KurzüberblickVerständlicher Überblick · zuerst lesen
SoFi Technologies is a U.S. digital bank that funds lending with deposits, distributes loans capital-light through its Loan Platform Business, and sells banking infrastructure through Galileo and Technisys. The report rates it Hold.
The operating record is strong. Q2 2026 adjusted net revenue was $1.206 billion at a 30% adjusted EBITDA margin, members rose 35% year on year to 15.8 million, and the quarter was SoFi's eleventh consecutive GAAP profit. Deposits reached $45.5 billion, funded more than 90% of average liabilities and cost 156 basis points less than warehouse borrowing. The fee engine is growing but has not taken over yet: Q2 net interest income of $788.2 million still outweighed management-defined fee revenue of $472.3 million, so credit, deposits and regulatory capital continue to set the earnings power of the whole company.
Valuation is where the report parts company with the market. At $18.22 the shares trade at 2.48 times Q2 tangible book of $7.34 per share. Run through a standard bank framework at a 10.5% cost of equity and 4% long-term growth, that multiple corresponds to roughly a 20% sustainable return on tangible equity. The 2026 guidance implies about 9%. The market is paying today for returns to roughly double, and SoFi has to grow into the multiple through fee income and operating leverage rather than through member additions alone.
Credit is the unresolved part. Vintages originated from Q4 2022 through Q3 2025 show 4.68% cumulative losses with 35% of principal still outstanding, inside SoFi's 7% to 8% lifetime tolerance, and the all-in annualised personal-loan charge-off rate improved to about 3.7%. The figures look sound, but SoFi's history as a public company has never included a conventional U.S. recession with the unsecured book anywhere near its current size.
The report puts fair value at $16 to $19 and an ideal buy range of $9.50 to $11.00, so the current price sits well above where it would consider buying. Combined credit, loan-platform and capital stress puts maximum loss at roughly 45% to 60%. The stance is a good franchise at a price that already assumes the transition works.
The above is a summary of the report's views and does not constitute investment advice. Markets carry risk; invest with caution.
EinleitungSoFi Technologies is a U.S. digital bank that funds lending with deposits, distributes loans capital-light through its Loan Platform Business, and sells banking infrastructure through Galileo and Technisys. Q2 2026 set records, with adjusted net revenue of $1.206bn, 15.8 million members up 35% and $45.5bn of deposits, yet net interest income of $788.2m still outweighed fee-based revenue of $472.3m, so bank economics remain decisive. At $18.22 the shares trade at 2.48 times tangible book, which under a Gordon framework implies roughly a 20% sustainable return on tangible equity against the 9% adjusted ROTE embedded in 2026 guidance. Rating Hold: fee growth and cheap deposits are real, but the price already pays for returns to double while the personal-loan book has never been tested by a U.S. recession.
Die Preise im Artikel entsprechen dem Stand bei Veröffentlichung; den Live-Preis zeigt das Bewertungsband oben.
Meta
- Ticker: SOFI.US
- Company: SoFi Technologies, Inc.
- Price & market cap: $18.22 per share, $23.53 billion market capitalisation as of 2026-09-04 close; market cap uses SoFi’s latest company-disclosed 1,291,570,324 common shares outstanding as of 2026-07-31. The company-based figure is preferable to vendor market-cap figures that use different diluted-share denominators.
- Currency: USD
- Report date: 2026-09-07
- Industry: Consumer Finance
- One-line positioning: A U.S. digital bank combining deposit-funded lending, capital-light loan distribution, consumer financial products, and Galileo/Technisys banking infrastructure.
Research scope: first-time coverage, with no prior SoFi report inherited or challenged. The analysis uses a horizontal × vertical framework, a balanced risk posture, a 12-month valuation view and a 3–5-year business-quality view. The research base date is 2026-09-07. U.S. exchanges are closed for Labor Day on September 7, so the reference price is the Friday, September 4 close.
Research Summary
SoFi has completed one transformation and is attempting a second.
The first one is finished. A specialist student-loan refinancer turned into a regulated, deposit-funded financial institution. Federal Reserve approval in January 2022, then completion of the Golden Pacific Bancorp acquisition on February 2, 2022, made SoFi Technologies a bank holding company and put SoFi Bank, N.A. inside the national-bank regulatory perimeter. The parent is supervised by the Federal Reserve; SoFi Bank is a national bank principally supervised by the OCC, with FDIC-insured deposits. The charter changed the economics of the company, because deposits could replace much more expensive warehouse financing.
The second is still under way. SoFi wants more growth to come from fees, distribution, transactions and infrastructure rather than from placing every new loan on its own balance sheet. The Loan Platform Business, or LPB, carries that effort: SoFi underwrites or refers loans for third-party capital providers and collects fees without funding all the credit itself. In Q2 2026, LPB contributed $143.3 million of adjusted net revenue; $140.9 million came from about $3.1 billion of personal loans originated for third parties plus referrals. Call it an effective revenue-to-originations ratio of roughly 4.5%, although this is not a contractual “take rate” because referral revenue sits in the numerator.
There is a catch. The consolidated numbers do not yet show a clean migration away from bank economics. Q2 2026 net interest income reached $788.2 million, up 52% year on year, against management-defined fee-based revenue of $472.3 million. Net interest margin rose to 5.98%. Deposits reached $45.5 billion and accounted for more than 90% of average liabilities. The capital-light business is growing quickly; in the latest quarter the balance-sheet bank grew faster still.
The best description of SoFi today is “company in transition”: its fee engine is becoming economically important, while credit, deposits and regulatory capital still determine the earnings power of the whole company.
The distinction matters because the stock market frequently prices SoFi more like a high-growth fintech platform than a bank. At $18.22, the shares trade at about 2.48 times Q2 tangible book value of $7.34 per share and about 30 times the company’s 2026 adjusted EPS guidance of roughly $0.60. Using the latest company-disclosed share count rather than a vendor’s diluted-market-cap convention gives a basic market capitalisation of $23.53 billion. Against that, the 2026 adjusted revenue midpoint of $4.80 billion represents roughly 4.9 times price-to-adjusted-revenue, and the $1.6 billion adjusted EBITDA target implies roughly 14.7 times market cap to adjusted EBITDA. Those fintech-style ratios look far less demanding than 2.48 times tangible book against a current adjusted return on tangible equity that remains below 10%.
The gap between those two readings is the central valuation disagreement. A Gordon-style bank framework shows what the current multiple demands: at a 10.5% cost of equity and 4% long-term growth, a 2.48-times tangible-book valuation mathematically corresponds to roughly a 20% sustainable return on tangible equity. Small changes in the assumed cost of equity move that figure. The message survives them. The market is capitalising a future SoFi that earns dramatically higher returns on tangible capital than the one visible today, and the company has to grow into the multiple through fee income, operating leverage and disciplined credit rather than merely grow members.
The operating momentum is formidable. Q2 2026 GAAP net revenue was $1.219 billion, adjusted net revenue $1.206 billion, adjusted EBITDA $357.8 million, GAAP net income $156.6 million and adjusted net income $160.4 million. Adjusted EBITDA margin reached 30%. Members rose 35% year on year to 15.8 million, with a record 1.1 million net additions, and total products reached 24.4 million. One caveat on that last figure: the definition of “products” changed in Q1 2026 when SoFi expanded the SoFi Invest metric from two product categories to four, so the 42% product-growth headline is not perfectly comparable with earlier periods.
The profitability is real too. Q2 was SoFi’s eleventh consecutive GAAP-profitable quarter. The 2025 quarterly net-income sequence ran approximately $71 million, $97 million, $139 million and $174 million, and H1 2026 added another $323 million. That half-year figure implies roughly $166 million in Q1 2026, so the quarterly GAAP line has stepped down twice from the $174 million posted in Q4 2025 even as quarterly revenue kept rising. Profitability has now lasted well beyond one unusually favourable quarter.
Credit is the unresolved part of the story. SoFi says it underwrites personal loans to a maximum expected cumulative life-of-loan net-loss tolerance of roughly 7–8%. Recent vintages from Q4 2022 through Q3 2025 had generated approximately 4.68% cumulative losses with around 35% of original principal still outstanding; the comparable 2017 vintage stood at 6.43% at the same seasoning. Across Q1 2020–Q1 2026 personal-loan originations, roughly 62% of principal had already paid down with cumulative losses around 6.8%. On the same base, a further 3% loss rate on the remaining 38% of principal would push the full cohort to an 8% cumulative loss.
