Crypto & Digital Assets
A profitable, fast-growing digital bank mispriced as a crypto name and now mispriced again by fear — but the entire bull/bear hinges on one binary question the market cannot yet resolve: are the +$2.0B of fair-value loan marks real, or is Muddy Waters right?
Research
The verdict
A profitable, fast-growing digital bank mispriced as a crypto name and now mispriced again by fear — but the entire bull/bear hinges on one binary question the market cannot yet resolve: are the +$2.0B of fair-value loan marks real, or is Muddy Waters right?
SoFi (founded 2011; IPO'd via SPAC June 2021; Nasdaq: SOFI) is a digital-first, member-centric financial-services platform that became a bank holding company in Feb 2022 and operates SoFi Bank, N.A. (a nationally chartered bank). The thesis the whole company is built on is the "Financial Services Productivity Loop": acquire a member cheaply with one product, cross-sell more, drive lifetime value up and acquisition cost down. It reports in three segments:
Business model in one line: SoFi makes money three ways — (1) net interest income on a deposit-funded loan book ($2,219.0M FY25, 61% of revenue), (2) fee-based revenue ($1.5B FY25, +59% — origination, Loan Platform, interchange, referral, brokerage), and (3) subscription/usage fees from the tech platform. The strategic engine is the bank charter: deposits ($37.5B at YE25, $40.2B at Q1'26) fund loans at a lower cost than warehouse/securitization, and give SoFi the optionality to hold loans for NII or sell them for gain-on-sale.
Customers / suppliers / competitors. Customers = 14.7M retail "members" (Q1'26) + enterprise clients of the tech platform (e.g. Banco Nación of Argentina on Cyberbank; multiple US consumer brands). "Suppliers" of a bank are its funding sources: retail deposits, warehouse lenders (Goldman Sachs admin agent on the revolver), and capital-markets buyers of its loans (large financial institutions, incl. a large JPMorgan loan-platform relationship). Competitors span three fronts: neobanks/fintech (Chime, Cash App/Block, Robinhood, Affirm), incumbent digital banks (Ally, Discover, Capital One, Marcus), and in the tech platform, Marqeta / Fiserv / FIS / Galileo's own former clients building in-house. Contract structure is a mix: lending is transactional (gain-on-sale or hold), tech-platform is multi-year, volume-based, often with minimum monthly payments — but, critically, not take-or-pay enough to stop a large client from leaving (see Lens 5).
A bank's "supply chain" is its funding-to-asset-to-distribution flow. Named stakeholders along it:
Upstream (funding inputs):
Midstream (SoFi itself): originates → underwrites (proprietary risk models on willingness/ability/stability) → either holds (LHFI at fair value $15.3B; at amortized cost $1.4B) or sells (LHFS $25.5B at Q1'26).
Downstream (distribution / loan buyers):
Chokepoints / single-source dependencies: (1) The bank charter is the keystone — it is what makes deposit funding (and thus the whole cost-of-funds advantage) possible; regulatory loss of it would break the model. (2) The Loan Platform Business is concentrated in a few capital-markets buyers — the JPMorgan transaction is large enough that Muddy Waters built a thesis around it (Lens 10/13). (3) The tech platform had client concentration severe enough that one departing client cut segment revenue 27% YoY (Lens 5) — a textbook single-source dependency.
Where the moat is real:
Bargaining power: Strong over members (switching costs rise as they hold more products; direct deposit is sticky). Weak over capital-markets loan buyers and tech-platform clients — both can walk, and one tech client did. Weak vs. deposit competition — SoFi's deposit growth is bought with high APYs, so the "cheap funding" advantage is real but not free.
Verdict on moat: A genuine but mid-tier moat — narrower than the bulls' "next-gen money-center bank" framing, wider than the bears' "subprime lender in a fintech costume." The charter is the moat; everything else is execution.
Every figure.
| Segment | FY25 net rev | FY25 YoY | FY25 contrib. profit | Q1'26 net rev | Q1'26 YoY | Q1'26 contrib. profit |
|---|---|---|---|---|---|---|
| Lending | $1,848.9M | +24% | $1,016.9M (55% margin) | $642.4M | +55% | $382.4M (60% margin) |
| Financial Services | $1,542.0M | +88% | $792.9M | $428.5M | +41% | $195.6M |
| Technology Platform | $450.2M | +14% | $144.4M | $75.1M | −27% | $12.0M (−61%) |
| Total net revenue | $3,613.4M | +35% | — | $1,100.4M | +43% | — |
Trend & cause:
Geography: Predominantly US. International (Latin America, Canada, Switzerland, Hong Kong) runs largely through the Technology Platform segment (e.g. Banco Nación, Argentina) — and Argentina triggers highly-inflationary FX adjustments in adjusted EBITDA.
