SoFi Technologies, Inc. (NASDAQ: SOFI) — The 30% Grower the Market Gave Up On, Priced Halfway to a Moat It Hasn’t Earned

Report date: 2026-06-11 | Price: $15.87 (2026-06-10 close) | Market cap: ~$20.4B | Diluted shares: ~1.283B | 52-week range: $13.97–$32.73

⚡ Claude’s Take

This block is the author’s own subjective opinion and general information only — not investment advice and not a recommendation to buy or sell any security. The analysis that follows takes no position and carries no price target; that discipline is intact everywhere except inside this fenced block. Do your own research.

Verdict: ACCUMULATE ON WEAKNESS — a speculative, position-sized BUY here in the mid-teens. Conviction: medium. Attractive entry zone ~$13–16 (where it trades, near the 52-week low); fair-value zone ~$19–24 on base-case execution; trim into the high-$20s+; the bull case supports ~$38–45 over three years if returns converge.

SoFi is a genuine ~30%-revenue grower whose stock has been halved (from $32.73 to $15.87) even as its fundamentals accelerated — Q1-2026 revenue grew 42% and normalized EPS more than doubled in 2025. The market gave up on it for three defensible reasons: ~16%/year share dilution, a return on tangible equity of only ~9% against management’s 20–30% aspiration, and earnings that are ~96% mark-to-model on an aggressive ~4.5% discount rate. Those are real concerns and the body below does not flinch from them. But at ~2.2x tangible book and a PEG below 0.7 on a 38–42% targeted EPS CAGR, the price is no longer underwriting the dream — it is underwriting half the gap between today’s 9% ROTCE and a mid-teens steady state. That is a more forgiving setup than the consensus “overvalued fintech” framing admits. The thing I weight most heavily: CEO Anthony Noto has been buying stock in the open market with his own cash ($15.73–$17.88) into the very weakness the market is selling — the single cleanest insider conviction signal available, and it points the same way the de-rated multiple does.

Framing: this is a quality-grower-at-a-de-rated-price with unproven returns — not a deep-value bank and not a proven compounder, but the contested transit between the two (the “Ally zone” at ~1x book vs. the “Robinhood zone” at ~3x+ book). The 3-year payoff skew is symmetric-to-favorable (bear ~−30%, base ~+25–50%, bull ~+150%), and a crowded ~13.5%-of-float short position adds upside squeeze asymmetry on any positive surprise. But size it as the higher-risk, mark-sensitive, dilution-prone name it is.

What flips me bullish (single trigger): ROTCE pushing decisively through ~15% and holding across a rate-cut cycle, with fee/noninterest income proving durable through a quarter of widening credit spreads. What flips me bearish (single trigger): a consumer-credit downturn that simultaneously spikes personal-loan losses and marks the fair-value book down — or another large, dilutive equity raise at a depressed price. Tag: “Paid to wait on a moat it hasn’t earned yet — but the CEO is buying and the dream is on sale.”

1. Executive Summary

SoFi Technologies is a vertically integrated, digital-native consumer-finance company that became a national bank holding company in February 2022 (via the ~$22.3M Golden Pacific Bancorp acquisition → SoFi Bank, N.A.). It runs a “one-stop-shop” financial app — borrow, save, spend, invest, protect — across three reportable segments: Lending (personal, student, home loans), Technology Platform (Galileo card-processing + Technisys core banking), and Financial Services (deposits, brokerage, credit card, interchange, and the capital-light Loan Platform Business). FY2025 total net revenue was $3,613M, up 35%, with reported GAAP net income of $481M.

The business is executing. Members grew to 13.7M (+35%) and products to 20.2M (+37%); Q1-2026 added a record 1.1M members in a single quarter and revenue accelerated to +42% YoY. Deposits have compounded from $26.0B at end-2024 to $40.2B (Q1-2026), giving SoFi a genuine cost-of-funds edge (~190bps cheaper than warehouse lines). Crucially, the apparent FY2024→FY2025 net-income “decline” ($499M → $481M) is an accounting illusion: FY2024 GAAP NI included a $271M one-off deferred-tax-valuation-allowance and restructuring benefit. Strip it out and normalized net income path is $(54)M → $227M → $481M — a doubling in 2025 — with normalized adjusted diluted EPS climbing $(0.10) → $0.15 → $0.39. Management guides FY2026 to ~$4,655M revenue (+30%), ~$825M adjusted net income, and ~$0.60 adjusted EPS, with a medium-term 30%+ revenue / 38–42% EPS CAGR.

But the quality of those earnings and the durability of the moat are genuinely contested. Three issues anchor the bear case and run through every section below. First, ~96% of SoFi’s ~$38B loan book is carried at fair value (mark-to-model), not amortized cost — the allowance for credit losses is a trivial $51M — and 2025 noninterest income was flattered by an ~84bp decline in the discount-rate assumption. The same mechanism reverses when credit spreads widen, making earnings pro-cyclical. Second, headline charge-offs are mechanically suppressed by ~$360M of delinquent-loan sales in 2025, so the true loss content is understated. Third, the supposedly high-margin “SaaS moat” — the Technology Platform — is in outright decline (Q1-2026 revenue −27%, contribution profit −61%) after anchor client Chime built its own processor (ChimeCore) and migrated off Galileo, a live demonstration that the platform’s best clients graduate away.

The central tension is returns versus price. SoFi earns a ~9% ROTCE today against management’s 20–30% target, yet trades at ~2.2x tangible book — a multiple the bank cross-section reserves for mid-teens ROTCE franchises. The embedded-expectations math says the market is underwriting roughly half the journey from 9% to the target. On its own ten-year history the stock is mid-cheap (P/E at the 22nd percentile, composite valuation at the 42nd percentile) after a ~50% peak-to-trough de-rating. The competitive verdict is that SoFi is a well-run, fast-growing, charter-advantaged consumer-fintech bank competing largely on price, marketing spend, and product breadth — not yet a structurally protected franchise. Whether the charter + integrated tech stack + scale-in-progress mature into a durable low-cost-producer moat is the entire investment question, and at a 9% ROTCE it remains an unproven bet rather than a demonstrated fact.

This report takes no position and sets no price target. It frames the valuation strictly as embedded expectations and scenarios.

2. Business Overview

SoFi makes money in three distinct ways, and the mix is shifting in an economically important direction.

Segment revenue (GAAP total net revenue, $M):

(Segment figures sum above consolidated net revenue because of inter-segment eliminations and corporate/other; directionally they are the right way to read the business. Source: FY2025 10-K segment footnote; Q1-2026 10-Q.)

The single most important structural fact in the table is the de-concentration away from balance-sheet lending: Lending fell from 65% of revenue in FY2023 to ~51% in FY2025, while Financial Services nearly doubled (+88% in FY2025). SoFi is deliberately morphing from a spread lender into a hybrid that earns an increasing share of revenue from fees (FACT).

Lending segment. The historical core. Personal loans dominate (weighted-average origination FICO ~749 — genuinely prime), alongside student-loan refinancing (SoFi’s founding product), in-school loans, and home loans (capacity added via the 2023 Wyndham Capital Mortgage acquisition). The historical model is a gain-on-sale + balance-sheet hybrid: SoFi originates, earns net interest income while it holds the loan, then sells whole loans to large financial institutions or securitizes them, retaining servicing. The recurring component is NII on held loans plus servicing; the transactional component is gain-on-sale and fair-value marks. Lending remains the largest contribution-profit engine — ~$1,017M of contribution profit at a 55% margin in FY2025 (FACT, 10-K) — but that margin compressed from ~60% as direct advertising rose ~50% to ~$328M, a reminder that loan growth is bought, not free.

