-of-capital haircut against deposit-funded peers, a dilution haircut for a diluted count 17% above basic, a litigation overhang, and a cyclicality haircut for realised volatility of 65–78% and a beta well above two. On adjusted EBITDA the stock is at 4.9x; on GAAP earnings after credit marks it is at 10–13x. The second number is the honest one, and it is no longer conspicuously cheap.
Peer market capitalisations referenced in this report are third-party aggregator estimates of varying dates in July 2026 and are indicative only, not a precise like-for-like screen. Pagaya’s own figures throughout are computed from primary filings and the 31 July quote.
Same model, different inputs. Click any card to load it.
Weighting those 25 / 50 / 25 gives $19.67, about 12% above the quote. We publish that number and do not lead with it, because a probability-weighted average of a distribution this skewed is arithmetic rather than insight: the bull case is 2.5 times the base and the bear is a quarter of it. Our stated fair-value range is $14–$22, with a central estimate of $16.50–$17.00 — essentially where the shares trade today.
| 2Q26 | Result | YoY | Versus guidance |
|---|---|---|---|
| Network volume | $3,535m | +33.5% | Beat top end of $2,875–3,075m |
| Total revenue & other income | $387.0m | +18.6% | Beat top end of $345–365m |
| FRLPC | $146.9m | +16.4% | — |
| FRLPC % of network volume | 4.16% | −61bps | Low end of the 4–5% target |
| Adjusted EBITDA | $123.5m | +43.2% | Beat top end of $100–115m |
| GAAP operating income | $105.8m | +87.4% | — |
| GAAP net income | $45.3m | +171.8% | Top of the $25–45m range |
| GAAP diluted EPS | $0.49 | +145% | Company record |
| Losses on investments in loans & securities | $42.3m | 3.0x | Credit marks on the retained book |
| Core operating expenses | $45.6m | −6.4% | 31% of FRLPC, a record low |
| Diluted share count, weighted | 97.2m | +22.1% | Basic rose 8.2% to 83.2m |
Read the first four rows together, because they contain the whole quarter. Volume grew 33.5%, revenue grew 18.6%, and gross profit grew 16.4%. The company is running faster to stand still on economics. FRLPC as a percentage of volume has now fallen from 4.96% in the third quarter of last year to 4.85%, then 4.63%, then 4.16% — four consecutive declines, arriving at the bottom of management’s own target range.
CFO Jon Dobres attributes this to two things: new partners and products entering at lower initial margins by design, and elevated benchmark rates compressing the funding-side spread. Both explanations are credible. Neither is temporary on any near-term horizon — the ten-year Treasury closed at 4.67% and the thirty-year at 5.21% on 30 July, and management explicitly assumes rates stay elevated through year end. Management also disclosed pricing ABS deals with more conservative loss assumptions, trading day-one revenue for vintage stability. That is prudent, and it is a margin headwind.
Auto was the whole story: more than 75% of year-on-year volume growth, a $4.8bn annualised run-rate, and volume up roughly 140% year on year. Management credits three product changes rolled out over six to nine months — dynamic offer optimisation varying down payment, APR, LTV and term in real time rather than issuing one static approval; better alignment on ticket sizes; and a new “counters” construct where partners route Pagaya applications they previously kept for themselves, because Pagaya’s approval closes better at the dealer desk. President Sanjiv Das says close rates on core auto segments improved by two to three times.
The credit-discipline tell is that conversion held at approximately 1% of application dollars while applications grew 29% and passed $300bn in a quarter for the first time. Flat conversion on a much larger pool is growth from flow, not from loosening the bar. Asked directly by Stephens whether underwriting had loosened, management said no, posture unchanged since December when the highest-risk tiers were exited. Personal loans hit an all-time high, with the Affiliate Optimizer Engine contributing over $1bn of volume. Point-of-sale is the weak spot: one POS partner rolls off in the back half of 2026, carrying very little FRLPC but visibly denting the volume line in the fourth quarter.
| Metric | 3Q26 guide | FY26 guide | Implication |
|---|---|---|---|
| Network volume | $3.425–3.625bn | $12.50–13.25bn | Roughly flat sequentially against $3.535bn — management is not extrapolating the auto step-function |
| Total revenue & other income | $370–390m | $1.425–1.525bn | Midpoint of $1,475m is our model base year |
| Adjusted EBITDA | $120–130m | $460–490m | Raised |
| GAAP net income | $42–52m | $155–180m | Raised roughly 25% at the midpoint; implies a 4Q26 exit rate above $200m annualised |
The FY26 midpoint of $12.875bn implies second-half volume of $6.72bn against $6.16bn in the first half, which is achievable and already embeds the POS roll-off. The guide is credible rather than heroic, which is consistent with a company that beat the top end of every single guided metric this quarter and raised the full year three consecutive quarters in 2025. Guidance accuracy is the strongest single item in our management assessment.