Q2 reported annualised personal-loan charge-offs were 2.62%, down from 3.03% in Q1. Management’s more economically useful “all-in” measure, which strips out the benefit of selling delinquent loans, came in around 3.7%, roughly 70 basis points better sequentially and 80 basis points better year on year. Late-stage delinquency was broadly stable year on year. The figures look sound. But SoFi’s public-company history has not included a conventional U.S. recession while its unsecured personal-loan book was at anything close to current scale.
The accounting makes that uncertainty more consequential. SoFi elects fair-value accounting for its major personal, student and home-loan portfolios. At June 30, it carried $44.30 billion of unpaid principal at $46.60 billion of total fair value, including $2.03 billion of cumulative fair-value adjustments. Personal loans alone accounted for $26.10 billion of unpaid principal and a $1.22 billion positive cumulative fair-value adjustment. That premium can contract when default, prepayment or discount-rate assumptions worsen. It already did in Q2: the total cumulative fair-value adjustment fell $11.6 million sequentially even though unpaid principal increased more than $5.9 billion, with management citing higher benchmark discount rates and a higher personal-loan annual default assumption among the offsets to portfolio growth.
A conventional loan-loss allowance ratio gives the wrong impression of the personal-loan book. Expected credit risk sits inside fair value rather than in a separate CECL allowance against those loans. SoFi’s positive $1.22 billion personal-loan fair-value adjustment is a net valuation premium, not a $1.22 billion credit-loss reserve. Initial and subsequent fair-value changes run through non-interest income within loan origination, sales, securitisation and servicing.
Technology Platform is the weakest leg. Galileo and Technisys once supplied much of the reason investors could imagine SoFi as a software company attached to a lender. The latest numbers do not validate that thesis. Q2 2026 Technology Platform revenue fell 23% to $84.5 million; contribution profit fell to $11.8 million and margin to 14% from 30% a year earlier. Enabled accounts were 135 million, down 16% year on year after a large client was fully offboarded, though accounts recovered by roughly two million sequentially.
Set that against the SPAC-era plan and it is more revealing still. SoFi’s 2021 merger materials envisaged roughly $3.7 billion of 2025 revenue and a far larger Technology Platform contribution to the mix. Actual 2025 adjusted net revenue of about $3.6 billion came surprisingly close to the consolidated target. The mix arrived by a different route: Financial Services and lending economics compensated for Technology Platform materially undershooting the software thesis.
Capital is thick at the parent and adequate at the bank. The Q2 filing showed estimated SoFi Technologies CET1 of 18.7% and total risk-based capital of 18.8%, against required minimums including the capital conservation buffer of 7.0% and 10.5%, respectively. SoFi Bank’s total risk-based capital ratio was 15.3%, down from 16.6% at year-end 2025 and still above the 10% “well-capitalized” benchmark. Rapid asset growth is consuming the cushion: consolidated risk-weighted assets rose from $37.2 billion at year-end to $48.7 billion by June.
How the market handled Q2 2026 clarifies what investors are now paying for. The quarter beat headline expectations and management raised adjusted-net-revenue guidance to $4.75–4.85 billion, yet held the $0.60 adjusted EPS outlook and the roughly $1.6 billion adjusted EBITDA target. TD Cowen described the beat as tempered by unchanged EPS guidance; management said it was leaning into higher-growth investment opportunities. The shares initially fell roughly 9–10%. The information in that reaction was the weaker conversion of incremental revenue into incremental 2026 profit, compounded by Technology Platform weakness. The size of the move also reflected a shareholder base that had repeatedly paid for upside guidance.
The central bull/bear disagreement is measurable. Bulls see deposit-funded lending, a rapidly scaling capital-light distribution business, 35% member growth, double-digit GAAP profitability and a large capital base as the ingredients for future 15–20% tangible returns. Bears see a company whose present return on tangible equity remains far below what a 2.5-times tangible-book multiple requires, whose personal-loan credit has not been recession-tested, whose software segment has gone backwards, and whose share count has risen materially.
That framework carries through the rest of this report. Member growth is evidence of distribution power. It is not yet evidence of 20% sustainable ROTCE.
Vertical History and Financial Evolution
SoFi began in 2011 as a response to a specific distortion in U.S. consumer credit. Graduates with good earning prospects often carried federal or private student debt priced with little differentiation for their post-graduation risk. The original model used an alumni-linked funding concept to refinance high-quality graduate borrowers at lower rates. Refinancing gave the company its first customer-acquisition engine and its underwriting identity: select relatively affluent borrowers, price them below legacy lenders, and try to keep the relationship as those borrowers accumulated more financial needs.
The model evolved because a student-loan specialist had a structural ceiling. Personal loans, mortgages, investing, cash management and credit products opened more ways to monetise each acquired household. Anthony Noto’s arrival as chief executive in 2018 accelerated the transition. The strategic logic settled into a “one-stop” consumer-finance company whose lending economics could fund customer acquisition while non-lending products increased lifetime value.
A second strategic turn came through infrastructure. SoFi bought Galileo in 2020 for roughly $1.2 billion, picking up payments and account-processing technology already used by other fintechs, then added Technisys in 2022 for cloud-native core-banking technology and Latin American clients. The ambition was to own more of the technology stack underneath both its own products and those of outside financial institutions. Buying Wyndham Capital Mortgage in 2023 added mortgage-origination capability.
The public-market birth came through a SPAC rather than a conventional IPO. SoFi agreed in January 2021 to combine with Social Capital Hedosophia Holdings Corp. V in a transaction that valued the company at roughly $8.65 billion in equity value and brought a large amount of cash onto the balance sheet. The merger closed in late May, and SoFi common stock began trading on Nasdaq under SOFI on June 1, 2021.
The market presentation was aggressive, and in hindsight unusually informative. The 2021 plan projected rapid revenue growth toward roughly $3.7 billion by 2025, a much larger share from Technology Platform and Financial Services, and eventual adjusted EBITDA margins above 30%. Actual 2025 adjusted net revenue landed close to the top-line aspiration. The software mix lagged sharply. SoFi largely reached the consolidated destination by a different road.
The history is best divided into four economic stages.
The first, from 2011 through the late 2010s, was product validation: student-loan refinancing proved that SoFi could attract desirable borrowers digitally and underwrite them differently from mass-market lenders. What lasted was the customer-acquisition brand among younger professionals. The limitation that also lasted was reliance on securitisation and wholesale funding.
The second stage, culminating in the 2021 listing, was platform expansion. Galileo, brokerage, cash management, personal lending and mortgage products enlarged the addressable wallet. Capital markets treated SoFi as a fintech growth asset, which left the share price highly sensitive to long-duration growth multiples and liquidity.
The third stage started in 2022 with the bank charter. The Federal Reserve approved SoFi Technologies’ bank-holding-company formation in January, and the Golden Pacific transaction closed on February 2. Deposits became a structural funding source. It altered lending margins more profoundly than any app redesign: SoFi could hold more loans with lower funding expense and could capture net interest income inside the bank.
The fourth stage began around 2024–25. GAAP profitability arrived, LPB started scaling, Financial Services turned profitable, deposits surged and the company raised a great deal of fresh equity while its stock price was high. That is where things stand: SoFi is deliberately combining a growing bank balance sheet with more capital-light fee generation rather than choosing one model exclusively.
The financial vertical makes the transition visible.
| Fiscal period | Adjusted net revenue | GAAP net income | Deposits at period end | Broad financial interpretation |
|---|---|---|---|---|
| 2021 | about $1.0bn | loss of about $0.48bn | pre-bank-charter | growth fintech funded through capital markets |
| 2022 | about $1.6bn | loss of about $0.32bn | about $7bn | bank charter begins changing funding economics |
| 2023 | about $2.1bn | loss of about $0.30bn | about $19bn | deposit scaling, still investing heavily |
| 2024 | about $2.6bn | $0.50bn† | $26.0bn | first full GAAP-profitable year |
| 2025 | about $3.6bn | about $0.48bn | $37.5bn | profitability broadens; two large equity raises |
| H1 2026 | $2.31bn adjusted | $0.323bn GAAP | $45.5bn | lending and fee engines both accelerating |
† 2024 GAAP net income included a large deferred-tax valuation-allowance benefit; adjusted net income was materially lower. The figures above are reconstructed from SoFi filings and quarterly releases rather than treated as a single homogeneous five-year series because both the charter and segment mix changed materially.