Technology Platform. Galileo (an API-based card-issuing and authorization-processing platform, acquired 2020 for ~$1.2B) plus Technisys (a cloud-native core-banking system, acquired 2022 for ~$1.1B all-stock). It sells infrastructure to banks, fintechs, governments, and non-financial brands across North America and Latin America under multi-year, fee-based contracts — in theory the most “recurring,” highest-quality, capital-light revenue SoFi has. In practice it is the problem child. Total Galileo accounts fell to 128.5M at year-end 2025 from 167.7M (−23%) because a large client — Chime — built its own in-house processor (ChimeCore) and migrated off Galileo. Segment revenue still grew +14% in FY2025 on pricing/mix and new logos (e.g., Banco Nación in Argentina), but by Q1-2026 the Chime departure overwhelmed that: revenue −27% and contribution profit −61% YoY. This is the textbook banking-as-a-service failure mode — your biggest, most successful clients eventually graduate in-house — and it is the clearest live evidence that the “platform moat” is weaker than the marketing implies (INTERPRETATION, grounded in 10-K account disclosures).

Financial Services. The growth engine and the strategic pivot. It houses SoFi Money (checking/savings — the deposit franchise), SoFi Invest (brokerage, robo-advisory, options, fractional shares, IPO access, alternatives), the SoFi Credit Card, SoFi Relay (personal-finance management), Lantern (a marketplace for non-qualifying applicants), SoFi Protect/Travel/At Work, the relaunched SoFi Crypto and SoFiUSD stablecoin (late 2025), and — most importantly — the Loan Platform Business (LPB). Financial Services swung from a contribution loss of $(0.3)M in 2023 to +$793M of contribution profit in 2025 (a ~67% incremental contribution margin), making it the #2 profit pool. Revenue here is interchange (card spend +54% YoY), deposit-funded NII, brokerage and referral fees, and LPB fees.

The Loan Platform Business — the defining recent change. Launched in Q3-2024, LPB is a capital-light, originate-for-a-fee model: SoFi originates personal loans on behalf of third-party partners, to those partners’ credit specifications, and/or refers pre-qualified borrowers — earning a fee per loan without retaining the credit risk on its own balance sheet. In FY2025 LPB generated ~$576M of loan-platform fees on ~$11.0B of personal loans originated for third parties (up from $2.1B of volume and $142M of fees in FY2024 — roughly a 5x jump), an implied blended take of ~5.2% of unpaid principal. In March 2026 SoFi announced more than $3.6B of new LPB delivery commitments across three partnerships (a global bank, an insurer, and a top-five private-equity firm), on top of existing Fortress/Blue Owl arrangements. LPB is the single biggest reason Financial Services profit exploded, and it is genuinely capital-efficient — but it is upfront, gain-on-sale-style recognition, it is counterparty-concentrated, and it depends on third-party (largely private-credit) appetite for the paper. The 10-K names this concentration as a risk factor.

The “Financial Services Productivity Loop.” Management’s flywheel thesis is: acquire a member (often via a loan), retain servicing for ongoing touchpoints, then cross-sell additional products at near-zero incremental customer-acquisition cost, lowering the blended CAC per product over time. The data partially supports it — products are growing faster than members, ~43% of new products in Q1-2026 were opened by existing members (up from ~32% in late 2023), Financial Services annualized revenue-per-product rose ~29% YoY to ~$104 (Q4-2025), and ~90% of new SoFi Plus subscribers are existing members. But products-per-member is still only ~1.5x. This is a real CAC-amortization engine, not a network effect (the distinction matters for the moat assessment in the relevant section): one member joining does not make the app better for another.

Verdict (the relevant section): A coherent, vertically integrated consumer-finance platform with a deliberate and economically sensible shift toward capital-light fee revenue. The business is real, diversified, and growing fast. The caution is that breadth is a feature, not a barrier — and the one segment sold as a durable, recurring SaaS moat (Technology Platform) is the one currently shrinking.

3. Industry Dynamics

SoFi competes across at least five largely separate, mostly commodity markets, and dominates none of them. That is the starting point for any honest moat discussion.

Unsecured personal lending. The US market is roughly $250B+ of balances across ~30M loans, with fintechs taking roughly half of new account balances. SoFi’s own 10-K describes the prime market as “highly fragmented and competitive” — a candid admission. There are essentially no barriers to entry; competitors include LendingClub, Upstart, Marcus (Goldman), credit unions, and the megabanks. SoFi underwrites a relatively narrow prime credit box (origination FICO ~749) and has built meaningful share within it, but “share of a narrow box in a fragmented market” is not market power.

Deposits / neobanking. Pure rate competition. The benchmark all-digital bank is Ally Financial: ~$143B of deposits, a ~3.4% NIM, and a retail deposit rate near 3.75%. Chime (≈9.5M active members, ~$2.2B FY2025 revenue, a debit-interchange model, no bank charter) and Cash App/Block (~58M monthly actives) are larger than SoFi by user count. And JPMorgan and Bank of America dwarf all of them on funding cost, scale, and fixed-cost absorption. SoFi’s deposit base ($40.2B) is growing fast and is ~97% direct-deposit-linked, but it is hot, top-of-market-APY money competing on price.

Brokerage. Robinhood, Schwab, Fidelity, and Webull set the terms; SoFi Invest is a small, introductory player using the brokerage primarily as a cross-sell anchor rather than a profit center.

Banking-as-a-service (Galileo’s market). A ~$35B market growing mid-teens, but a structurally low-margin processing race with negative selection: Marqeta holds an estimated ~15–18% of API card-issuing; Unit, Green Dot, Stripe Issuing, and others compete on price. The defining industry feature — that your largest, most successful clients eventually build in-house (exactly what Chime did with ChimeCore) — caps the achievable profit pool and undermines the recurring-revenue thesis.

Regulatory landscape — one genuine structural asset amid rising burden. The national bank charter is the single real barrier SoFi possesses. A charter is scarce and hard to obtain; it gives SoFi cheap deposit funding (versus Chime’s warehouse/sponsor-bank dependency) and lets it keep net interest income rather than renting a balance sheet. This is the most durable structural feature SoFi has. But it is shared — Ally, Marcus, and every chartered digital competitor have the same advantage; it is the charter’s edge, not SoFi’s proprietary one. Offsetting it, the compliance surface is widening: the CFPB began supervising SoFi Bank in January 2024; Regulation W constrains the Galileo/Technisys non-bank subsidiaries to deal with SoFi Bank at arm’s length; and the 2025–2026 crypto/stablecoin relaunch (SoFiUSD, tokenized deposits) opens an entirely new, still-unsettled regulatory front.

The student-refi cycle — a real but reversible, policy-driven tailwind. SoFi’s founding product is private student-loan refinancing, which was gutted for years by the federal moratorium and forbearance. That has reversed: federal forbearance interest resumed August 1, 2025 for ~7.6M borrowers, and the SAVE plan was struck down by the 8th Circuit on March 10, 2026. Both push creditworthy federal borrowers toward private refinancing — SoFi’s product — and Q1-2026 student-refi originations hit a record $2.6B (+2.2x YoY). This is a genuine multi-year tailwind, but it is policy-driven and could reverse with the political cycle; it is not a structural moat.