All unless noted.
| Metric | Q1'26 | Q1'25 | YoY |
|---|---|---|---|
| Total net revenue | $1,100.4M | $771.8M | +43% |
| Net interest income | $693.0M | $498.7M | +39% |
| Noninterest income | $407.4M | $273.0M | +49% |
| Net income (GAAP) | $166.7M | $71.1M | +134% |
| Diluted EPS | $0.12 | $0.06 | +100% |
| Adjusted EBITDA | $339.9M | $210.3M | +62% |
| Adjusted net revenue | $1,087.2M | $770.7M | +41% |
| NIM | 5.94% | 6.01% | −7bps |
| Total deposits | $40.2B | — | +7% q/q |
Beat/miss vs. consensus: Revenue and EPS beat; the stock fell ~12–13% anyway. The miss was in the forward guide and segment mix, not the print.
What drove it: Lending (+55%, origination +68%) and Financial Services (+41%) more than offset the Galileo collapse (−27%). Fee-based revenue $386.8M (+23%). Loan Platform added $140.8M to adjusted net revenue across two segments.
Margins: GAAP net income margin 15%; adjusted EBITDA margin 31%. The Adjusted EBITDA trajectory is a clean up-and-to-the-right: $210.3M → $249.1M → $276.9M → $317.6M → $339.9M over the trailing five quarters.
Guidance / tone: Management reaffirmed (did not raise) FY26 guidance — adjusted net revenue ~$4.655B (+30%), adjusted EBITDA $1.6B (34% margin), adjusted EPS ~$0.60. A reaffirmation after a beat read as a de facto downgrade to a market that wanted a raise — and the Tech Platform must nearly triple from a $75M quarterly run-rate to hit ~$325M FY guidance, which the market does not believe.
Balance-sheet flags (the important part):
Market reaction tells you: the market is no longer trading SoFi on the print — it is trading the quality of the print (the Muddy Waters overhang) and the durability of growth (Galileo, capital intensity, credit).
No transcripts on the research shelf (transcripts=0); this lens is ``.
Management tone across the FY25→Q1'26 calls has been relentlessly, almost defensively, upbeat — Noto frames every quarter as "the strongest in the company's history" (the exact phrase appears verbatim in both the 10-K and 10-Q MD&A). The recurring phrases: "Financial Services Productivity Loop," "durable growth," "fee-based revenue," "diversified funding," "record." The new vocabulary in late 2025/2026: "stablecoin," "SoFiUSD," "blockchain remittance," "first national bank to issue a stablecoin on a public permissionless blockchain" — a deliberate pivot to attach SoFi to the crypto/stablecoin narrative just as that theme re-rated. The thing they stopped emphasizing: the Technology Platform as a growth engine — for two years it was the "diversification" story; now it's discussed in damage-control terms ("exit of a large client").
Sentiment read: Management conviction is high and consistent; what changed is the market's willingness to take the optimism at face value, post-Muddy Waters. A promotional CEO + a short report alleging promotion-via-accounting is a combustible combination (Lens 9, 13).
| Company | Ticker | Mkt cap | Fwd P/E | EV/Sales | Notes |
|---|---|---|---|---|---|
| SoFi | SOFI | ~$20B | ~26x (Feb-2026 prints cited "~50x" on trailing) | n/a (bank, EV/sales not meaningful) | NIM 5.94%, ROE ~4.6% [calc] |
| Ally Financial | ALLY | n/a | ~17x | n/a | Incumbent digital bank; "valuation cushion" vs SoFi |
| Robinhood | HOOD | n/a | n/a | n/a | Brokerage-led fintech peer |
| Affirm | AFRM | n/a | n/a | n/a | BNPL lender peer, often unprofitable |
| Credit Services industry median | — | — | ~9.4x | — | SoFi trades ~179% above median |
Read: On a forward-P/E basis SoFi at ~26x is expensive vs. Ally (~17x) and the credit-services median (~9.4x) but is being priced as a 30%+ grower, not a bank. The bull says "it's a fintech compounding at 30% — 26x is cheap"; the bear says "it's a fair-value-marked consumer lender — it should trade like Ally, i.e. ~35% lower." The comps cannot settle this because the disagreement is about which industry SoFi belongs to. ROE of ~4.6% (FY25 net income $481.3M / ~$10.5B equity ) is low for a bank and undercuts the "premium franchise" case — though it is depressed by the fair-value accounting and reinvestment.
Mostly ; the IPO-to-now total-return is .
The 5-year tape, indexed (June-1-2021 = $100): SoFi $100 → $70.42 (YE21) → $20.35 (YE22) → $43.93 (YE23) → $67.99 (YE24) → $115.58 (YE25), vs Nasdaq Composite $175.08 and S&P Financials $156.06 over the same window. SoFi has badly lagged both its index and the financials sector since IPO — the YE25 strength has since reversed.