Revenue growth came first from more members and lending, then increasingly from deposits, Financial Services monetisation and LPB. The change in earnings quality matters more than the raw CAGR. SoFi moved from repeated losses to recurring GAAP profits while using deposits instead of warehouses for most incremental funding. In Q2 2026 the average rate on interest-bearing deposits was 3.07%, versus 4.63% on warehouse facilities; in Q2 2025 those rates were 3.36% and 5.23%. Measured against warehouse borrowing, deposit costs fell 29 basis points while warehouse costs fell 60 basis points, a roughly 48% “funding beta.” So the deposit advantage narrowed from 187 basis points in Q2 2025 to 156 basis points in Q2 2026, although deposits remained materially cheaper.
The deposit mix is getting more complicated. Q2 deposits included $37.0 billion of savings balances, $4.0 billion of demand deposits and $4.38 billion of time deposits; brokered deposits had increased to about $4.53 billion from $1.40 billion at year-end 2025. That does not negate the funding advantage, but it means the marginal dollar of deposit growth is becoming less obviously “cheap retail checking and savings.”
An eight-quarter net-interest/non-interest reconstruction shows why the “fee mix is rising” narrative needs qualification.
| Quarter | GAAP net revenue | Net interest income | Implied GAAP non-interest revenue | NIM |
|---|---|---|---|---|
| Q3 2024 | $697m | $431m | about $266m | 5.57% |
| Q4 2024 | $734m | about $471m | about $263m | 5.91% |
| Q1 2025 | $772m | about $499m | about $273m | about 5.9% |
| Q2 2025 | $855m | $518m | about $337m | 5.86% |
| Q3 2025 | $962m | $585m | about $377m | 5.84% |
| Q4 2025 | about $1.01bn | $617m | about $396m | 5.72% |
| Q1 2026 | about $1.10bn | $693m | about $408m | 5.94% |
| Q2 2026 | $1.219bn | $788m | about $431m | 5.98% |
Figures are rounded because SoFi’s management presentations use adjusted and GAAP revenue alongside segment-specific non-interest definitions; “implied GAAP non-interest revenue” above is total GAAP net revenue less reported consolidated NII.
The direction is striking. Non-interest dollars are far larger than two years ago, yet NII has accelerated so quickly that the non-interest share of consolidated GAAP revenue is not monotonically rising. Management-defined fee-based revenue was $377.5 million in Q2 2025, about 44% of adjusted net revenue. In Q2 2026, $472.3 million of fee-based revenue represented about 39%.
SoFi is becoming more capital-light at the margin in selected activities, but it is not yet becoming less bank-like at the consolidated level.
LPB shows the other side of the picture.
| Quarter | Personal originations | Third-party LPB volume† | Implied SoFi-funded / retained route‡ | LPB loan-platform fees | Effective fee/volume |
|---|---|---|---|---|---|
| Q3 2024 | about $4.9bn | about $1.0bn | about $3.9bn | about $56m | about 5.6% |
| Q4 2024 | about $5.2bn | $1.1bn | about $4.1bn | $63.2m | 5.7% |
| Q1 2025 | about $5.5bn | $1.6bn | about $3.9bn | $92.8m | 5.8% |
| Q2 2025 | $7.0bn | $2.4bn | about $4.6bn | $127.4m | 5.3% |
| Q3 2025 | about $8.3bn | $3.4bn | about $4.9bn | $164.9m | 4.9% |
| Q4 2025 | about $7.8bn | $3.7bn | about $4.1bn | $190.9m | 5.2% |
| Q1 2026 | $8.3bn | $3.0bn | about $5.3bn | $138.3m | 4.6% |
| Q2 2026 | $10.7bn | $3.1bn | about $7.6bn | $140.9m | 4.5% |
† Third-party volume includes personal loans originated for third parties in LPB; disclosed fee revenue also includes referrals, so the final column is an effective ratio rather than contractual pricing. ‡ This is a routing approximation, calculated as total personal originations less disclosed third-party LPB originations; it should not be read as the exact quarter-end balance-sheet addition because loans can be sold or securitised after origination. Figures in early quarters are rounded from company segment disclosures.
LPB clearly has structural significance. Third-party volumes rose from roughly $1 billion a quarter in late 2024 to above $3 billion. Q2 2026 is also a warning against extrapolating: the third-party share of personal originations fell to roughly 29% from nearly half around Q4 2025, because total personal-loan demand exploded. The fee rate drifted from the mid-5% area toward the mid-4% area.
The evidence supports a hybrid reading. LPB is a real capital-efficiency mechanism, not an accounting narrative, and its contribution can scale without a matching increase in risk-weighted assets. But current growth also reflects a very strong personal-loan cycle, and SoFi is happily putting billions of those loans on its own balance sheet when economics and capital allow it.
Technology Platform tells a different vertical story.
| Quarter | Technology Platform revenue | YoY trend | Contribution margin | Enabled accounts |
|---|---|---|---|---|
| Q3 2024 | about $102m | positive | low-30%s | about 160m |
| Q4 2024 | about $103m | positive | about 31% | roughly mid-160m |
| Q1 2025 | $103.4m | +10% | 30% | 158.4m |
| Q2 2025 | $109.8m | +15% | 30% | 160.0m |
| Q3 2025 | $114.6m | +12% | 28% | 157.9m |
| Q4 2025 | $122.4m | +19% | 39% | 128.5m |
| Q1 2026 | $75.1m | -27% | 16% | 132.9m |
| Q2 2026 | $84.5m | -23% | 14% | 134.8m |
SoFi attributes the account contraction primarily to a large client offboarding completed by year-end 2025; the client is not identified in the comparable public segment disclosure. Q2 accounts did rise sequentially. Revenue and contribution margin remain far below the prior-year level.
The original software thesis has weakened as a result. Galileo and Technisys still carry strategic value, since SoFi controls technology that can serve both its own products and outside institutions. The numbers no longer justify treating Technology Platform as an independent high-growth SaaS asset until revenue growth and margins recover.
The balance sheet grew faster than the share count only after a large amount of capital was raised. That share history matters, because per-share value cannot be reconstructed from corporate growth alone.
| Disclosure date | Common shares outstanding |
|---|---|
| 2022-10-31 | 927.9m |
| 2023-10-31 | 958.7m |
| 2024-10-31 | 1,085.1m |
| 2025-04-30 | 1,105.4m |
| 2025-07-31 | 1,197.5m |
| 2025-12-31 | 1,270.6m |
| 2026-07-31 | 1,291.6m |
From October 2022 through July 2026, the basic share count rose about 39%; from October 2024 alone it rose about 19%. Equity compensation, share-funded convertible-note repurchases and the two 2025 common-equity offerings all played a part. Q2 2026 diluted weighted-average shares were 1.352 billion, 14% above the 1.183 billion in Q2 2025.
The 2025 tangible-book jump has to be read through that capital issuance. Tangible book ended 2025 around $8.9 billion, or $7.01 per share, versus roughly $4.47 per share a year earlier. SoFi earned approximately $481 million of GAAP net income during 2025, which explains only a small part of the roughly $4 billion absolute increase. SoFi raised roughly $1.5 billion of common equity in the summer and another $1.5 billion in December; the December offering was explicitly announced at $1.5 billion and priced when the stock was far above tangible book. The remainder of the bridge comes from employee/equity activity and accounting movements including the interaction of intangible amortisation with tangible book.
A rough economic reconciliation:
| 2025 tangible-book bridge | Approximate contribution |
|---|---|
| Beginning tangible book | about $4.9bn |
| Two common-equity raises, net of issuance costs | roughly $2.9bn |
| 2025 GAAP net income | about $0.48bn |
| Other equity / tangible-book accounting effects | roughly $0.6bn |
| Ending tangible book | about $8.9bn |
The per-share reading is more balanced than the absolute $4 billion increase. Issuing stock well above tangible book was accretive to tangible book per share, which is why the 2025 capital raises made economic sense for the company even though they diluted ownership percentages. The December offering was announced when the shares were near $30, against year-end tangible book of only $7.01 per share.
Convertibles remain another denominator risk. SoFi repurchased parts of its 2026 convertible notes in several transactions, settling portions with shares, and as of early 2026 roughly $428 million of principal remained; the remaining notes could represent about 19.1 million shares at an approximately $22.41 conversion price. Separately, SoFi issued $862.5 million of 1.25% convertible notes due 2029 in March 2024, and the conditional conversion threshold was met during Q1 2026, although no conversion had been reported through the early-May filing date.
The price history mirrors these changing narratives. The post-SPAC stock first traded as a long-duration fintech growth name, fell into the low-single-digit/high-single-digit region during the 2022–23 tightening and profitability doubts, then re-rated as the bank charter, deposits and GAAP profitability gained credibility. By July 28, 2025 the close was $21.02. The stock subsequently reached a 52-week high of $32.73, and on December 4, just before the second $1.5 billion equity offering, the close was about $29.60. Q1 2026’s record results produced a roughly 12% sell-off because guidance was left unchanged; Q2’s record results produced another initial high-single-digit decline because profit guidance again did not move with the revenue beat. The September 4 close is $18.22.