Confident, occasionally over-polished. Gal Krubiner used “flywheel,” “moat,” “engine” and “compounding” repeatedly; the letter and the call together contain three separate flywheel constructions. Das was the most measured voice, saying that those who have been through several cycles are extremely humble about what they do not know — a good instinct from someone who ran CitiMortgage through the last one. Dobres did not oversell: asked by KBW whether the negative $23m capital-markets line was a run rate, he said no, then immediately conceded that elevated rates will continue to pressure the funding side. That is an honest answer rather than a spun one.
One. Krubiner cited a total addressable market of almost $1 trillion on the call while the shareholder letter published the same morning says almost $900bn. Small, but it is the kind of number that should not drift within a single morning. Two. Management spent the closing remarks correcting perceptions rather than celebrating — Krubiner listed four “fundamental small misperceptions” covering product breadth, borrower quality, earnings power and the balance sheet. When a management team with a record quarter uses its last three minutes on rebuttal, it tells you what the market actually disbelieves.
| Item | Timing | Why it matters |
|---|---|---|
| FRLPC % in 3Q26 | ~early Nov 2026 | The highest-leverage single item in the report. At or above roughly 4.2% validates the mix explanation. Below 4.0% breaks the guided range and the earnings model needs re-rating. |
| Losses on investments in loans and securities | Quarterly | $42.3m in 2Q, tripled year on year. If this keeps growing faster than volume, the asset-light framing fails. |
| ABS issuance cadence | Monthly to bi-monthly | A $900m AAA-rated personal-loan deal closed in July and a record $750m auto deal shortly after quarter end. A pulled or downsized deal would be read as a credit signal. |
| Named regional bank partners | “Coming quarters” | Promised for three quarters now. The highest-quality validation available — and further slippage damages credibility. |
| Potential Fitch upgrade | 6–18 months from 15 Jul 26 | Outlook already Positive. Lower funding cost feeds straight into FRLPC and validates the balance sheet against the short thesis. |
| 4Q26 volume against the POS roll-off | ~early Feb 2027 | Margin-neutral by management’s account, but the market anchors on the volume headline. |
| 2025–26 vintage seasoning | Continuous | Visible monthly in ABS trustee and remittance data, months before it reaches the P&L. The most valuable independent verification available to an outside investor. |
Pagaya is not a lender in the way most people use the word, and it is not purely a software vendor either. Thirty-six lending partners — banks, credit unions, auto lenders, fintechs, point-of-sale providers — pipe consumer credit applications into Pagaya through a pre-built API that lives inside the partner’s own loan origination system. Pagaya’s models approve roughly 1% of application dollars. The approved loans are bought and packaged into asset-backed securities sold to 174 institutional investors. Pagaya earns a cash fee on each loan it places.
Historically this was decline monetisation — loans the partner had already rejected. Management now says more than 45% of flow comes from non-decline sources: pre-screen, affiliate marketing, and counters, where Pagaya’s approval replaces a partner’s less attractive one. The second quarter was the largest funding quarter in company history at $3.7bn across six ABS transactions, with the last three upsized.
| Revenue line | 2Q26 | 2Q25 | What it is |
|---|---|---|---|
| Revenue from fees | $365.6m | $317.7m | Cash fees from partners and funding vehicles — the real engine. Growth on this line alone was 15.1%. |
| Interest income | $22.2m | $10.7m | Yield on retained ABS bonds. Doubled as the book shifted toward bonds; now 5.7% of the total. |
| Investment income, net | −$0.8m | −$2.1m | Small and volatile |
| Total revenue & other income | $387.0m | $326.4m | — |
| Less: production costs | −$218.7m | −$191.5m | Cost of acquiring and funding the loans |
| = FRLPC | $146.9m | $126.2m | The company’s preferred gross-profit measure |
One presentational point is worth flagging here rather than in the red-flag section, because it is about arithmetic rather than intent. Interest income is reported gross inside “total revenue and other income,” but it is earned on a bond portfolio funded largely with $875m of debt, and the $19.7m of interest expense sits below the operating-income line. Presenting the income gross while the funding cost is reported elsewhere flatters the revenue growth rate. On fee revenue alone, growth was 15.1% rather than 18.6%.