What moves the stock:
Pattern: Pre-2026, SoFi reacted to macro (rates) and milestones (profitability). Post-March-2026 it reacts to one thing: the credibility of its accounting. The market has stopped paying for growth until the Muddy Waters question is settled.
CEO Anthony Noto — the dominant figure. Ex-Goldman Sachs (TMT banker, took Twitter public), ex-CFO of Twitter, ex-NFL CFO; CEO of SoFi since 2018.
The one fact that cuts hardest in his favor: Noto bought stock in the open market three times in 2026 — 56,000 shares @ ~$17.9 (Mar 2), 15,545 @ ~$16.0 (May 11, right after the Q1 drop), and 13,888 @ ~$18.06 (Jun 16) — and made no open-market sales all year. A CEO accused of accounting games who is buying his own stock with cash into the teeth of a short report is making a costly, falsifiable signal. Weigh it accordingly.
Acting as a forensic equity analyst. This is the center of gravity of the entire dossier.
The mechanism that creates the risk — fair-value loan accounting. SoFi elects the fair-value option on its personal, student and home loans. As of Q1'26:
| Loan type | Unpaid principal | Total fair value | Cumulative FV adjustment |
|---|---|---|---|
| Personal | $22,317.9M | $23,682.4M | +$1,203.0M (≈ +5.4% premium to par → the "105.4% mark") |
| Student | $14,510.6M | $15,336.8M | +$756.9M |
| Home | $1,562.3M | $1,648.0M | +$78.7M |
| Total | $38,390.9M | $40,667.2M | +$2,038.7M |
So SoFi carries its loan book at a ~$2.0B premium to unpaid principal, and the change in that mark each period flows through noninterest income (gain on the way up, loss on the way down). This is GAAP-permitted, but it means a meaningful slice of reported revenue/EBITDA is non-cash, model-driven mark-to-model on illiquid consumer loans — exactly the structure that flatters earnings in a benign credit environment and reverses violently in a bad one.
Where cash flow diverges from earnings: the +$2.0B mark is unrealized; the cash test is whether loans sell at or above the marks. SoFi's defense (which the filings support) is that in Q1'26 it sold $3.8B of loans and execution was at or above balance-sheet marks; the personal-loan mark declined 27bps q/q (discount-rate driven) — i.e. they are not aggressively marking up.
Where credit is deteriorating (the bear's strongest on-filing number): net charge-offs on transferred personal loans rose to $229.7M in Q1'26 from $128.9M in Q1'25 (+78%). Delinquent (90–120 day) personal loans are marked down ~82% (FV $24.0M on $105.6M UPB). The fair-value model marks down at 10/30/60/90 days — so the question is whether the model's default assumptions keep pace with a worsening consumer.
SBC flattering non-GAAP: $262M SBC FY25 (+$234M D&A) is added back to reach the $1,053.9M adjusted EBITDA. Standard, but material — adjusted EBITDA is >2x GAAP net income.
Off-balance-sheet structures: 23 nonconsolidated VIEs + 1 consolidated VIE (securitization trusts). SoFi states VIE creditors "have no recourse against our general credit" and its exposure is limited to retained residual/securitization investments ($202.2M). MW characterizes these as "Enron-esque off-balance-sheet structures."
The Muddy Waters report (the must-surface item) — March 17, 2026:
Regulatory findings (required sub-section):
Forensic verdict: SoFi is not a fraud on the evidence available — the marks are GAAP-permitted, the SEC has filed nothing, the auditor signed off, the CEO is buying. But the bear has identified the real vulnerability: a consumer-loan book carried +$2.0B above par with charge-offs up 78%, where a meaningful chunk of "EBITDA" is mark-to-model. This is a credit-cycle bet dressed as a fintech-growth bet. The honest position: the accounting is aggressive-but-legal, and its truthfulness will be revealed by credit, not by argument.
Built bottom-up from FY25 actuals + management FY26 guide. Every input labeled; output ``. No forecast.ts logged (watchlist breadth mode — per skill).