The two-year attribution is necessarily an event-study inference rather than a regression. Roughly half of the decline from the $32.73 52-week high to $18.22 can be associated directly with company-specific repricing around Q1/Q2 guidance conversion, dilution concerns and Technology Platform disappointment; the remainder reflects the compression of high-beta fintech valuations as rate and risk-appetite expectations changed. The supporting evidence is that Q1 and Q2 alone produced approximately 12% and 9–10% initial post-result declines despite record operating figures.
Crypto adds another chapter, though it currently contributes far less earnings than the lending/deposit machine. SoFi reintroduced crypto investing in Q4 2025 from within its regulated banking structure, with the product launched on December 22; by year-end it counted 63,441 crypto products. In Q1 2026 it began work around SoFiUSD, a dollar-reserve stablecoin, and announced interoperability/settlement development with Mastercard.
Federal supervisory policy loosened in parallel. The Federal Reserve withdrew earlier crypto-asset and dollar-token guidance on April 24, 2025; in August it sunset the dedicated novel-activities supervision program and returned such activities to ordinary supervision; in December 2025 it replaced a previous policy statement with one explicitly intended to facilitate responsible innovation. The regulatory burden remains: digital assets inside a national bank remain subject to ordinary prudential supervision, BSA/AML controls, custody requirements and consumer law.
Student lending is simultaneously exposed to federal policy. Interest began accruing again for borrowers in the now-defunct SAVE plan on August 1, 2025, and the Department of Education implemented a new repayment framework in 2026. From July 1, 2026, qualifying federal Direct Loan borrowers using autopay received a temporary 1% interest-rate reduction through June 30, 2028. That reduction marginally worsens the refinance proposition for SoFi, because a private refinance gives up federal protections and now competes with a lower effective federal rate. Simplification of repayment and reduced forgiveness options pull the other way, making private refinancing economically relevant to more borrowers.
Business Model, Moat, Industry, and Competition
SoFi’s business machine has three reported segments, but five economic engines.
Start with lending. SoFi originates personal, student and home loans, and it can retain them, sell them into capital markets, securitise them or originate them on behalf of third parties. Personal loans are the economic centre of gravity: yields are high enough to create a large spread over deposits, while the borrowers are generally prime or near-prime rather than classic subprime. Q2 personal originations reached $10.7 billion, student loans $2.69 billion and home loans $1.39 billion.
Deposits are the second engine. A digital bank can pay customers an attractive savings rate and still fund loans more cheaply than warehouse facilities allow. Q2 average interest-bearing deposit cost was 3.07%, versus 4.63% on warehouses, which management estimates saved about $713 million of annualised interest. This funding advantage is one of SoFi’s most tangible competitive assets.
Third comes Financial Services. Money, Invest, card, Relay, referred loans and related products throw off interchange, brokerage, referral, deposit-spread and other revenue. Q2 Financial Services revenue was $466.3 million: $249.1 million of NII and $217.2 million of non-interest income. Almost an even split, which is why segment profitability matters more than raw product counts.
LPB counts as a fourth engine even though it sits across reporting segments. It takes underwriting and distribution capability originally built for SoFi’s own loan book and sells access to outside capital. Q2 adjusted revenue contribution was $143.3 million. At an annualised run rate above $570 million, LPB is already economically larger than Technology Platform’s current revenue run rate.
Technology Platform is meant to be the least capital-intensive engine. Galileo processes accounts and payments; Technisys supplies core-banking infrastructure. Yet at $84.5 million of quarterly revenue, down 23% year on year, it is no longer carrying the valuation narrative.
The cost structure produces real operating leverage. Much of product engineering, compliance, banking infrastructure and corporate cost does not rise one-for-one with members. Financial Services is the cleanest demonstration: the segment went from years of investment losses to contribution profitability as revenue scaled over the same infrastructure. Consolidated adjusted EBITDA margin reached 30% in Q2 2026 from 29% in Q2 2025 despite heavy continued member growth.
Credit expense and funding expense are the limits to that leverage. A software company can add a customer without reserving capital against an unsecured loan. SoFi cannot. Every retained personal loan consumes balance sheet, capital and credit capacity. The 2025 equity raises and the increase in risk-weighted assets from $37.2 billion at year-end 2025 to $48.7 billion in June 2026 make this visible.
Three moats deserve serious weight.
The bank charter is the strongest. National-bank status lets SoFi gather deposits nationwide under a unified regulatory structure and use those deposits to fund lending. The Q2 funding-cost differential is hard-dollar evidence of economic value.
Distribution and cross-sell come second. Member count rose to 15.8 million and Financial Services revenue is now hundreds of millions per quarter. An existing Money or Relay relationship creates a lower-friction path to a personal loan, brokerage account, card or refinance offer. The evidence supports a distribution advantage; the company’s proprietary “member” and “product” definitions should prevent investors from treating those figures as equivalent to monthly active users or paying accounts.
The third moat is underwriting plus capital-market distribution. LPB suggests SoFi’s credit-acquisition machinery has value to external capital providers. It becomes a stronger moat if outside buyers stay through a weak credit cycle. Today it is proven in a healthy consumer environment, not through recession.
Technology Platform is a potential moat rather than a proven one. Client switching costs in core financial infrastructure can be large, yet losing a single large client was enough to cut enabled accounts by double digits and drive 2026 revenue sharply lower. That is evidence of concentration and execution risk.
Management’s record is mixed in a useful way. Anthony Noto’s team deserves credit for obtaining the charter, scaling deposits, reaching sustained GAAP profitability, building Financial Services contribution profit and raising common equity when the stock traded at a large premium to tangible book. Those actions strengthened per-share tangible capital despite headline dilution. Set against that, Technisys/Galileo has delivered far less than the original capital-markets thesis, and the repeated reliance on stock issuance and equity compensation means ordinary shareholders must judge growth per diluted share rather than corporate growth.
No perfect comparable exists. SoFi sits between a consumer bank, an unsecured marketplace lender, a digital bank and a financial super-app. Four comparables each describe a different piece of the economics.
Capital One is the credit benchmark, a much larger deposit-funded consumer bank with deep card and unsecured-credit experience. Q2 2026 net revenue was $15.9 billion, NIM 8.01%, net charge-offs about $3.6 billion and an overall NCO rate around 3.23%; its CET1 ratio was about 13.7%. This is what mature consumer-credit economics look like across cycles.
LendingClub is the closest structural analogue to LPB plus a bank charter, originating unsecured personal loans, retaining selected ones and selling the rest to investors. Q2 2026 net charge-offs on loans held for investment were about $40.6 million, with credit improving from the prior year. Its much smaller ecosystem and member funnel make SoFi the stronger consumer platform. What LendingClub offers is a cleaner test of whether marketplace loan buyers remain available when credit weakens.
Nu Holdings shows what the market can reward when a digital bank proves very high returns on equity. It reported Q2 2026 net income of $1.1 billion, revenue near $5.9 billion and ROE of 33%; NII was $3.7 billion and risk-adjusted NIM 12.4%. Geography and nominal interest rates make direct margin comparison inappropriate, but the 33% ROE is the relevant lesson. Nu earns the type of return that can justify a several-times-book digital-bank multiple.
Robinhood is the consumer-app and capital-light benchmark. Q2 2026 revenue rose 32% to $1.31 billion, net income reached $573 million, and Gold subscriptions reached 4.8 million. Robinhood can monetise brokerage, prediction markets, subscriptions and interest without putting a SoFi-sized unsecured consumer-loan book on its own balance sheet. Its valuation can rationally lean harder on growth and EBITDA than on tangible book.
A cross-section illustrates the category problem.
| Dimension | SoFi | Capital One | Nu | Robinhood |
|---|---|---|---|---|
| Q2 2026 quarterly revenue | $1.22bn GAAP | $15.9bn | about $5.9bn | $1.31bn |
| Current return signal | roughly 7% annualised Q2 GAAP ROTE† | mature bank returns | 33% ROE | asset-light profit model |
| Credit exposure | high, unsecured personal | high, cards/consumer | high, LatAm consumer | limited direct consumer-loan balance sheet |
| Capital measure | 18.7% consolidated CET1 | about 13.7% CET1 | jurisdiction-specific | not a bank valuation anchor |
| Approximate valuation frame | 2.48x TBV; 30x 2026 adjusted EPS | low-to-mid 1x TBV range | several-times book | platform-growth multiple |
† Q2 annualised GAAP net income divided by roughly average H1 tangible book; this is a research calculation, not a company-defined ROTE. Peer valuation ranges are cross-checks rather than strict like-for-like measures because accounting, geography and capital regimes differ.