Total assets of $1,693m at 30 June, of which $1,040m is investments in loans and securities carried at fair value — 61% of assets and 1.75 times shareholders’ equity. Against that: $253m of secured borrowing, $150m of exchangeable notes and $472m of long-term debt, so $875m gross and $626m net of $249m of unrestricted cash. Shareholders’ equity of $594m, up from $480m at year end, sits on an accumulated deficit of $793m against $1,409m of paid-in capital. Debt to equity is 1.47 times. Fitch revised its rating outlook to Positive on 15 July. This retained book is the reason the company is not a pure-play software business, and it is the single most important thing to understand about the equity.
| Company | Model | How it differs |
|---|---|---|
| Upstart | AI lending marketplace | The closest analogue. Sells its model to banks and credit unions but owns the borrower-facing funnel and brand, converting roughly 16% of applications against Pagaya’s 1%. Higher revenue multiple, weaker GAAP profitability. |
| Happen (ex-LendingClub) | Chartered digital bank | Rebranded in June 2026. Deposit-funded, so a structurally lower cost of capital — the advantage Pagaya most conspicuously lacks. |
| SoFi | Diversified fintech bank | Also deposit-funded and far larger. Same structural advantage. |
| Enova | Balance-sheet near-prime lender | Two decades of proprietary underwriting data. Takes full credit risk at higher APRs, with a much smaller distribution surface. |
| Affirm | BNPL / point-of-sale | Overlaps only in POS. Merchant-side network, far larger, real consumer brand. |
| The partners themselves | In-housing | The genuine long-term threat. Every partner has an incentive to build the capability once it proves valuable. Pagaya’s answer is embeddedness and data scale. |
| Dimension | Score | Our read |
|---|---|---|
| Distribution | 8/10 | The strongest advantage. The API sits inside 36 partners’ loan origination systems, reaching roughly 30,000 dealerships, above 40% of the US auto market. Onboarding compressed two to three times and new partners scale four to five times faster because the integrations are pre-built. A competitor cannot replicate this by writing better models; it has to win the same enterprise sales cycles one at a time. |
| Switching costs | 7/10 | Once embedded in a bank’s origination workflow, removal means re-engineering a regulated system and reopening credit-policy questions. Dobres’s framing — difficult to win but highly durable once established — matches the evidence: more than twenty partners added over five years with no disclosed material losses. |
| Data | 7/10 | Over $4 trillion of cumulative applications and above $300bn in the June quarter alone. A genuinely large and proprietary dataset covering near-prime borrowers across a cycle. The caveat is that the incremental value of the millionth application is far below the thousandth, and rivals have comparable depth in overlapping segments. |
| Network effects | 6/10 | Real but two-sided and asymmetric. More partners means more applications, better models, better offers, more dealer routing; separately, 174 institutional investors create funding depth a new entrant cannot match. The weakness is that neither side is locked in by the other — a partner does not care how many ABS investors exist, only about approval quality. |
| Scale | 6/10 | Roughly $12.9bn of expected FY26 volume gives ABS pricing power and fixed-cost absorption, and core opex at 31% of FRLPC is genuinely best-in-class. But at a $14bn origination run-rate against a $900bn market, Pagaya has no market-structure scale. |
| Customer loyalty | 6/10 | Partner retention appears high. But Pagaya has no relationship with the end borrower, who experiences the partner’s brand. That protects it from consumer churn and denies it any consumer franchise. |
| Intellectual property | 5/10 | Proprietary models, but nothing patent-protected or structurally unavailable, and underwriting talent is mobile. Krubiner’s remark that agentic AI lets the team do more with the same cuts both ways — the same tooling is available to every competitor and to every partner’s in-house team. |
| Cost advantage | 5/10 | Excellent on operating cost: near-zero marketing spend, flat opex, roughly $8m of average contribution per partner. Poor on cost of capital, which is the cost that actually matters in lending. This is the single dimension where the moat is narrowing as benchmark rates stay high. |
| Regulatory barrier | 4/10 | Licensing and fair-lending compliance create modest friction, but the company’s own risk factors flag FTC and CFPB scrutiny of AI underwriting as a live issue. Regulation is currently more a risk than a barrier, and true-lender reclassification would be genuinely damaging. |
| Brand | 3/10 | Effectively no consumer brand, by design, and limited B2B brand beyond the sector. Weak, but largely irrelevant to the model. |
Pagaya is winning the fight for shelf space inside lenders and losing ground on what it can charge for it. Expanding: distribution, switching costs, funding network. Stable: the data advantage and the operating cost structure. Narrowing: cost of capital against deposit-funded rivals, and take-rate — which is the visible economic expression of the moat. FRLPC falling from 4.96% to 4.16% while volume grew 33% is what a narrowing pricing moat looks like, whatever the mix explanation. Over two years distribution probably dominates. Over five, the question is whether take-rate stabilises above 4%, and that is fundamentally an interest-rate and competitive-intensity question rather than one Pagaya controls.