Marathon capital-cycle read. Consumer fintech sits in the mid-to-late boom phase of the capital cycle: abundant capital, a fresh wave of IPOs (Chime listed in 2025), heavy marketing spend, and growth being extrapolated into rich multiples. In Marathon’s framework, high returns attracting capital is precisely the condition that erodes future returns. This is not a consolidating, supply-disciplined industry; it is fragmenting and well-funded. That argues against paying a premium for any single participant absent a demonstrated, durable advantage.

Verdict (the relevant section): structurally mediocre-to-poor. SoFi operates in commodity markets with low entry barriers, intense and well-capitalized competition, and a capital cycle in its late-boom phase. The one genuine structural asset — the bank charter — is shared with several competitors. A good operator can earn good returns here for a time; the industry structure does not grant durable excess returns.

4. Competitive Position

Apply the Greenwald test directly: a genuine competitive advantage shows up as sustained excess returns on capital. SoFi’s ROTCE was ~9% in Q4-2025 against management’s 20–30% target. Greenwald’s profitability screen for a real moat is sustained after-tax ROIC/ROE in the 15–25%+ range. SoFi fails that test today. The moat is a forecast, not a demonstrated fact — and that is the most important sentence in this report.

Naming the candidate mechanisms one by one:

• Cost advantage (deposit funding): REAL but SHARED. Deposits ($40.2B) genuinely lower SoFi’s funding cost. The 10-K quantifies it: total interest-bearing deposit cost was 3.40% in FY2025 versus warehouse facilities at 5.31% — roughly 190bps cheaper — and the deposit build let SoFi cut wholesale debt from ~$3.1B to ~$1.8B. This is a true edge over charter-less neobanks (Chime, Cash App). But it is the charter’s advantage, equally available to Ally, Marcus, and every chartered digital bank. It is not proprietary.

• Switching costs (direct-deposit stickiness): PARTIAL / largely THEORETICAL. Direct-deposit members are demonstrably stickier and more profitable, and retained loan servicing creates recurring touchpoints. But consumer financial switching costs are inherently low — moving a checking account is a weekend’s work — and products-per-member of ~1.5x shows the relationship is not yet deep enough to lock members in. There is emerging stickiness, not a moat-grade lock-in.

• Network effects: ABSENT. Pressure-tested hard, there is no two-sided network. One member joining does not improve the product for another. The “Productivity Loop” is CAC amortization dressed up as a flywheel. It should not be credited as a network effect, and any bull case that leans on “network effects” is mislabeling the mechanism.

• Economies of scale + captivity (Greenwald’s strongest combination): EMERGING at best. SoFi remains sub-scale versus JPM/BAC/Ally on funding cost and fixed-cost absorption. The vertically integrated Galileo/Technisys stack is a cost-efficiency play — owning the rails rather than renting them — and could eventually lower unit costs enough to matter. But in the BaaS market that stack carries negative captivity (clients graduate, as Chime did), and SoFi is not yet at the scale where unit-cost advantages compound decisively.

• Brand/intangibles: real awareness, not a barrier. Heavy marketing (stadium naming rights, national TV) built genuine consumer awareness. But brand without an underlying barrier earns average returns, and SoFi spends ~30% of revenue on sales and marketing to sustain that awareness — the opposite of a brand that pulls customers in for free.

Head-to-head: Chime is larger in active users (~9.5M) but charter-less and interchange-dependent — and it just proved it can build away from SoFi’s own technology. Ally is far larger in deposits ($143B vs $40B) with an established ~3.4% NIM and a higher current ROTCE than SoFi, yet trades at ~1x tangible book because it carries no growth premium. Cash App has ~58M monthly actives. SoFi’s genuine differentiator is breadth — it is the only one of these that lends, banks, brokers, and sells the underlying tech infrastructure, all under one charter. But breadth is replicable piece by piece; it is a feature set, not a barrier to entry.

Verdict (the relevant section): a thin, emerging moat — not a protected franchise. SoFi is best characterized as a well-run, fast-growing, vertically integrated consumer-fintech bank competing largely on price, marketing spend, and product breadth. The charter + integrated stack + scale-in-progress could mature into a durable low-cost-producer advantage — but only if returns climb to and hold the mid-teens-plus level, and at 9% ROTCE that is an unproven bet. The LPB pivot, while financially smart, actually reduces the balance-sheet scale that a lending moat would rest on, trading it for high-margin but counterparty-concentrated fee income. If the moat were real today, it would already show up in the returns. It does not yet.

5. Growth History and Forward Opportunities

Member and product trajectory (FACT, company disclosures/transcripts):

Members have grown more than 20x from ~650K in 2018, and the ~35% YoY cadence has been remarkably steady for years — a rare consistency. Q1-2026 added a record 1.1M members and 1.8M products in one quarter. The growth is overwhelmingly organic (the cross-buy flywheel), not acquired — the only recent M&A is the tiny Composer AI-portfolio tuck-in.

Monetization is improving alongside the unit growth (FACT): cross-buy is rising (43% of new products opened by existing members in Q1-2026, up from ~32% in late 2023), Financial Services annualized revenue-per-product climbed ~29% YoY to ~$104 (Q4-2025), and ~70% of SoFi Invest sign-ups are existing members. This is the right shape — more products per member at a rising revenue per product — and is the strongest evidence for the flywheel narrative.

The strategic pivot to capital-light / fee-based revenue is the central growth story. Management reports fee-based revenue of ~$443M in Q4-2025 (+53% YoY), ~44% of revenue now fee-based and ~50% non-lending on a trailing basis, and 85% of products non-lending. In Q1-2026 it introduced a “cash revenue” framing (~$1.1B = ~$690M cash NII + ~$390M non-interest cash) explicitly to counter fair-value-accounting skeptics — a framing device that itself should be validated against the 10-Q rather than taken at face value (OPEN QUESTION). The three capital-light pillars are LPB (origination fees, no retained credit risk), interchange (~$25B annualized card spend, +54% YoY), and the Technology Platform.

Forward opportunities management is selling. (1) LPB expansion — ~$11B of third-party originations in FY2025, ~$576M of fees, and $3.6B+ of new partner commitments announced March 2026; notably, Q1-2026 LPB volume was deliberately throttled sequentially to retain NII-generating loans on the now over-capitalized balance sheet, which complicates the simple “maximize fee income” narrative. (2) A new-product stack with near-zero current revenue — the SoFi Crypto relaunch (Nov-2025, ~30 coins, first national bank to offer it natively), the SoFiUSD stablecoin (Dec-2025, fully reserved in a Fed master account, with a Mastercard 24/7-settlement deal), Big Business Banking (Apr-2026), the SoFi Plus paid subscription (relaunched at $10/month, ~160K paying members by mid-May-2026), and a Smart Card. Management frames each as “easily a $100M revenue business,” which is aspiration, not evidence. (3) Technology Platform new logos (13 new revenue-generating clients in Q1-2026, a top-3 US telecom launching in 2026), meant to backfill the Chime loss. (4) Deposit and rate-cut optionality — student refi (+2.2x YoY) and home loans (+2.4x YoY) are growing even with zero 2026 rate cuts now assumed, so any cut is upside.

Guidance. FY2026 (reaffirmed in Q1-2026 even after removing all rate-cut assumptions): revenue ~$4,655M (+30%), adjusted EBITDA ~$1.6B (~34% margin), adjusted net income ~$825M, adjusted EPS ~$0.60, members +30%+. Medium-term (2025→2028): revenue CAGR ≥30%, EPS CAGR 38–42%, and the long-dated 20–30% ROTCE aspiration that management pointedly declined to put a date on.