SoFi’s niche is a chartered digital-bank challenger that can use a consumer app as both a deposit-gathering machine and a credit-distribution machine. It takes profit pools from traditional banks through lower physical distribution costs and from marketplace lenders through direct deposit funding, and it competes with brokerages for financial engagement.
The most dangerous future competitor is not necessarily another neobank. A scaled bank that combines cheap deposits with modern digital acquisition can attack SoFi’s funding and loan economics; an asset-light platform can attack engagement without assuming credit risk; and loan buyers can simply demand higher yields if consumer credit weakens.
Industry cyclicality here is multidimensional. SoFi is exposed to the macro cycle through borrower losses, the rate cycle through NIM and refinance demand, the capital-market cycle through loan-sale pricing, and the policy cycle through student lending and crypto. Personal lending is the most fragile in a recession. Student and home refinancing are more rate-sensitive, LPB is exposed to investor risk appetite, and Technology Platform runs on an enterprise technology cycle but currently lacks enough growth to diversify the other cycles materially.
The regulatory perimeter is broad. The Federal Reserve supervises the holding company; the OCC supervises the national bank; consumer products remain subject to federal consumer law and CFPB authority where applicable; deposits carry FDIC regulation/insurance considerations. Crypto has moved from a special supervisory posture toward normal bank supervision after the Fed’s 2025 withdrawals and sunset of its novel-activities program. That “normal” supervision still requires capital, liquidity, operational resilience, AML and consumer controls.
Current Fundamentals and Market Debate
Q2 2026 was operationally excellent.
| Q2 2026 measure | GAAP | Adjusted | YoY / interpretation |
|---|---|---|---|
| Net revenue | $1,218.7m | $1,205.6m | adjusted +40% |
| Net income | $156.6m | $160.4m | adjusted +65% |
| EBITDA | n/a | $357.8m | +44%; 30% margin |
| Diluted EPS | $0.12 | company FY guide about $0.60 | Q2 beat consensus |
| Members | company-defined 15.8m | n/a | +35% |
| Products | company-defined 24.4m | n/a | +42%; Invest definition changed in Q1 |
At the revenue and net-income level, the Q2 GAAP/non-GAAP reconciliation is modest.
| Reconciliation item | Q2 2026 |
|---|---|
| GAAP net revenue | $1,218.7m |
| Aggregate specified revenue adjustments | $(13.1)m |
| Adjusted net revenue | $1,205.6m |
| GAAP net income | $156.6m |
| Aggregate net adjustments | +$3.8m |
| Adjusted net income | $160.4m |
| Adjusted EBITDA | $357.8m |
Adjusted net revenue excludes specified fair-value changes in servicing rights/residual interests and gains or losses on debt extinguishment, among other defined adjustments. Note what it leaves in: the adjustment does not remove all fair-value movements on loans held at fair value, and those movements stay embedded in the loan-origination/sales line.
That leads to an accounting issue central to SoFi.
At June 30, 2026:
| Fair-value loan portfolio | Personal | Student | Home |
|---|---|---|---|
| Unpaid principal | $26.10bn | $16.13bn | $2.07bn |
| Accumulated interest | $0.181bn | $0.082bn | $0.009bn |
| Cumulative fair-value adjustment | +$1.223bn | +$0.705bn | +$0.100bn |
| Carrying fair value | $27.51bn | $16.92bn | $2.18bn |
Personal and home loans are classified as held for sale; student loans at fair value are classified as held for investment.
The total $2.03 billion fair-value adjustment is about 4.6% of unpaid principal. That is a cumulative balance-sheet valuation premium, not current-quarter income. A common analytical mistake is to read the change in that balance as quarterly mark-to-market profit. Sales, charge-offs, new originations and repayments all change the portfolio composition, which makes the shortcut invalid.
The earnings release never presents the quarterly amount of pure loan fair-value “gain” as a clean stand-alone number. SoFi combines initial and subsequent fair-value changes with loan origination, sale, securitisation and servicing economics in non-interest income. The identifiable Lending non-interest revenue pool in Q2 was roughly $116 million on an adjusted basis after subtracting Lending NII from Lending revenue, and that is an upper bound on the pure mark, since it also contains realised sales and servicing economics. A more precise fair-value-gain number would be false precision.
None of this makes SoFi’s earnings fictional. Cash interest is large: consolidated NII was $788 million in Q2, and fee-based revenue was $472 million. The fair-value-sensitive lending non-interest pool is a minority of total revenue. What it does mean is that changes in discount rates and expected credit performance can alter reported lending revenue before the underlying cash losses fully materialise.
Credit currently looks better, not worse. Reported annualised personal-loan NCO declined from 3.03% in Q1 to 2.62% in Q2, and management’s all-in measure was about 3.7%. Student-loan charge-offs were roughly 0.61%. The company’s 90-day delinquency trend was broadly stable year on year: $104.0m of personal loans were 90 or more days past due at 30 June 2026, close to 0.4% of personal-loan unpaid principal.
The fair-value model still turned slightly more conservative on personal defaults in Q2. The weighted-average annual default assumption rose to about 4.77% from 4.57% in Q1; the discount rate rose to 4.97% from 4.61%, with management attributing most of that increase to a 37-basis-point rise in benchmark rates. Personal-loan coupon declined slightly, while prepayment assumptions increased.
Keep annual charge-offs separate from the company’s 7–8% life-of-loan tolerance. A 3.7% annualised all-in NCO rate cannot be compared directly with an 8% cumulative lifetime loss ceiling. Personal loans amortise quickly, so the relevant question is whether cumulative vintage losses stay within underwriting expectations. The evidence available through Q2 says they have. The recession evidence remains absent.
Capital gives SoFi room for error, though balance-sheet growth is consuming it.
| Capital measure | 2025-12-31 | 2026-06-30 | Regulatory reference |
|---|---|---|---|
| SoFi Technologies CET1 | 22.8% | 18.7% | 7.0% required incl. buffer |
| SoFi Technologies total capital | 22.9% | 18.8% | 10.5% required incl. buffer |
| SoFi Technologies leverage | 18.8% | 16.5% | 4.0% |
| SoFi Bank total capital | 16.6% | 15.3% | 10.0% well-capitalized benchmark |
| Consolidated RWA | $37.2bn | $48.7bn | n/a |
By themselves the declining capital ratios are not a credit warning. They show SoFi deploying the large capital base raised in 2025 into rapid asset growth, and they explain why a genuinely capital-light LPB matters: maintaining 20%+ balance-sheet growth indefinitely would require retained earnings or further external capital.
Technology Platform is the clearest fundamental disappointment. H1 2026 revenue was roughly $160 million against about $213 million in H1 2025, down approximately 25%, and contribution margin fell from around 30% to roughly 15%. Enabled-account recovery of two million sequentially in Q2 is encouraging but far too small to establish a turn.
Financial Services is the opposite. Q2 revenue of $466 million rose sharply, NII and non-interest revenue were almost evenly balanced, interchange increased about 55%, annualised spend across Money and card products reached approximately $28 billion, and brokerage revenue was roughly 2.5 times the prior-year level.
This is why member growth has genuine economic content, even though it should not be treated as a profit metric. More members are generating deposits, spend, brokerage activity and referred-loan revenue. The stronger test is revenue and contribution profit per acquired member, not the member count by itself.
Q2’s stock sell-off was predominantly an expectations event. Management lifted full-year adjusted net revenue to $4.75–4.85 billion, from the earlier outlook around the mid-$4.6 billion range, but kept roughly $1.6 billion adjusted EBITDA, $825 million adjusted net income and $0.60 adjusted EPS. The market got more revenue and no more expected 2026 profit.
That repeated the Q1 pattern. SoFi had just delivered Q1 adjusted revenue growth of 41%, adjusted EBITDA growth of 62% and record originations, yet the stock fell roughly 12% when management left full-year targets unchanged. By Q2, investors had learned to look past the quarterly beat and ask whether it flowed through to annual earnings.
The direction of the Q2 sell-off contained information: incremental revenue was not being converted into a higher near-term profit forecast. The magnitude was amplified by valuation and a market conditioned to expect upward revisions.
Management’s defence is rational: investing through a period of rapid member acquisition can maximise longer-term economics. The burden of proof rises at a 30-times adjusted EPS multiple, because shareholders are already paying for a large portion of that future conversion.