We ran a deliberately hostile read of the filings, looking for the reason not to own it. Severity is scored one to ten, where one is cosmetic and ten is disqualifying.
| Area | Severity | Evidence, and why it matters |
|---|---|---|
| Non-GAAP adjustments | 8/10 | The single largest issue, set out in full in Section 01. Impairment losses of $37.97m in the quarter and $74.35m across the half are added back to reach adjusted net income of $101.0m against GAAP of $45.3m. The definition also sweeps in non-recurring expenses of $3.4m in the quarter, and recurring non-recurring expenses have appeared in every period disclosed. Mitigant: full reconciliations are published prominently, and GAAP operating income of $105.8m stands on its own. |
| Free cash flow reality | 7/10 | Headline 1H26 operating cash flow of $117.9m against $6.6m of capex looks like $111m of free cash flow. In the same period the company purchased $496.0m of investments in loans and securities against $345.7m of maturities and prepayments and $19.4m of sales — a net portfolio build of $130.9m. Adjust for it and first-half free cash flow is approximately negative $20m. The retained book is not optional; it is risk retention required to sell the ABS. Any FCF-based valuation that ignores it is wrong by roughly $130m per half-year. |
| Fair-value / Level 3 assets | 7/10 | $1,040m of investments carried at fair value — 61% of total assets and 1.75x shareholders’ equity. These are retained tranches of Pagaya-sponsored deals; there is no deep secondary market for most of them, and the marks depend on the company’s own loss assumptions. A 10% error in carrying value is about $104m, or 17% of book equity. The company points to $19m of note sales this quarter and $365m of year-to-date back-book capital return as evidence the marks are realisable. That is real evidence, and it is partial. |
| Customer / vertical concentration | 7/10 | Thirty-six partners sounds diversified. Economically it is not: more than 75% of year-on-year volume growth came from auto, driven by specific product unlocks at individual named partners, and one POS partner rolling off is material enough to be flagged in guidance. Average customer contribution of roughly $8m against $538m of LTM FRLPC implies the top handful of partners carry a large share. Concentration is the risk factor Pagaya itself lists first. |
| Margin quality | 6/10 | FRLPC has declined four consecutive quarters, from 4.96% to 4.16%, while volume grew 33%. Both of management’s explanations are plausible and both point to continued pressure. The forward guide of 4–5% is a wide band that would accommodate another 16bps of decline without technically missing. |
| Revenue quality | 6/10 | Interest income reported gross inside total revenue while the funding cost sits below the operating-income line. On fee revenue alone, growth was 15.1% rather than 18.6%. |
| Dilution | 6/10 | Weighted diluted shares rose 22.1% year on year to 97.2m while basic rose 8.2% to 83.2m. The 14.1m gap — 16.9% of basic — is warrants, exchangeable notes, RSUs and options. Company history is worse than the current run-rate suggests: a 1:12 reverse split in March 2024, and $1,409m of paid-in capital against a $793m accumulated deficit. Mitigant: SBC halved year on year and the company has repeatedly stated it neither needs nor plans to issue equity. |
| Leverage | 5/10 | $875m gross debt against $594m equity and $249m unrestricted cash. On the other hand interest expense is falling, coverage is improving, the debt is substantially matched against the investment portfolio, and Fitch moved the outlook to Positive citing exactly those improvements. A manageable structure that would become unmanageable quickly if ABS markets closed. |
| Short-seller and legal overhang | 5/10 | Iceberg Research published on 11 February 2025 alleging Pagaya misled investors on risk distribution and used financial manoeuvres to inflate fees and mask impairments; shares fell over 13% that day and law-firm investigations followed. No adjudicated finding exists and the company has continued to grow and pass audits since. But note the uncomfortable symmetry: the 2026 accounts show a growing retained book and $74m of half-year impairment add-backs — the two things the report was about. That does not make the allegations correct. It does mean the market’s discount is not irrational. |
| Management language | 5/10 | Heavy narrative density, and a TAM figure that moved between $900bn and $1trn on the same morning. Counterweight: Das and Dobres both volunteered specific unflattering facts unprompted — the POS roll-off, rate pressure, and conservative ABS loss assumptions costing day-one revenue. |