Verdict (the relevant section): high-quality growth, with two caveats. Growth is organic, accelerating (unusual at this scale), monetizing better, and mix-shifting toward durable capital-light fees — that is genuinely high quality. The caveats: it is partly rate-cycle-flattered (a 5.85% NIM and a benign credit environment do a lot of work), and the forward story increasingly rests on unproven crypto/stablecoin/Big-Business-Banking products with no disclosed revenue yet. The base of growth is high-quality; the forward incremental case leans on optionality that has not yet shown up in the P&L.

6. Financial Quality

Verdict up front: the economics genuinely improve with scale, but reported profitability is real and lower-quality than the headline — mark-to-model, rate-cycle-aided, and propped by upfront fee and gain-on-sale recognition.

Segment economics (contribution profit, $M; CP = segment revenue less directly attributable expense and, for FS, credit provision — corporate SBC/D&A sit below the line):

All three segments now generate contribution profit. The operating-leverage story is real: Financial Services swung from a $(0.3)M loss in 2023 to +$793M in 2025 (a ~67% incremental margin). But two cautions jump off the table — Financial Services’ “profit” is heavily LPB-fee-driven rather than organic spread/interchange, and the Technology Platform, the segment meant to embody durable capital-light SaaS economics, is contracting (Q1-2026 CP −61%).

The fair-value crux — the heart of the quality-of-earnings concern. Of ~$38.0B of loans at year-end 2025, roughly $36.5B (~96%) is carried at fair value ($22.9B held-for-sale + $13.7B held-for-investment), with only ~$1.5B at amortized cost. The consequence is that the allowance for credit losses is just $50.9M — it covers only the ~4%-of-book amortized-cost sliver. Credit risk on the other 96% does not flow through a reserve; it flows through mark-to-model noninterest income. The personal-loan fair-value model used a weighted-average annual default assumption of ~4.46% and a discount rate of ~4.46% at year-end 2025, down from ~5.3% a year earlier — an ~84bp decline. A ~4.5% discount rate on unsecured prime consumer paper is aggressive (a thin risk spread), and the 10-K explicitly attributes a chunk of FY2025’s higher fair-value gains to “a larger decrease in discount-rate assumptions” (e.g., +$292M on student loans alone). In plain terms: falling discount rates lifted 2025 revenue, and the same mechanism reverses — marking the book down through the income statement — when credit spreads widen. This is not fraud; it is pro-cyclical, lower-quality earnings, and it means 2.2x this tangible book is richer than 2.2x a clean accrual bank’s book, because reported equity embeds unrealized mark-to-model gains.

Credit quality — improving on the surface, with a suppression mechanism underneath. Personal-loan annualized net charge-offs were ~2.80% in Q4-2025, below the 4.46% modeled default (a model cushion exists), but rose to ~3.03% in Q1-2026. The important QoE flag: SoFi sold ~$360M of delinquent personal loans during 2025 (the 10-K confirms this verbatim), selling late-stage delinquencies before they charge off — which mechanically lowers the reported NCO. Headline charge-offs therefore understate the true loss content. The direction is best described as normalizing-to-modestly-deteriorating, partly masked by sales — not the unambiguous “improving credit” management narrates.

Net interest margin and the deposit edge. Consolidated NIM was ~5.85% in FY2025 (vs ~5.80% FY2024), high because of the high-yield unsecured consumer asset base. The deposit funding advantage is real and quantified (3.40% deposit cost vs 5.31% warehouse — ~190bps), and is the single most defensible economic feature SoFi has — though, again, shared with every chartered digital bank.

Capital and dilution. H2-2025 raised ~$3.2B of common equity — July 31 (82.7M shares at $20.85 ≈ $1.7B net) and December 8 (54.5M shares at $27.50 ≈ $1.5B net) — driving stockholders’ equity from $6.5B (YE2024) to $10.5B (YE2025) to $10.8B (Q1-2026). Capital ratios tell a nuanced story: holdco CET1 jumped to ~22.8% on the raise, but SoFi Bank CET1 actually fell from 17.3% to 16.4% (YE2025) to ~15.3% (Q1-2026) as risk-weighted assets outgrew the capital downstreamed into the bank. The holdco cushion is real and downstreamable, but bank-level capital is on a declining path — a metric to watch. Goodwill ($1,393M, Galileo/Technisys) plus intangibles (~$232M) net against the $10.5B of equity to leave tangible equity of ~$8.9B and tangible book value per share of ~$7.2 (YE2025, management-disclosed) — so at $15.87 the stock trades at ~2.0x book and ~2.2x tangible book.

Dilution is the chronic wound. Shares grew from ~1,095M (YE2024) to ~1,271M (YE2025) — roughly 16% in a single year — and from ~800M to ~1.28B since 2021, driven by the equity raises, SBC, and convertible-note conversions. Stock-based compensation was $262M in FY2025, but as a percentage of revenue it has improved to 7.3% (from 12.8% in FY2023) — a genuine positive that partially offsets the share-count narrative.

The normalization that fixes the headline. The reported FY2024→FY2025 net-income “decline” ($498.7M → $481.3M) is an artifact: FY2024 GAAP NI included a $258.4M deferred-tax-valuation-allowance release plus a $13.0M restructuring tax benefit (~$271M total). Stripping the one-offs:

Normalized net income +112% in FY2025; normalized EPS path $(0.10) → $0.15 → $0.39. Adjusted EBITDA likewise rose $432M → $666M → $1,054M (+58% in FY2025, ~29% margin), though EBITDA adds back ~$0.5B of real D&A and SBC, so normalized net income is the more honest measure. The right cash-flow analogs for a bank are pre-provision net revenue (≈ adjusted EBITDA ~$1,054M) and ROTCE (~9%, Q4-2025) — free cash flow is not meaningful when balance-sheet loan growth dominates operating cash flow.

Verdict (the relevant section): economics improve with scale, and normalized earnings doubled — but the quality is mark-driven and pro-cyclical. Four caveats temper the headline: (1) ~96% of loans marked to model at an aggressive ~4.5% discount rate, so 2025 income was lifted by falling discount rates; (2) headline NCOs suppressed by ~$360M of delinquent-loan sales; (3) LPB fees front-loaded and counterparty-dependent; (4) the high-margin Technology Platform shrinking. The clean positives — the deposit funding edge and the normalized-earnings doubling the GAAP optics hide — are real. Net: investors are paying ~2.2x tangible book for a ~9%-ROTCE business on the bet that returns roughly double toward target.

7. Capital Allocation

M&A scorecard — a below-average acquirer historically, saved by one cheap, transformative deal.

In Q3-2025 SoFi aggregated the Galileo and Technisys reporting units into a single “Technology Platform” unit (~$1.3B goodwill) — which, fairly noted, also makes future segment-level impairment harder to isolate. The net read on CEO Anthony Noto’s M&A record: below average historically (Technisys is a genuine black mark, Galileo has underdelivered), redeemed almost entirely by the ~$22M charter purchase.

Capital raises and dilution. The defining negative is chronic share growth — ~16% in FY2025 alone. But the 2025 vintage of raises was opportunistically timed: ~$3.2B issued near 52-week highs ($20.85 and $27.50, versus the current $15.87), explicitly to pay down higher-cost warehouse debt and fund growth flexibility. That is a defensible, even shrewd, use of an elevated stock — a clear improvement over the dilutive 2023–2024 convertible-for-equity swaps (e.g., ~92.7M shares issued to retire 2026 notes). Management’s defense that tangible book value per share rose 57% YoY to ~$7.21 despite the raises is accurate — but TBV-accretive is not the same as EPS-non-dilutive, and the per-share earnings drag from ~16% more shares is real.