The bull evidence is specific. Member growth remains 35%; deposits exceeded $45 billion; LPB has become a roughly $140 million quarterly fee engine; Financial Services is profitable at scale; credit is improving; capital is well above minimums; and consolidated NIM is close to 6%.
The bear case is just as specific. Technology Platform revenue is down more than 20%; the effective LPB fee ratio has slipped toward 4.5%; bank capital ratios are falling as assets expand; brokered/time deposits are growing faster; personal loans have never been tested at this scale through recession; and share count increased 19% from October 2024 through July 2026.
Valuation, Stress, Risk, and Tracking
The valuation framework starts with a deliberate refusal to use standard industrial-company free cash flow.
For a bank, reported operating cash flow is dominated by movements in loans, deposits and other financial assets. SoFi itself explains that cash can be used in operating activities when held-for-sale originations outrun loan sales and repayments. Treating those loan originations as conventional “cash burn” produces nonsense. A five-year OCF/net-income ratio changes sign and is economically unstable; it is not an appropriate valuation anchor.
Maintenance capex versus growth capex is similarly secondary. SoFi does not disclose a maintenance/growth split reliable enough to construct industrial-style owner earnings. The economically important reinvestment is the regulatory capital required to support risk-weighted assets. This report instead defines bank owner economics through tangible-book growth, return on tangible equity and the capital required for balance-sheet growth.
The magnitude is visible in H1 2026. Consolidated RWA increased roughly $11.4 billion in six months. Holding a 15% CET1 ratio against that incremental risk weight would require roughly $1.7 billion of capital before any internal RWA optimisation; H1 GAAP net income was only $323 million. SoFi started from an unusually large capital surplus after its 2025 equity raises, which is what allowed the ratio to fall from 22.8% to 18.7%. That surplus cannot fund the same rate of risk-asset growth indefinitely.
This is why a conventional “owner earnings yield” understates how much LPB matters. Capital-light fee growth lets revenue expand without the same proportional CET1 retention.
At $18.22, the simple headline valuation is:
| Metric | Current valuation |
|---|---|
| Price / Q2 tangible book value per share | 2.48x |
| Price / 2026 adjusted EPS guidance | about 30.4x |
| Market cap / 2026 adjusted net revenue midpoint | about 4.9x |
| Market cap / 2026 adjusted EBITDA target | about 14.7x |
| 2026 adjusted earnings yield | about 3.3% |
Inputs are the September 4 price, Q2 tangible book of $7.34 per share, adjusted EPS guidance around $0.60, adjusted revenue guidance midpoint of $4.80 billion and EBITDA guidance of about $1.6 billion.
The bank method should carry more weight than EV/EBITDA. Deposits are an operating funding source, so enterprise value calculations obscure the core economics rather than illuminate them. Adjusted EBITDA is useful for understanding how the market currently talks about SoFi. Price-to-tangible-book against sustainable ROTCE is the better valuation discipline.
Current tangible-book valuation is neither at the euphoric end of SoFi’s recent range nor bank-like. The stock has fallen more than 40% from the 52-week high while tangible book has kept rising, compressing P/TBV materially from its late-2025 peak. The short post-charter history rules out a statistically reliable percentile. Call it the middle-to-upper portion of the post-charter range, and far above mature consumer-bank levels.
The expectation embedded in 2.48 times tangible book is demanding. Using the simplified relation:
P/TBV = (sustainable ROTCE − long-term growth) / (cost of equity − long-term growth)
and assuming a 10.5% cost of equity with 4% perpetual growth, current valuation corresponds to about 20% sustainable ROTCE. A 12% sustainable return supports only about 1.23 times tangible book under those assumptions; 15% supports about 1.69 times; 18% supports roughly 2.15 times.
This is not a literal price-target formula. It identifies the expectation gap. SoFi’s 2026 adjusted net-income guidance of roughly $825 million, set against average 2026 tangible book around the low-$9 billions, points to an adjusted ROTE around 9%. The market is paying for that return to approximately double over time.
Nu explains why such a premium can eventually be deserved; its Q2 ROE was 33%. Capital One explains why a standard bank multiple is lower. SoFi occupies the uncomfortable space between them: much faster growth than a mature bank, much lower current tangible returns than a proven high-ROE digital bank.
My 12-month valuation scenarios use bank logic first and fintech logic as a cross-check.
| Dimension | Conservative | Base | Optimistic |
|---|---|---|---|
| Revenue / margin assumptions | growth drops toward high-teens/low-20s; Tech stays weak; LPB moderates | revenue growth stays mid/high-20s; Tech stabilises; EBITDA margin low-30s | around 30% growth persists; Tech reaccelerates; fee mix improves materially |
| Tangible-book / return assumption | forward TBV/share about $7.8–8.1; sustainable ROTCE 10–12% | TBV/share about $8.1–8.5; ROTCE moves toward 15–17% | TBV/share about $8.5–9.0; ROTCE reaches 19–21% |
| Bank multiple assumption | about 1.5–1.7x forward TBV | about 1.9–2.2x | about 2.5–2.8x |
| 12-month fair-value area | roughly $12–14 | roughly $16–19 | roughly $22–25 |
| Fintech cross-check | 2027 earnings multiple compresses to high-teens/low-20s | mid/high-20s earnings multiple | 30x+ justified by sustained 30% growth |
| Key catalyst | credit stays inside tolerance | fee growth + margin conversion | 20% ROTCE path becomes credible |
| Key risk | loan losses / buyer retreat | Tech remains a drag | expectations already near perfect |
| Three-year value case | about $12–13 | about $22–24 | about $34–36 |
| Implied three-year annualised return from $18.22 | about -13% to -11% | about +6% to +10% | about +23% to +25% |
| Permanent-loss risk | stress erodes TBV and multiple | returns stall below cost of equity | high-growth multiple collapses before ROTCE arrives |
This is valuation-scenario analysis within a research framework, not investment advice.
The conservative scenario is not a recession. It assumes SoFi remains profitable but earns only a modest premium to its cost of equity. In the base scenario, management closes roughly half the gap between current tangible returns and the return implied by today’s stock price. The optimistic scenario requires the market’s central thesis to become operational reality.
The stress case is more severe: double SoFi’s own 7–8% maximum expected cumulative personal-loan loss assumption.
At Q2, personal-loan unpaid principal was $26.10 billion. Take 7.5% as the midpoint of management’s tolerance and double it to 15%: that is roughly $1.96 billion of incremental lifetime loss on the current principal balance in a purely mechanical stress. A 22% tax rate gives an illustrative after-tax equity impact around $1.53 billion.
Against roughly $9.5 billion of tangible book, that would remove about 16%. Against $48.7 billion of consolidated RWA, an immediate $1.53 billion CET1 hit would cut the consolidated CET1 ratio by roughly 3.1 percentage points before second-order RWA changes, from 18.7% toward the mid-15% area. The bank-level ratio would have less room, since its total capital ratio starts at 15.3%. This is deliberately severe and not a forecast; actual fair-value marks would depend on timing, prepayments, discount rates, portfolio turnover and tax treatment.
Fair-value accounting would transmit the shock earlier than cash charge-offs. Personal loans currently carry a $1.22 billion positive cumulative fair-value adjustment. A roughly $2 billion adverse pre-tax credit revaluation of the kind implied by the mechanical stress would more than erase that premium, before any offsets from coupon value, discount-rate movements and realised cash flows.
The second leg of the stress is LPB. Q2 loan-platform fees of $140.9 million annualise above $560 million. Suppose external loan demand falls 75% and effective fee economics compress 20%: annual fee revenue would drop by roughly $450 million. This is relatively capital-light revenue, so a large portion of that would come straight out of contribution profit. SoFi would then face a choice, either retaining more originations and consuming capital during a credit downturn, or cutting originations and losing revenue and member monetisation.
The doubled-loss stress does not look like an immediate insolvency scenario; it looks like a tangible-book, capital and valuation-reset scenario capable of cutting the equity value by roughly half.
That is the permanent-loss path investors should care about.
The margin-of-safety test is demanding. Current $18.22 is roughly 30–52% above the $12–14 conservative fair-value area. The margin of safety against the conservative case is zero.
The most fragile base-case assumption is the rise in sustainable ROTCE toward 15–17%. Cut the improvement to 70% of the modelled increase and sustainable returns land around 13–14%. A bank framework would then support closer to 1.5–1.7 times forward tangible book, or roughly $12–15 rather than $16–19.
A flat-earnings thought experiment is similarly unattractive. If earnings per share stay around $0.60 for three years and the stock finishes that period at a mature 20-times earnings multiple, the terminal price is roughly $12, a negative annualised return from $18.22. Even at the unchanged current 30-times multiple, the investor earns almost nothing from earnings growth, because there is none and SoFi pays no meaningful dividend. The return would be inferior to a normal long-duration government-bond hurdle unless multiple expansion rescued it.