| Insider activity | 4/10 | Roughly $33.5m of cumulative insider selling over 24 months, concentrated in co-founders Yahav Yulzari and Avital Pardo, with no recorded open-market purchases for most of that stretch. That changed in June 2026: Krubiner bought 16,230 Class A shares on the open market at about $15.43 — a modest sum, but an unforced signal. Most 2026 dispositions are explicitly tax withholding on RSU vesting, which is not informative. Net read: historically negative, currently neutral to slightly positive. |
| Governance | 4/10 | Israeli foreign-private-issuer origins, a dual-class structure with founder control, three co-founders in executive and board seats, and a CFO transition completed 15 June 2026 — the second in the company’s public life. Not alarming; worth monitoring. |
This is not a fraud profile and it is not a clean one. The operating business generates real GAAP profit, real cash fees and demonstrable operating leverage, and the accounting is disclosed rather than hidden. What earns the 6.5 is the systematic direction of every judgement call: impairments excluded from adjusted earnings, portfolio build excluded from the cash-flow narrative, funding costs reported below the revenue line, and a wide guidance band on the one metric that is deteriorating. Individually each is defensible. Collectively they all point the same way, and an investor who reads only the adjusted numbers will materially overestimate this business.
We score management 3.2 out of 5. The operating discipline is real and improving: costs held flat through 33% volume growth, share-based compensation halved, interest expense cut 15%, and the credit posture explicitly tightened in December 2025 and held even as volume accelerated. Exiting your highest-risk tiers and then delivering a record volume quarter seven months later is the behaviour of operators rather than promoters. Guidance accuracy is the strongest single line at 4.5 out of 5 — every guided metric beaten at the top end this quarter, and FY25 raised three consecutive times and landed inside the ranges.
Where they fall short is shareholder-facing. Transparency scores 3.0: genuinely good on operating KPIs — conversion rates, vintage losses against 2021 and 2022 comparables, investor counts, borrower FICO, income and DTI, opex bridges — but the non-GAAP framework itself is aggressive, the free-cash-flow narrative omits the portfolio build, and the TAM drifted within a single morning. Disclosure quality is high; disclosure framing is not neutral. Buybacks score 2.0, because there are none: defensible while retaining capital for portfolio growth and a ratings upgrade, but with the stock at roughly 10x forward earnings and management publicly arguing it is misunderstood, the absence of any repurchase is a gap between words and actions. Krubiner built this from nothing in ten years, which is a genuine achievement; he also took it public by SPAC in 2022 at a valuation that subsequently collapsed roughly 95% and required a 1:12 reverse split. A strong operator whose capital-markets history is poor.
BULLNetwork volume up 33%, applications up 29%, auto up 140% — and every bit of it came from product, not from lowering the credit bar. Conversion held at exactly 1%. You cannot fake that combination. They exited their riskiest tiers in December and still put up a record volume quarter seven months later.
BEARVolume up 33%, gross profit up 16%. Take-rate down for four straight quarters, from 4.96% to 4.16%, now at the floor of their own range. They are buying volume with price and calling it a flywheel. And losses on the retained book tripled to $42m — if credit really is this benign, why are the marks growing three times faster than the volume?
BULLTen times forward earnings and under five times EBITDA for a business compounding GAAP net income at triple digits. Upstart trades at roughly five times this multiple and is barely profitable. The reverse test says you need 17% revenue growth to justify today’s price — they just did 19% in a quarter where volume grew 33%.
BEARStart with the one number that decides this. Adjusted EPS of $1.07, GAAP EPS of $0.49, and the bridge is $38m of credit impairment on loans they themselves chose to underwrite, added back as though it were a restructuring charge. For a credit business that is not an adjustment — it is the cost of the product. Half of the headline profitability is an accounting posture.