Anchors: FY25 GAAP EPS $0.39, FY25 adj net rev $3,591.4M, adj EBITDA $1,053.9M. FY26 guide: adj net rev ~$4,655M (+30%), adj EBITDA $1.6B, adj EPS ~$0.60.
| Path | FY26E adj EPS | FY27E adj EPS | FY28E adj EPS | Key assumptions |
|---|---|---|---|---|
| Bull | $0.62 | $0.92 | $1.30 | Hits guide; Lending +30%, FS +35%, Galileo stabilizes & re-grows; credit benign (marks hold); modest dilution (~+5%/yr). |
| Base | $0.60 | $0.78 | $0.98 | Meets FY26 guide; Lending +25%, FS +30%, Galileo flat-to-down; charge-offs rise but model keeps pace; dilution ~+6%/yr drags EPS ~3pts below pretax growth. |
| Bear | $0.50 | $0.48 | $0.42 | Credit normalizes → fair-value marks reverse (a one-time non-cash hit to noninterest income), Loan Platform sale prices compress, Galileo keeps bleeding, capital intensity forces balance-sheet retention (less gain-on-sale), dilution continues. EPS stalls then falls despite revenue growth. |
The projection's load-bearing variable is not revenue growth — it is the direction of the fair-value marks. In the base/bull, the +$2.0B premium is sustained and EPS compounds. In the bear, a credit turn forces mark reversals that flow straight through noninterest income, and EPS can fall while revenue rises — the nastiest kind of miss.
Brier forecast (for later logging, not created here): "SOFI FY26 adjusted EPS ≥ $0.58, resolves 2027-02-28" — p ≈ 0.62 (management reaffirmed it post-Q1 with a beat; the risk is a 2H credit/marks event, not operating shortfall).
Bull case (narrative). SoFi is the first genuinely new money-center bank in a generation — a digital-native, charter-holding institution compounding revenue at 30%+, now GAAP-profitable and self-funding through a $40B deposit base. The Financial Services segment has crossed into real profitability ($793M contribution, from a loss two years ago), the cross-sell loop is empirically lifting products-per-member, and the stablecoin/crypto optionality (first national bank to issue a permissionless-chain stablecoin under the GENIUS Act) is a free call option on the single hottest regulatory tailwind in finance. At ~$18 / ~26x forward, with a CEO buying stock and a beat-and-reaffirm quarter, the market is handing you a 30% grower at a price that bakes in a fraud that hasn't been proven. The contrarian point the market refuses to see: the Muddy Waters report may be the best thing that happened to SoFi shareholders — it created a 40% discount on a business whose fundamentals accelerated through the attack.
Bear case (narrative). SoFi is a pro-cyclical consumer lender wearing a fintech multiple, and three things can permanently impair it: (1) Credit. Charge-offs on transferred loans are up 78% YoY; the book is marked +$2.0B above par; when the consumer cracks, the marks reverse and "EBITDA" evaporates — and it lends to the lower-income, prime-but-fragile borrower most exposed to a downturn. (2) The growth engine is decaying where it's supposed to be durable — Galileo, the high-multiple "infrastructure" story, shrank 27% on a single client loss and must triple to hit guidance. (3) Trust. A promotional CEO + a credible short report alleging EBITDA inflated 90% and dilution-for-bonuses + a securities probe = a multiple that should compress toward Ally's ~17x, i.e. ~35% downside, before any fundamental miss.
Pre-mortem (18 months out, thesis broke — what happened?): US consumer credit deteriorated in 2H26–27; SoFi's fair-value model lagged the turn; a quarter arrived where mark reversals turned noninterest income negative, GAAP EPS missed badly, and — fairly or not — it was read as partial validation of Muddy Waters. The multiple collapsed to a lender's multiple. Galileo never recovered. Dilution kept the share count climbing. The stock saw single digits.
Are multiples too high? For a bank, yes (26x fwd vs Ally 17x). For a 30% grower, no. The entire debate is a single bet on which one SoFi is — and credit will be the referee.
Contrarian view: The consensus has bifurcated into "fraud" vs "screaming buy," and both are overconfident. The truth is more boring and more useful: SoFi is a good, real, fast-growing bank with aggressive-but-legal accounting whose stock is now cheap enough that the asymmetry has flipped to the upside — conditional on the consumer credit cycle not breaking in the next 12 months. It's a macro-credit bet with a fintech kicker, not a fraud forensic.
Dismantling the bull case.
A levered, sub-NAV Ethereum proxy run by the best operators in the trade — own the discount-closing buyback, not the "treasury premium" that already died; the bet is ETH plus a 0.83x→1.0x mNAV mean-reversion, fighting a 95% drawdown and a structurally bigger rival.
A real, cash-generative super-app now valued like a crypto-beta call option — the bull thesis (tokenization + a 27M-customer flywheel) is genuine, but ~37x forward earnings re-prices violently every time the crypto/options trading cycle exhales, and the prediction-markets engine that drove Q1 growth is one adverse court ruling from being switched off.
A bitcoin miner whose P&L now swings on the BTC mark, not operations — repricing itself as a 1.7 GW AI/HPC power landlord on the back of one $311M AMD lease and a Starboard-pushed pivot; the ~$10.6B cap is paying for an optionality that is still 95% promise.