Margin-of-safety sufficiency verdict: none.
The most important permanent-capital risks can be stated as observable pathways.
Credit is medium probability and high impact. The trigger is personal all-in annualised NCO moving above roughly 5.5% for two quarters, or vintage projections breaking the 7–8% cumulative tolerance. Deterioration first lowers fair values, then raises realised losses, then consumes bank capital. If the multiple moves toward conventional consumer-bank levels at the same time, the equity loss is much larger than the earnings loss.
Third-party loan liquidity is medium probability and high impact. Watch LPB originations, the effective fee-to-volume ratio and loan-sale/securitisation spreads. A retreat by buyers removes high-margin fee revenue and can force SoFi either to retain more credit when capital is least attractive or to reduce origination volume.
Funding is medium probability and medium-to-high impact. The indicators here are NIM, average deposit cost, brokered deposits and time-deposit growth. The deposit advantage remains large, but Q2 showed deposit rates falling less than warehouse rates, while brokered deposits had increased sharply. A shrinking deposit advantage cuts NII and undermines one of the charter’s largest economic benefits.
Technology Platform underperformance is medium probability and medium impact on current earnings, higher impact on valuation. Revenue is already down 23% and contribution margin 14%. Continued contraction would not threaten the bank, but it would remove the main evidence for a software multiple.
Dilution is medium probability and high long-term per-share impact. Basic shares rose 19% from October 2024 to July 2026. Future raises done well above tangible book can still be economically accretive; repeated issuance nonetheless transfers a growing share of corporate value to new holders and employees. The correct indicator is diluted EPS and TBV per diluted share, not total net income.
Valuation compression is high probability over a multi-year holding period and medium-to-high impact. A stock priced at 2.5 times tangible book while earning a single-digit adjusted ROTE has little protection if the market concludes that 15–20% sustainable returns are unreachable.
Positive catalysts are equally concrete. LPB fees holding above $140–150 million a quarter while total originations keep rising would establish that third-party distribution is structural. Technology Platform returning to positive year-on-year growth, with contribution margin above 20%, would restore a missing part of the thesis. Confidence in underwriting improves substantially if personal all-in NCO stays below 4% as unemployment weakens. A stable NIM near 6% while deposit growth remains predominantly retail would strengthen the charter economics, and sustained TBV-per-share growth without further large equity issuance would improve the per-share case.
Negative catalysts are guidance conversion failures, LPB buyer retreat, fair-value markdowns, deterioration in personal-loan vintages, a further shift toward expensive brokered/time deposits, another equity raise at a lower price, or Technology Platform remaining in contraction through 2027.
The tracking dashboard is:
| Indicator | Current / normal zone | Alert threshold | Timing |
|---|---|---|---|
| Adjusted net-revenue growth | 30%+ currently | below 20% YoY | quarterly |
| Adjusted EBITDA margin | about 30–34% | below 27% | quarterly |
| NIM | about 5.7–6.1% | below 5.5% | quarterly |
| Personal all-in annualised NCO | roughly 3.5–4.5% | above 5.5% for 2 quarters | quarterly |
| Late-stage personal delinquency | about 0.4% currently | above 0.8% | quarterly |
| LPB effective fee/volume | roughly 4.5–5.5% | below 4.0% | quarterly |
| SoFi Bank total capital | around 15% | below 12.5% | quarterly |
| Technology contribution margin | current 14%; desired >20% | below 15% for 2 more quarters | quarterly |
| Organic diluted-share growth | desired below 5% YoY | above 7% YoY | quarterly |
| Next earnings | expected 2026-10-27† | date change / guidance preannouncement | event |
† October 27 is an unconfirmed market estimate as of the research base date; SoFi had not yet formally confirmed the date.
Read the dashboard in combinations. A 5.3% NCO rate with stable LPB and 18% CET1 is manageable. The same 5.3% alongside falling LPB demand and a 13% bank capital ratio is a different investment.
Cross-Synthesis, Conclusion, Uncertainties, and Sources
Vertically, SoFi has proven three capabilities.
It proved, first, that a digitally acquired borrower relationship can be expanded into a much broader financial relationship. The company that listed in 2021 relied heavily on a lending identity; by Q2 2026 Financial Services alone produced $466 million of quarterly revenue and more than 15 million members sat inside the broader ecosystem.
Second, it proved that acquiring a bank charter could create economic value rather than merely regulatory prestige. Deposits are now more than 90% of average liabilities, and the deposit/warehouse funding gap produces hundreds of millions of annualised interest savings. That capability should persist unless SoFi loses its ability to gather attractive retail deposits.
And it proved that management can change the earnings model. Financial Services moved from an investment drag to a major contributor; LPB has created more than $140 million of quarterly adjusted revenue; the group has posted eleven consecutive profitable quarters.
Technology infrastructure has not been proven to the same standard. Galileo/Technisys remains an asset, but the 2021 idea that software infrastructure would become one of the dominant contributors to 2025 economics did not happen. The striking historical fact is that SoFi nearly reached the consolidated revenue promised in the SPAC presentation while missing the business mix behind that promise, with lending and Financial Services carrying more of the burden.
That matters because different revenue deserves different capitalisation. A recurring processing fee with limited credit exposure is worth a different multiple from a personal-loan spread whose cash flow depends on borrower defaults and deposit costs. Treating the two identically would overstate the durability of consolidated earnings.
Horizontally, SoFi’s advantage is unusual rather than absolute. Capital One has deeper cycle-tested credit and far greater scale, LendingClub offers a cleaner marketplace-lending model, Nu has already reached the high ROE that SoFi’s multiple assumes, and Robinhood has a more capital-light relationship with consumer engagement. SoFi’s differentiation is that it combines pieces of all four: nationwide deposits, prime unsecured lending, external loan distribution and a consumer financial app.
That combination is potentially powerful, because each piece can support another. Deposits lower funding costs. Lending generates NII and acquisition economics, Financial Services increases engagement, and LPB lets underwriting scale without equivalent capital. Technology can lower internal infrastructure costs and, eventually, add external fee economics.
The combination also creates a valuation trap. A conglomerate of several attractive activities does not automatically deserve the highest multiple available to any one of them.
The stock’s current price effectively assumes SoFi can raise tangible returns from around the single digits toward the high teens or approximately 20% while maintaining high growth. Several routes lead there: LPB can become a materially larger share of revenue; Financial Services can compound faster than balance-sheet assets; operating expenses can scale slower than revenue; Technology can recover; or SoFi can recycle loans faster and reduce capital intensity.
The present data supports the direction of travel, not the destination.
The fee shift is the clearest example. LPB has gone from a small operation to roughly $140 million quarterly, and Financial Services non-interest revenue is above $200 million quarterly. Yet Q2 NII grew 52% and represented roughly two-thirds of GAAP net revenue. SoFi is becoming a better bank and a more meaningful fee platform at the same time. The earnings base remains bank-dominated.
The credit question controls the three-to-five-year outcome. A recession that leaves all-in personal-loan NCO in the 4–5% range and lifetime vintages below 8% would do more for SoFi’s valuation credibility than another year of 30% member growth, because it would show that the bank’s underwriting engine survives stress. A recession that drives lifetime losses into the low-to-mid teens would produce fair-value marks, capital consumption and third-party loan-buyer retreat at the same time.
Capital is the bridge between those outcomes. The 2025 equity raises created a lot of capacity, and issuing stock at a large premium to tangible book was intelligent corporate finance. That capital is already being spent: consolidated RWA rose more than $11 billion in six months and CET1 fell from 22.8% to 18.7%. LPB must increasingly carry growth if management wants to maintain very high originations without returning repeatedly to the equity market.
The next 12 months hinge on credit, LPB and earnings conversion. Member additions are secondary. The most informative sequence would run: all-in NCO stays around 4% or lower, LPB remains above $3 billion of quarterly third-party volume, fee revenue grows faster again, Technology turns positive year on year, and management starts translating revenue beats into higher adjusted earnings.
Over three years, sustainable ROTCE is the decisive variable. A mid-teens return would make a roughly 2-times tangible-book franchise defensible. An 18–20% return combined with continuing double-digit tangible-book growth could justify today’s platform premium and potentially more. A 10–12% return would make the present valuation difficult to sustain regardless of member growth.