BULLThen ignore adjusted entirely. GAAP operating income was $105.8m, up 87%, and it sits above the credit-mark line. GAAP net income was $45m, up 172%, after every impairment. At $17.62 that is 10x forward GAAP earnings. The GAAP number alone supports the case.
BEARThen value it on GAAP and accept that 10x is not cheap for a levered, cyclical credit intermediary with a narrowing spread and an open short-seller file. Upstart’s 48x is not the right comparison. A specialty finance company’s 8–12x is.
BULLCore opex has not moved in eighteen months while profits grew 400%. Marketing spend is essentially zero. That is a software cost structure attached to a credit distribution network, and the platform handles two to three times current volume without new investment.
BEAROperating cash flow was $118m in the half. They also put $131m net into the loan portfolio, because you cannot sell the ABS without retaining the risk. Free cash flow was negative $20m. Every data provider publishing a double-digit free cash flow yield on this name is wrong.
BULLThe portfolio build is growth capital, not a leak. They pulled $365m back out of the back book year to date through resecuritisations and are shifting the mix toward bonds they can borrow against or sell. That is a maturing balance sheet, not a black hole. And credit is fine — personal-loan cumulative losses on recent vintages are 28–34% below the 2021 peak, auto 53–61% below 2022, with average borrower income around $120,000 and FICO around 680.
BEARThat vintage data is management-selected and benchmarked to the worst possible comparison periods — 2021 for personal loans and 2022 for auto, both peak-loss vintages. Beating your worst year by 30% is not the same as underwriting well. Ask again in 2027 when the 2026 auto vintages reach month-on-book 18. And they fund through ABS markets while Happen and SoFi fund with insured deposits. With the ten-year at 4.67% that gap is permanent.
CHAIRThe bear has the stronger analytical case; the bull has the stronger empirical case. The bear’s arguments are structural and verifiable from the filings today: the adjustment framework really does add back cost of goods sold, take-rate really has fallen four quarters running, free cash flow after portfolio build really is negative, and the funding disadvantage really is permanent. None of that is contestable. The bull’s arguments are about what has actually happened: six consecutive GAAP-profitable quarters, costs flat through 33% volume growth, a top-of-range beat on every guided metric, a Fitch outlook upgrade, and $3.7bn of institutional money voting with its capital in a single quarter. That is also not contestable. The two sides are not arguing about facts — they are arguing about which set of facts is load-bearing. The model splits the difference and lands at roughly the current price, which is the honest answer.
We rate PGY fairly valued, with the distribution of outcomes skewed wide rather than up. At $17.62 the market has already discounted most of the bear case and none of the bull case. Note also that the shares have rallied roughly 66% off the April low of $10.40 and are up 37% over three months — a good part of the re-rating has already happened. This is a barbelled payoff, not a mispriced compounder, and it should be sized as one.
One. Two consecutive quarters of FRLPC stabilising or improving — the highest-leverage single item, and the third-quarter print is the first test. Two. A Fitch upgrade following the Positive outlook, lowering cost of capital and validating the balance sheet against the short thesis. Three. Named regional bank partners going live, proving the model works with true deposit-funded institutions. Four. Credit marks growing slower than volume for two or three consecutive quarters. Five. A shift to GAAP-first reporting, or an authorised buyback — either would close the gap between what management says about the balance sheet and what it does with it.
What is genuinely uncertain, and what nobody in this debate knows: whether FRLPC stabilises above 4% or continues drifting; whether the 2026 auto vintages perform as the 2025 ones did; whether operating leverage survives the regional-bank onboarding wave; and whether the Fed’s next move is a cut or a hike. On the last of those, market commentary in late July was openly debating a hike, which is not the assumption embedded in most sell-side models of this business.
What should be verified next, in priority order: monthly ABS trustee and remittance reports for the 2025–26 vintages, which are independent, third-party and available before any of it reaches the profit and loss account; the fair-value hierarchy footnote and any critical audit matter in the FY26 annual report; the third-quarter FRLPC print; the proxy, to establish whether executive incentives are tied to GAAP or adjusted metrics; and partner concentration disclosure in the next annual filing.