Convertible structure. The 0% 2026 notes (~$428M remaining) carry a $22.41 conversion price and a $32.02 capped-call cap; the 1.25% 2029 notes ($862.5M) convert at $9.45 with a $14.54 cap and became convertible in Q1-2026 — SoFi intends to cash-settle principal and issue shares only on the excess, limiting dilution but creating a future cash claim. Capped calls (cost ~$200M across both) partially neutralize dilution within the strike-to-cap bands.

No dividend, no buyback — correct here. For a growing, deposit-funded bank holding company that must downstream capital into SoFi Bank to support loan growth and meet regulatory minimums, retaining 100% of capital is the right call. A buyback while diluting 16%/year would be incoherent, and SoFi is not doing that.

Spend intensity. Sales & marketing was ~$1,095M in FY2025 — 30.3% of revenue and the single largest opex line and margin governor; technology & product development ~$648M (~18%); SBC $262M (7.3%, and falling as a share of revenue). Growth here is bought, and the marketing line is the lever to watch if member growth ever needs defending.

Insider signal — the one genuine conviction tell, and it is bullish. Against a backdrop of routine grants, option exercises, and planned 10b5-1 sales by other insiders, CEO Anthony Noto made repeated discretionary open-market purchases (SEC code P, not 10b5-1) into the 2026 share-price weakness: ~56,000 shares at $17.88 (Mar-2), ~28,900 at $17.32 (Mar-17), ~15,878 at $15.73 (May-8), and ~15,545 at $16.00 (May-11). This continues a multi-year serial-buying pattern since 2022. Noto’s beneficial ownership is ~24.0M shares (~1.9%). Per the relevant section, discretionary open-market CEO buying is the rarest and most credible insider signal — and here it aligns precisely with the de-rated multiple.

Compensation and governance. Noto’s FY2025 total pay was ~$30.3M (50% PSU / 50% RSU, plus a ~$744K personal-aircraft perk). Bonus metrics are reasonable and capital-aware (Adjusted Net Revenue 35%, Adjusted EBITDA 35%, Return on Tangible Equity 15%, New Members 15%); the PSU metric is absolute tangible-book-value growth with a relative-TSR modifier and a hard forfeiture trigger if the total risk-based capital ratio falls below 10.5%. The flag: 2025 say-on-pay passed with only ~76% support amid explicit investor pushback on the magnitude of CEO pay — the company responded by tightening the 2026 plan (raising the rTSR target to the 55th percentile and the target-payout threshold to 100% of target).

Verdict (the relevant section): mixed but improving — not yet proven. Against management: ~16%/year chronic dilution, the Technisys misallocation, and rich, softly-approved CEO pay. For management: the transformative charter purchase, opportunistically well-timed 2025 raises, declining SBC intensity, sensible capital retention, comp tied to reasonable capital metrics, and — most tellingly — the CEO repeatedly buying stock with his own money. The trajectory tilts positive even though the history is below average.

8. Changes and Headwinds — Last ~2 Years

Strategic shifts (mostly thesis-strengthening):

• The LPB launch and scaling (Q3-2024 onward) turned SoFi from a pure spread lender into a hybrid fee model — ~$576M of capital-light fees in FY2025 and $3.6B+ of new partner commitments in March 2026. This is the most important structural change of the period.
• The crypto/stablecoin re-entry (2025–2026) — SoFi Crypto trading (Nov-2025), the SoFiUSD stablecoin (Dec-2025, reserved in a Fed master account with a Mastercard settlement deal), and Big Business Banking (Apr-2026) — a bet that being a national bank operating natively in both fiat and crypto is a differentiator. Promising but unproven, with new regulatory surface.
• SoFi Plus relaunched as a paid $10/month subscription (Apr-2026), an early but interesting recurring-revenue and cross-sell lever.

Regulatory / policy:

• Student-loan policy is a clear tailwind: federal forbearance interest resumed (Aug-2025) and the SAVE plan was struck down (Mar-2026), reversing years of headwind to SoFi’s founding refi product (record $2.6B student-refi originations in Q1-2026).
• CFPB supervision of SoFi Bank since January 2024 raises the compliance burden; the crypto/stablecoin push opens a fresh, unsettled regulatory front.
• Leadership is stable — Noto (CEO since 2018) remains in place with no major turnover.

Recent news/events: the AZI news tape (checked 2026-06-11) is quiet and routine — minor/neutral items on SpaceX-IPO retail access via SoFi Invest, generic bank-stock moves on a strong May jobs report (lower rate-cut odds), and a new AI coaching tool. No thesis-changing news, no short report, no controversy.

Headwinds:

• The Technology Platform is structurally impaired by the Chime/ChimeCore departure (accounts −23%; Q1-2026 revenue −27%, CP −61%; FY2026 segment guide cut to ~$325M) — the clearest evidence the “platform moat” is weaker than marketed.
• The rate environment turned less friendly — FY2026 guidance now assumes zero cuts (down from two), removing the student/home-refi accelerant, though management held the full-year guide.
• Valuation compression — the stock is down ~50% from its $32.73 high to $15.87 despite accelerating fundamentals, as the market discounts dilution, fair-value earnings quality, and the forward crypto narrative.
• Credit normalization is benign now but cycle-dependent, and the suppression of headline NCOs via delinquent-loan sales means the reported trend flatters the underlying.

Verdict (the relevant section): net thesis-strengthening on the business, but the changes concentrate the bull case into unproven optionality while the market de-rated the narrative. The student-refi reversal and the LPB diversification are real, durable-ish positives; the crypto/BBB bets and the Tech Platform weakness are the offsetting question marks. The most striking fact of the period is that the business strengthened while the stock halved.

9. Risk Analysis

The three risks that matter most. Credit (#1) and fair-value accounting (#2) are correlated and define the bear case. In a consumer recession the unsecured personal-loan loss rate spikes and credit spreads widen — so the same event raises realized losses and marks the ~$36.5B fair-value book down through noninterest income. The $50.9M allowance is not a buffer; it covers only the ~4% amortized-cost sliver, while the other 96% has no reserve, only the model. The modeled default (~4.5%) does sit above realized NCO (~2.8–3.0%), but realized NCO is itself suppressed by ~$360M of delinquent-loan sales, so the apparent cushion is partly cosmetic. This is the highest-conviction bear vector — not that SoFi is uniquely reckless, but that its earnings and equity are mark-sensitive and the disclosure makes the suppression mechanism explicit.

Valuation/sentiment (#11) is where all the others get priced. At ~2.2x tangible book and ~28x forward earnings on a 9% ROTCE, the market is underwriting the 20–30% ROTCE target. If credit, the Tech Platform, or rate-driven NIM/refi disappoint, the de-rating is amplified by the gap between price and current returns — and the 173M-share short position adds two-way violence to any surprise.

Catastrophic / total-loss assessment (honest). A total loss (equity to zero) requires SoFi Bank insolvency, which the capital stack (Bank CET1 ~15%, holdco ~22%, FDIC insurance, a prime book, no near-term debt wall) makes a remote tail, not a base case — likelihood LOW. The realistic adverse outcome is not zero-equity but permanent per-share impairment via a forced dilutive capital raise at a distressed price in a downturn. Given chronic ~16%/year dilution and a 9% ROTCE that retains capital below the cost of equity, the equity-issuance reflex — not a bank-failure event — is the genuine through-cycle wealth leak to underwrite.