Over five years, the question is whether SoFi becomes a high-ROE bank with unusually good digital distribution or remains a fast-growing bank that needs large amounts of capital to keep growing. The former compounds per-share value. The latter can produce rapid revenue growth while ordinary shareholders receive much less of it, because new equity and retained capital fund the balance sheet.
What the market may be misjudging most is the difference between revenue quality and revenue quantity. Q2’s sell-off indicates investors have started demanding earnings conversion. That is healthy. The market may still underappreciate, though, how valuable LPB could become if it survives the next credit downturn. A capital-light loan-distribution business that can still originate several billion dollars per quarter during stress would materially alter SoFi’s sustainable ROTE.
The bull reasons are traceable:
- Deposits reached $45.5 billion and fund more than 90% of average liabilities, creating roughly $713 million of annualised savings versus warehouse funding.
- LPB produced $140.9 million of loan-platform fees on roughly $3.1 billion of third-party personal-loan volume in Q2, creating a large capital-light revenue stream.
- Q2 member growth was 35%, adjusted revenue growth 40%, adjusted EBITDA margin 30%, and GAAP profitability reached an eleventh consecutive quarter.
- Personal-loan realised credit improved in Q2, with reported NCO down to 2.62% and all-in NCO around 3.7%, while cumulative vintages remained within the company’s 7–8% lifetime tolerance.
- Consolidated CET1 of 18.7% leaves substantial capacity above required minimums even after rapid H1 balance-sheet growth.
The bear reasons are equally concrete:
- Current 2.48-times tangible book mathematically requires roughly 20% sustainable ROTCE under reasonable bank assumptions, while 2026 adjusted return on tangible equity is still around the high-single digits.
- Technology Platform revenue fell 23% and contribution margin fell to 14%, weakening the software component of the valuation case.
- The public credit history does not include a conventional recession at the present $26 billion personal-loan principal scale; doubling the company’s lifetime-loss tolerance creates an illustrative near-$2 billion pre-tax credit hit.
- Basic shares outstanding increased approximately 19% from October 2024 to July 2026, and convertible/equity-compensation dilution remains.
- The company raised 2026 revenue guidance after Q2 without raising adjusted EPS or EBITDA, evidence that incremental revenue is not yet flowing directly into incremental near-term shareholder earnings.
The first pre-mortem script is a 2027 consumer-credit shock. Unemployment rises enough to push SoFi’s all-in personal NCO above 6%, and projected lifetime losses move from the 7–8% underwriting range toward 11–12%. A $1 billion-plus fair-value markdown hits tangible equity, LPB buyers demand much higher yields and third-party originations halve. Bank total capital falls toward 11–12%, forcing SoFi to slow retained lending. Sustainable ROTCE expectations fall from the high teens to 10–12%, and P/TBV compresses from around 2.5 times toward 1.2–1.4 times. With stressed TBV per share around $6–7, the shares could trade in the $7–10 range.
The second script is quieter operationally and just as damaging to the stock. Credit stays acceptable in 2027–28, but Technology Platform never recovers, LPB fee rates slip below 4%, brokered deposits keep rising and diluted shares grow another 10–15%. SoFi still reaches several billion dollars more revenue, yet sustainable ROTCE stalls around 12–13%. Investors stop treating it as a platform and value it around 1.5 times tangible book. A future $9–10 TBV per share would then support only $13–15, a poor multi-year return out of a materially larger company.
My final view rests on a simple distinction. SoFi has become a materially better business than the one that came public in 2021. The bank charter works, deposits create a measurable cost advantage, Financial Services has moved into real profitability, LPB is large enough to matter, and management has built a deep layer of regulatory capital. The remaining objections concern price and proof through a credit cycle rather than the viability of the franchise.
At $18.22, shareholders are being asked to pay today for a future high-ROTE version of SoFi. The base case can justify roughly the current price if fee businesses keep scaling and tangible returns move into the mid-teens. The conservative case is materially lower. The optimistic case requires a 19–21% sustainable ROTCE profile that has not yet been earned. That asymmetry leaves little protection against a credit or execution error.
【Company-profile scores】
- Fundamental quality: medium
- Growth: high
- Moat: medium
- Financial soundness: strong
- Management credibility: high
- Valuation attractiveness: medium
- Risk level: high
- Suitable investor type: long-term growth
【Investment rating】
- Rating: Hold
- One-line thesis: Fee growth and cheap deposits are real, but $18.22 already requires a much higher sustainable ROTCE than SoFi earns today.
- Current-price classification: acceptable hold
- Whether to wait for a better price: yes. A purchase becomes substantially more attractive around $9.50–11.00 if credit remains inside the 7–8% lifetime-loss framework, bank capital stays comfortably above minimums and LPB remains structurally intact. The opportunity cost is that SoFi could reach mid-teens ROTCE without such a drawdown, leaving a patient investor underexposed to a genuine re-rating.
- Target holding horizon: 3–5 years
- Expected annualized return: conservative approximately -13% to -11%; base approximately +6% to +10%; optimistic approximately +23% to +25%, using the three-year scenario values above.
- Max-loss risk: roughly 45–60% in the combined credit/LPB/capital pre-mortem, with a stressed $7–10 share-price region.
- Reassessment-trigger signals: personal all-in NCO above 5.5% for two consecutive quarters; modeled vintage losses exceeding the 8% lifetime ceiling; SoFi Bank total capital below 12.5%; LPB effective fee/volume below 4% combined with a 25%+ decline in third-party volume; Technology Platform contribution margin remaining below 15% through another two reporting periods.
【Ideal Buy Price】9.50–11.00 USD
Basis: at least a 20% discount to the $12–14 conservative 12-month fair-value estimate, subject to credit and capital remaining fundamentally intact.
Acceptable hold price: 16.00–19.00 USD, corresponding to the base case in which sustainable ROTCE rises toward the mid-teens.
Clearly overvalued price: 28.00–32.00 USD, beginning more than 10% above the optimistic $22–25 fundamental value area and requiring the high-ROTE outcome before it is demonstrated.
【Valuation Range】
- current: 18.22 (close as of 2026-09-04)
- bear (conservative · ideal buy zone): [9.50, 11.00]
- base (fair · acceptable hold zone): [16.00, 19.00]
- bull (optimistic · above the clearly-overvalued line): [28.00, 32.00]
Research uncertainties remain material. First, fair-value loan accounting discloses no single clean quarterly number for pure loan mark-to-market gains, so separating marks from realised loan-sale and servicing economics beyond the disclosed revenue categories would require management-level loan-accounting detail.
Second, SoFi has no recession-scale public credit record comparable with Capital One or older card issuers. Vintage seasoning is encouraging. It is not a substitute for unemployment-driven stress.
Third, the comparable public disclosure does not identify the large Technology Platform client that offboarded. That limits analysis of whether the loss was company-specific, contract-specific or a broader threat to Galileo’s competitive position.
Fourth, October 27, 2026 is an estimated rather than company-confirmed next earnings date as of the base date.
Fifth, the $23.53 billion market cap deliberately uses the company’s latest disclosed basic share count. Market-data vendors can report a higher figure by applying a more current or diluted denominator, and the diluted economic claim on the company is indeed larger than basic market capitalisation, because of employee awards and convertibles.
The most important primary sources for this report are SoFi’s Q2 2026 Form 10-Q and SEC-filed earnings release, the Q1 2026 and four 2025 quarterly releases, the 2025 Form 10-K, Federal Reserve documentation of the 2022 bank transaction and subsequent digital-asset supervisory changes, and Department of Education releases covering the 2025–26 federal student-loan regime.
For market-expectation and price-reaction attribution, I used SoFi’s own stock information, historical quote data and Reuters reporting on Q1/Q2 2026. The Reuters Q2 report is especially useful, because it identifies unchanged $0.60 EPS guidance as the issue TD Cowen raised despite the revenue-guidance increase.
Peer operating data comes primarily from Q2 2026 releases from Capital One, Nu Holdings, LendingClub and Robinhood rather than from third-party screening databases.
Other tickers mentioned
- COF.US: cycle-tested U.S. consumer-bank and unsecured-credit benchmark.
- LC.US: closest listed U.S. chartered-bank and marketplace-personal-loan analogue.
- NU.US: high-growth digital-bank benchmark showing the ROE required to sustain a premium-to-book valuation.
- HOOD.US: capital-light consumer-finance platform benchmark for member engagement and growth valuation.
This report is based on public information and does not constitute investment advice. Markets carry risk; invest with caution.
Vollständige Analyse
Zum Lesen der vollständigen Analyse anmelden
Registriere dich kostenlos, um den vollständigen Text, die Baillie-Wachstumsscorecard und die Volltextsuche freizuschalten.
Anmelden / Kostenlos registrieren