Imagine you walk into a car dealership and apply for a loan. The lender looks at your file and either says no, or offers you terms that don’t quite work. In the old world that’s the end of it — you walk out and the dealer loses the sale. Pagaya is the company sitting invisibly behind that lender’s software, saying “wait, we’ll take that one.” It runs its own models over the application, and if it likes what it sees, it tells the lender to make you an offer — often a better-structured one. Pagaya then finds big institutional investors to actually put up the money. You never hear the name Pagaya. You just get a loan.
A fee on every loan it places. That is genuinely it — think of it as a very sophisticated matchmaker, lenders on one side and big money managers on the other, taking a slice of everything it introduces. There is one wrinkle that matters. To convince investors to buy these loan bundles, Pagaya has to keep a piece for itself — skin in the game. That piece is about $1 billion sitting on its balance sheet right now, and when those loans go bad, Pagaya eats some of the loss. So it isn’t purely a fee business. There is real credit risk in there.
Because the numbers just got a lot better. On 30 July the company reported its sixth straight profitable quarter, with loan volume up a third and profit up 172%, then raised its forecast for the year by about a quarter. And here is the part that excites people: its costs haven’t gone up in eighteen months. Same team, same technology, a third more business flowing through. Almost every extra dollar falls to the bottom line.
The cut they take is shrinking — four quarters in a row now. They’re moving more loans but keeping less of each one. They borrow expensively. Competitors like Happen (the old LendingClub) and SoFi are actual banks that fund loans with customer deposits, which is much cheaper. With rates still high, that’s a real disadvantage. If the economy turns, this hurts more than most — their borrowers are ordinary Americans with credit scores around 680, and if unemployment rises, Pagaya’s own billion-dollar slice takes the hit. And a short-seller went after them in February 2025 claiming the accounting hid losses. Nothing was ever proven, but the cloud hasn’t fully lifted.
Pagaya reports “adjusted” earnings of $1.07 a share alongside real earnings of $0.49. The difference is mostly loan losses, which it excludes. For a company whose whole job is picking which loans to make, loan losses aren’t an unusual one-off — they’re the cost of doing business. Use the $0.49. That habit will serve you well across every company you ever look at: when a company shows you two profit numbers, understand exactly what sits between them before you decide which to believe.
If you remember one thing: Pagaya is a good operating business attached to a risky balance sheet, and the company’s preferred way of presenting itself emphasises the first and downplays the second. Nothing here is hidden — the reconciliations are all published — but you have to go and read them. That is not a reason to avoid the stock. It is a reason to size any position as what it is: a volatile, cyclical, leveraged bet on the American consumer, not a steady compounder.
Our standing rule is that every figure traces to a primary source and every derived input is labelled as derived. This report leans heavily on a single quarterly release and its call, so it is worth being explicit about which category each load-bearing number falls into.
| Status | What it covers |
|---|---|
| Verified from Pagaya | All 2Q26, 1H26 and 2Q25 income statement, balance sheet, cash flow and non-GAAP reconciliation figures; network volume, FRLPC and FRLPC percentage across 2Q25–2Q26; core operating expenses, share-based compensation, interest expense and losses on investments in loans and securities; 3Q26 and FY26 guidance; the $3.7bn funding quarter, six ABS transactions and 174 institutional investors; conversion rate, application volume and borrower FICO, income, homeownership and DTI statistics; the POS partner roll-off; management commentary and analyst questions from the 30 July call; the Fitch outlook revision of 15 July 2026; the June 2026 open-market purchase by the CEO. |
| Our estimate or judgement | All LTM aggregates, enterprise value and every valuation multiple; every input and output of the FCFE model, including the retention charge, the cost of equity, both sensitivity grids, the reverse test and the $8–$11 stress zone; the $14–$22 fair-value range and the $16.50–$17.00 central estimate; all scores in Sections 06, 07 and 08; the reading of the closing remarks as rebuttal; and the characterisation of the 224% reported beat as not like-for-like. |
| Still open | The 2Q26 figures are drawn from the earnings release, shareholder letter and call. Prior-period figures for 3Q25 and 4Q25 are as disclosed in company press releases and are lightly rounded, and 1Q26 is derived by subtracting 2Q26 from the six-month figures in the 2Q26 release. Peer market capitalisations are third-party aggregator estimates of varying dates in July 2026 and are indicative only. The fair-value hierarchy footnote and any critical audit matter on the investment portfolio will be reconciled against the FY26 annual report when it publishes. |