10. Valuation Discussion (Embedded Expectations)

SoFi cannot be valued on a single lens — it is a chartered bank (so book value and ROTCE discipline apply) and a 30%-revenue grower (so the market reaches for P/E and PEG). The two lenses give opposite verdicts, and that contradiction is the valuation thesis.

The bank lens (P/TBV vs. ROTCE). SoFi trades at ~2.2x tangible book on a ~9% ROTCE. In the bank cross-section, P/TBV scales tightly with ROTCE — JPM ~2.85x at ~20%, BAC ~1.9x at ~14%, GS ~1.8–2.0x at ~14–16%, Citi ~1.4x at ~8%. A 9%-ROTCE bank “should” trade ~1.3–1.5x tangible book. At 2.2x, SoFi is already priced where a mid-teens-ROTCE franchise trades — i.e., the market is paying in advance for ROTCE convergence that has not happened.

The growth lens (P/E and PEG). Trailing adjusted P/E is ~40x ($0.39 FY2025 adjusted EPS); forward is ~26–28x ($0.60 guide / ~$0.58 consensus). On the medium-term 38–42% EPS-CAGR target, PEG is ~0.66–0.70 — cheap, if that CAGR is real. This is the bull’s single strongest number.

Peer comp — SoFi sits in no-man’s-land:

Ally is the cautionary anchor — what a chartered digital bank earns at maturity without a growth premium: ~1x book, ~10x earnings, at a current ROTCE that exceeds SoFi’s 9%. Robinhood is the aspirational anchor — what the market pays a fintech that actually delivers 20%+ returns. SoFi’s price implicitly bets it migrates from the Ally zone toward the Robinhood zone. On P/TBV it already trades like a mid-teens-ROTCE bank; on P/E it trades far below HOOD. It is priced as neither a mature bank nor a proven compounder, but as the transit between the two.

Embedded-expectations / reverse-DCF (the crux). Using the bank justified-multiple identity — Justified P/TBV = (ROTCE − g)/(COE − g), so implied steady-state ROTCE = P/TBV·(COE − g) + g — at P/TBV 2.2x and a cost of equity of ~11.5%:

The low-growth read says the market prices ~17–21% steady-state ROTCE — roughly double today’s 9% and just below the 20–30% target. The honest read for a 30% grower uses a higher near-term g, which lowers the implied terminal ROTCE to ~16–17%. Either way, the conclusion is the same: the market is underwriting mid-teens-to-high-teens steady-state ROTCE — clearly above today’s 9%, clearly below the aspiration. The price closes roughly half the gap to target. That is internally reasonable but leaves a thin margin of safety: if ROTCE stalls at ~10–12%, the justified P/TBV (COE 11.5%, g 5%) falls to ~1.0–1.3x — material downside from 2.2x even if book value keeps growing. Conversely, a 20–25% ROTCE, if hit and held, justifies ~2.6–3.5x+ tangible book.

Scenario analysis (illustrative zones, NOT a price target). Driven off the normalized adjusted-EPS path ($0.15 → $0.39 → $0.60 guide) and ROTCE convergence over 3–5 years × an exit multiple:

The 3-year payoff skew is symmetric-to-favorable (bear ~−30%, base ~+25–50%, bull ~+150%) — but it rests entirely on ROTCE convergence, which is unproven (9% today) and partly accounting/rate-cycle-aided. The bear case is a genuine ~30%+ drawdown in which the multiple and the earnings fall together (a bank de-rate plus normalized credit), not a soft landing.

What the market prices correctly vs. incorrectly. Correctly: the genuine 30%+ revenue growth, rising cross-buy and monetization, the deposit cost-of-funds edge, the capital-light LPB pivot’s capital efficiency, the normalized-EPS doubling the GAAP optics hide, and the scarce charter — the de-rating to mid-cheap-on-own-history is a rational response to dilution and rate risk. Potentially incorrectly (the bear’s edge): (1) mark-to-model book quality — 2.2x this tangible book is richer than 2.2x a clean accrual bank’s, because reported equity embeds unrealized FV gains lifted by falling discount rates; (2) dilution drag — a 30% revenue CAGR is only ~14% per share after ~16% annual dilution, so the cheap-looking PEG is on aggregate, not per share; (3) rate sensitivity — NIM and FV marks both move with rates, and FY2026 guidance already flipped from two cuts to zero.

No price target; no recommendation. The body’s role is to frame the embedded expectations: the price underwrites roughly half the journey from a 9% to a low-twenties ROTCE, with a thin margin of safety and a payoff skew that rewards convergence and punishes a credit-cycle reversal.

11. Variant Perception

Consensus view. A high-growth fintech bank that finally reached GAAP profitability and guides to 30% revenue / 38–42% EPS growth, but whose stock halved ($32.73 → $15.87) on dilution fatigue, rate uncertainty, and “9% ROTCE, not 25%” skepticism. Sell-side is mixed-constructive. The ~13.5% short interest (173M shares, ~2.5 days to cover) is itself a variant-perception signal — a crowded short plus a serial-buying CEO creates squeeze asymmetry on any positive surprise.

Strongest bull case. A capital-light flywheel (LPB fees +307% in FY2025, deposit-funded, capital-efficient) on top of a scarce national bank charter and a vertically integrated tech stack — a genuine 30% grower available at a de-rated ~26–28x forward / PEG ~0.66 / mid-cheap on its own ten-year history, with the CEO buying stock in the open market with his own cash. If ROTCE converges to even 16–18%, justified P/TBV is ~2.5–2.8x and EPS compounds to ~$1.00+ by 2028; the de-rating reverses toward the Robinhood zone. “The cheap 30% fintech grower the market gave up on.”

Strongest bear case. A moatless commodity consumer lender (SoFi’s own 10-K admits prime lending is fragmented and competitive; the Greenwald ROIC test fails at 9% ROTCE) whose earnings are lower-quality than they look — ~96% mark-to-model loans on an aggressive ~4.5% discount rate, headline NCOs suppressed by ~$360M of delinquent-loan sales, LPB fees that are upfront-recognized and counterparty-concentrated, and the supposed high-margin SaaS moat (Tech Platform) in outright decline as Chime built ChimeCore in-house. Layer ~16%/year dilution and a 9% ROTCE against a 2.2x tangible-book multiple pricing mid-teens-plus, into an untested consumer-credit cycle. “A commodity lender with mark-to-model earnings, priced for a 25%-ROTCE franchise it isn’t.”

The five assumptions that matter most — and what would falsify each side:

1. ROTCE convergence (the master variable). Bull needs 9% → mid-teens-plus; bear says it stalls ~10–12%. Falsify the bull: 4–6 quarters of ROTCE stuck below ~12% while book/share grows. Falsify the bear: ROTCE through ~15% and holding across a rate-cut cycle.
2. Fee/mark durability. Bull: LPB + capital-light fees are a structural mix shift. Bear: rate-cycle / private-credit-appetite artifacts; FV marks reverse when spreads widen. Falsify the bear: fees and noninterest income hold through a quarter of widening credit spreads. Falsify the bull: a spread move marks the FV book down and noninterest income drops sharply.
3. Credit normalization. Bull: FICO-749 underwriting and benign vintages hold. Bear: headline NCOs understate loss content because of delinquency sales. Falsify either way: the personal-loan NCO trend reported net of the delinquent-loan-sale adjustment.
4. Dilution / per-share compounding. Bear: ~16%/year share growth caps per-share returns. Falsify the bear: share growth decelerates below ~5% with the company self-funding. Falsify the bull: another large equity raise at a depressed price.
5. Technology Platform. Bear: structurally impaired (Chime gone, accounts −23%). Falsify the bear: Tech Platform revenue re-accelerates to double digits on new wins replacing Chime-scale volume.

The honest framing: a quality-grower-at-a-de-rated-price with unproven returns, contested between an Ally-zone bear (~1x tangible book, mature digital bank) and a Robinhood-zone bull (~3x tangible book, fintech compounder). The CEO’s open-market buying plus the crowded short create squeeze asymmetry; the mark-to-model earnings plus the dilution create a real downside if the credit cycle turns.

12. Fact vs. Interpretation Table

13. Open Questions

1. What is the cumulative unrealized fair-value gain embedded in the ~$36.5B fair-value loan book, and how much of reported equity would reverse if discount rates rose 100bps? (Not cleanly disclosed; the single most important QoE unknown.)
2. What is the personal-loan net charge-off trend net of delinquent-loan sales? The ~$360M of 2025 sales mechanically lowers the headline; the underlying trend is the real credit signal.
3. How concentrated is LPB across individual buyers, and how committed (vs. discretionary) are the $3.6B+ of new commitments if private-credit appetite contracts?
4. Can the Technology Platform replace Chime-scale volume, or does +14% FY2025 revenue on −23% accounts reverse once pricing/mix benefits exhaust?
5. Will products-per-member deepen beyond ~1.5x enough to convert the Productivity Loop narrative into a measurable retention/LTV moat?
6. What is the realistic timeline to a mid-teens ROTCE? Management pointedly declined to date it; the entire embedded-expectations case rides on the answer.
7. Does the SoFiUSD stablecoin / crypto build generate material, durable revenue, or is it optionality the bull case over-weights?
8. Will share count keep growing ~16%/year, and at what point does self-funding turn dilution into accretion?

14. What Must Be True

For the BULL case to be right:

• ROTCE must converge from ~9% toward the mid-teens-plus and hold across a rate cycle — the master variable. Falsification test: four-plus consecutive quarters of ROTCE stuck below ~12% while tangible book per share grows would break the thesis that returns are converging.
• Fee/capital-light income (LPB, interchange, tech) must prove structural, not cyclical — surviving a quarter of widening credit spreads without a sharp drop in noninterest income. Falsification test: a credit-spread widening that marks the fair-value book down and collapses noninterest income would expose the earnings as rate-cycle artifacts.
• Dilution must decelerate to single digits with the company self-funding growth. Falsification test: another large equity raise at a depressed price.

For the BEAR case to be right:

• Returns must stall and the credit cycle must turn, with personal-loan losses (net of delinquency sales) rising and the fair-value book marking down — proving the 9% ROTCE is the ceiling, not a waypoint. Falsification test: ROTCE pushing through ~15% on durable fee income would refute the “commodity lender priced for a franchise it isn’t” thesis.
• The Technology Platform must keep shrinking and the moat must stay un-evidenced in the returns. Falsification test: Tech Platform revenue re-accelerating to double digits on new wins would undercut the “negative-captivity BaaS” bear point.

The two cases share a single fulcrum: does ROTCE converge? Everything else — the multiple, the dilution math, the credit outcome — is downstream of that one question.

15. Source Appendix

Primary filings (SEC EDGAR, CIK 0001818874):

• SoFi Technologies FY2025 Form 10-K, filed 2026-02-17 (period ending 2025-12-31) — segment results, fair-value loan footnotes, allowance, deposits, capital ratios, risk factors.
• SoFi Technologies FY2024 Form 10-K, filed 2025-02-24 — deferred-tax-valuation-allowance release, prior-year comparatives.
• SoFi Technologies FY2023 / FY2022 / FY2021 Form 10-Ks — multi-year revenue, goodwill impairment (Technisys), member history.
• SoFi Technologies Q1-2026 Form 10-Q, filed 2026-05-07 (period ending 2026-03-31) — latest segment, deposit, NCO, capital data.
• SoFi Technologies DEF 14A proxy statements, filed 2026-04-30 and 2025-04-15 — CEO compensation, incentive metrics, say-on-pay, ownership.
• SoFi Technologies Form 4 corpus (insider transactions), 2021–2026 — CEO Anthony Noto discretionary open-market purchases (code P).
• SoFi Technologies Form 8-K filings, 2024–2026 — equity offerings, segment reorganization, exec/board changes.

Quantitative data (EDGAR XBRL via edgar.sh concept): RevenuesNetOfInterestExpense, NetIncomeLoss, StockholdersEquity, Deposits, NoninterestIncome, InterestIncomeExpenseNet — all reconciled to the 10-K/10-Q.

Transcripts (read in full): SoFi Q1-2026 (2026-04-29), Q4-2025 (2026-01-30), Q3-2025 (2025-10-28), Q2-2025 (2025-07-29) earnings calls; J.P. Morgan 54th Annual Global Technology, Media & Communications Conference presentation (2026-05-19); plus the 2021–2024 earnings-call archive.

Market/aggregated data (third-party, used for orientation and reconciled to filings): AZI fundamentals snapshot and own-history valuation index (P/E 22.5th, composite 41.5th percentile); yfinance quote (price $15.87 at 2026-06-10, market cap, EV, 52-week range); short-interest fields (173M shares, ~13.5% of float, 2.54 days to cover).

Industry/regulatory context (public): 8th Circuit ruling vacating the SAVE plan (2026-03-10); federal student-loan forbearance interest resumption (2025-08-01); CFPB supervision of SoFi Bank (effective 2024-01-01); OCC crypto-permissibility guidance for national banks (2025); Chime FY2025 disclosures and ChimeCore migration; Ally Financial public deposit/NIM disclosures.

All non-obvious facts are cited to a primary filing or dated public source in the body. Management commentary is treated throughout as hypothesis requiring external validation, not as evidence.

APPENDIX A — Standard Diligence Questionnaire

SoFi Technologies, Inc. (NASDAQ: SOFI) — Standard Diligence Questionnaire Appendix

Supplemental to the research memo (not counted toward its length standard). Report date 2026-06-11. Fact/Interpretation/Assumption labels applied where material.

General

What thoughtful questions have other investors asked about this company? The recurring institutional questions are: (1) Is the fair-value/mark-to-model accounting flattering reported earnings, and what reverses in a downturn? (2) Will the ~16%/year share dilution ever stop? (3) When does ROTCE close the gap from ~9% to the 20–30% target — and is the target credible? (4) Is the Technology Platform (Galileo) structurally impaired after Chime? (5) Is LPB a durable capital-light franchise or a cyclical, counterparty-dependent fee stream? These map onto the bull/bear axes in the Variant Perception section.

Cyclicality & Earnings Nature

Are earnings at a cyclical high or low? INTERPRETATION: mid-cycle but quality-flattered — NIM (5.85%) and fair-value marks are aided by the current rate/credit environment, so reported earnings sit above a normalized trough but the business is still scaling into operating leverage (so below its structural potential). The honest read is “good environment, immature returns.” Driven by external environment or internal actions? Both: internal (member/product growth, fee-mix shift, LPB) is the primary driver; external (rates, student-loan policy, credit cycle) is a meaningful swing factor. How stable are revenues? Member/product growth is remarkably steady (~35%/year), but the composition is volatile — gain-on-sale and fair-value marks swing with rates, and LPB fees are step-functions tied to partner commitments. Outlook for products/services? Growing: ~30% revenue CAGR guided through 2028; the addressable consumer-finance market is large, US-centric, and structurally shifting toward digital. How big will this market be? Large and growing (digital consumer banking/lending), but fragmented and competitive — share, not market size, is the constraint.

Business Quality & Competitive Moat

Is the industry getting more or less competitive? More — fintech is in the late-boom phase of the capital cycle (Marathon), with abundant capital, a fresh IPO wave (Chime), and heavy marketing spend. How profitable is the business (ROIC/ROE)? ROE ~6.6% TTM, ROA ~1.26%, ROTCE ~9% (Q4-2025) — below the cost of equity today, against a 20–30% target. FACT. How profitable is the industry / barriers to entry? Low barriers in lending, brokerage, and neobanking; the bank charter is the one real barrier, but it is shared. The 10-K calls prime lending “highly fragmented and competitive.” Can the business be easily understood? Mostly — three segments — but the fair-value loan accounting and segment contribution-profit definitions require care. Foreign low-cost-labor risk? Not material (US consumer-finance/regulated bank). Do brands matter? Yes for awareness, no as a barrier — SoFi spends ~30% of revenue on marketing to sustain the brand. Nature of competition? Price (deposit APY, loan APR), marketing spend, and product breadth — not a protected niche. Customers’ switching costs? Low-to-emerging — direct-deposit stickiness is real but products-per-member (~1.5x) shows the lock-in is shallow.

Financial Condition & Balance Sheet

Assets not fully recognized on the balance sheet? The brand, member base, and Galileo/Technisys technology are partly intangible; conversely, the embedded unrealized fair-value gains in the loan book are recognized but model-dependent. Off-balance-sheet liabilities? LPB shifts new credit risk off-balance-sheet (a positive), but loan-sale reps/warranties and securitization retained interests are contingent exposures. How conservative is the accounting? INTERPRETATION: aggressive-leaning on the loan book — ~96% at fair value with a ~4.5% discount rate and a $51M allowance; headline NCOs suppressed by ~$360M of delinquent-loan sales. Not fraudulent, but the opposite of conservative. How CapEx-hungry? Low physical capex (digital), but high customer-acquisition capex (S&M ~30% of revenue) and capitalized software — the economic equivalent of growth capex.

Capital Allocation & Management

How much FCF, and how is it used? FCF is not a meaningful metric for a balance-sheet-growing bank; capital is retained and downstreamed into SoFi Bank to fund loan growth. No dividend, no buyback — appropriate. Significant acquisitions recently? No major ones since 2023 (only the tiny Composer tuck-in); historical M&A is Galileo, Technisys (impaired), Golden Pacific (the charter win), Wyndham. Buying back shares? No — and it should not, while diluting. Issuing shares to insiders / dilution? Yes — chronic ~16%/year dilution from raises, SBC, and convertible conversions. The 2025 raises were opportunistically timed near highs. Compensation policy? CEO pay ~$30.3M (2025), 50% PSU / 50% RSU, metrics tied to revenue/EBITDA/ROTCE/TBV growth and a capital-ratio forfeiture trigger — reasonable, but say-on-pay passed at only ~76%. Motivations of management? Mixed signal but net positive — rich pay drew investor pushback, but the CEO is buying stock in the open market with his own cash (the strongest alignment signal available).

Valuation & Market Data

ADR/MLP/K-1? No — ordinary US common stock, NASDAQ-listed. Dividend policy? None. How profitable? GAAP-profitable since FY2024; normalized adjusted EPS $0.39 (FY2025), guided ~$0.60 (FY2026); ROTCE ~9%. Is net income diverging from cash from operations? Yes by design — operating cash flow is dominated by loan-balance-sheet growth, so the NI-vs-CFO gap is structural, not a red flag per se; the more relevant divergence is GAAP NI vs. cash revenue (the fair-value mark component).

Risks & Downside

What would cause the stock to decline? A consumer-credit downturn (losses spike + fair-value marks reverse), another large dilutive raise, an LPB counterparty pullback, further Tech Platform client losses, ROTCE stalling, or a rate path that removes the refi tailwind. See the relevant section. Catastrophic-loss risk? LOW (tail) — a total loss requires SoFi Bank insolvency, which the capital stack (Bank CET1 ~15%, holdco ~22%, FDIC, prime book) makes remote. Chance of a total loss? Remote. The realistic adverse outcome is permanent per-share impairment via a forced distressed equity raise, not equity-to-zero.

Recent News & Events

Has the business environment changed recently? Yes — favorably on student-loan policy (SAVE struck down March 2026; forbearance resumed August 2025), less favorably on rates (FY2026 guide now assumes zero cuts). The crypto/stablecoin launch opens new product and regulatory surface. Significant acquisitions? No. Change in accounting policies? No major change; the fair-value treatment is long-standing. Q3-2025 aggregated Galileo/Technisys into one reporting unit. Recent changes — new markets, facilities, management? New products (crypto, SoFiUSD stablecoin, Big Business Banking, SoFi Plus); ~$3.2B of equity raised in H2-2025; stable leadership; CFPB supervision since 2024.

APPENDIX B — Source Appendix

15. Source Appendix

Primary filings (SEC EDGAR, CIK 0001818874):

• SoFi Technologies FY2025 Form 10-K, filed 2026-02-17 (period ending 2025-12-31) — segment results, fair-value loan footnotes, allowance, deposits, capital ratios, risk factors.
• SoFi Technologies FY2024 Form 10-K, filed 2025-02-24 — deferred-tax-valuation-allowance release, prior-year comparatives.
• SoFi Technologies FY2023 / FY2022 / FY2021 Form 10-Ks — multi-year revenue, goodwill impairment (Technisys), member history.
• SoFi Technologies Q1-2026 Form 10-Q, filed 2026-05-07 (period ending 2026-03-31) — latest segment, deposit, NCO, capital data.
• SoFi Technologies DEF 14A proxy statements, filed 2026-04-30 and 2025-04-15 — CEO compensation, incentive metrics, say-on-pay, ownership.
• SoFi Technologies Form 4 corpus (insider transactions), 2021–2026 — CEO Anthony Noto discretionary open-market purchases (code P).
• SoFi Technologies Form 8-K filings, 2024–2026 — equity offerings, segment reorganization, exec/board changes.

Quantitative data (EDGAR XBRL via edgar.sh concept): RevenuesNetOfInterestExpense, NetIncomeLoss, StockholdersEquity, Deposits, NoninterestIncome, InterestIncomeExpenseNet — all reconciled to the 10-K/10-Q.

Transcripts (read in full): SoFi Q1-2026 (2026-04-29), Q4-2025 (2026-01-30), Q3-2025 (2025-10-28), Q2-2025 (2025-07-29) earnings calls; J.P. Morgan 54th Annual Global Technology, Media & Communications Conference presentation (2026-05-19); plus the 2021–2024 earnings-call archive.

Market/aggregated data (third-party, used for orientation and reconciled to filings): AZI fundamentals snapshot and own-history valuation index (P/E 22.5th, composite 41.5th percentile); yfinance quote (price $15.87 at 2026-06-10, market cap, EV, 52-week range); short-interest fields (173M shares, ~13.5% of float, 2.54 days to cover).

Industry/regulatory context (public): 8th Circuit ruling vacating the SAVE plan (2026-03-10); federal student-loan forbearance interest resumption (2025-08-01); CFPB supervision of SoFi Bank (effective 2024-01-01); OCC crypto-permissibility guidance for national banks (2025); Chime FY2025 disclosures and ChimeCore migration; Ally Financial public deposit/NIM disclosures.

All non-obvious facts are cited to a primary filing or dated public source in the body. Management commentary is treated throughout as hypothesis requiring external validation, not as evidence.