Wells Fargo & WFC

Reference table

1. What the business is, and how the disclosure actually works

Wells Fargo & Company was founded in 1852, listed on the NYSE since 1972-06-01, is headquartered at 333 Market Street, San Francisco, and employs approximately 197,466 people. Chief executive Charles W. Scharf. CIK 0000072971, CUSIP 949746101, ISIN US9497461015.

The 10-Q filed 2026-07-28 opens its Financial Review with the description we use as the reference point:

> "Wells Fargo & Company is a leading financial services company that has approximately $2.3 trillion in assets. We provide a diversified set of banking, investment and mortgage products and services, as well as consumer and commercial finance, through our four reportable operating segments: Consumer Banking and Lending, Commercial Banking, Corporate and Investment Banking, and Wealth and Investment Management. Wells Fargo ranked No. 38 on Fortune's 2026 rankings of America's largest corporations. We ranked fourth in assets and fifth in the market value of our common stock among all U.S. banks at June 30, 2026."

The four reportable segments — and a vendor table that is half current and half obsolete

This is defect class 7 and Wells Fargo is exactly the case it exists for: the company HAS reorganised its segments, and the vendor file contains both taxonomies stacked on top of each other as if they were comparable.

The seg_prod block supplies nine fiscal years. The FY2025 row uses the CURRENT four-segment structure and it is correct — the four names match the 10-Q verbatim. The FY2024 row and every year back to FY2020 uses "Community Banking" and "Wholesale Banking", a structure Wells Fargo abandoned. And the FY2019 row is not a segment table at all: it contains two lines, "Credit Card Rewards and Rebates" $1.5B and "Trailing Commission" $1.2B, which are fragments of a revenue-disaggregation footnote.

We verified the FY2025 tie and it is exact.

Those six lines sum to $83,699 million — which is EXACTLY our independently rebuilt 2025 total revenue (net interest income of $47.484B plus noninterest income of $36.215B). Not approximately: to the dollar. That is a clean reconciliation and it is also, incidentally, the decisive proof that the vendor's own income-statement "revenue" field of $123,529 million is on a different and wrong basis — see Section 2.

The FY2024 row does NOT tie. Its four lines sum to $83,759M against a rebuilt 2024 total revenue of $82,296M — a $1,463 million (1.8%) overstatement, because the FY2024 vintage carries no Corporate line and no Reconciling Items line while the FY2025 vintage does. Only FY2025 is used anywhere in this dive. No year-on-year segment comparison is drawn, because the vendor file does not permit one.

What the segments do, from the 10-Q's own descriptions:

• Consumer Banking and Lending — "offers diversified financial products and services for consumers and small businesses… checking and savings accounts, credit and debit cards, as well as home, auto, personal, and small business lending. We also provide personalized wealth management and financial planning services through our branch channel."
• Commercial Banking — "provides financial solutions to private, family owned and certain public companies. Products and services include banking and credit products across multiple industry sectors and municipalities, secured lending and lease products, and treasury management."
• Corporate and Investment Banking — "delivers a suite of capital markets, banking, and financial products and services to corporate, commercial real estate, government and institutional clients globally… corporate banking, investment banking, treasury management, commercial real estate lending and capital markets, equity and fixed income solutions as well as sales, trading, and research capabilities."
• Wealth and Investment Management — "provides personalized wealth management, brokerage, financial planning, lending, trust and fiduciary products and services to affluent, high-net worth and ultra-high-net worth clients."

One structural note the filings make explicit and the vendor cannot show: in the third quarter of 2025 certain business customers were transferred from Commercial Banking to Consumer Banking and Lending. The 10-Q cites this transfer four separate times as a driver of segment revenue, loans, deposits and expense in both directions. Segment-level year-on-year comparisons within the last twelve months are therefore contaminated even in the filings themselves, and we do not draw them.

Geography, customer concentration and competitors

seg_geo is EMPTY. There is no geographic revenue disaggregation available from any source in this file, and we do not estimate one. The only geographic disclosure recoverable from the 10-Q is on the lending side: non-US commercial and industrial loans of $81.9 billion at 2026-06-30 (against $81.0B at year-end), of which $40.3B is in the "financials except banks" industry, $13.2B securities-based, $8.3B banks and $2.0B oil, gas and pipelines. Wells Fargo is overwhelmingly a domestic institution and the file supports no more precision than that.

Customer concentration is not a meaningful concept here and no 10% customer disclosure exists — that is the structural advantage of a business funded by tens of millions of retail deposit accounts. The concentration risk in a bank is not customer concentration, it is asset-class concentration, and Section 2 handles it.

Competition, from the 10-K: "Our subsidiaries compete with financial services providers such as banks, savings and loan associations, credit unions, finance companies, mortgage banking companies, insurance companies, investment banks, investment advisory firms, and mutual fund companies. They also face increased competition from nonbank institutions such as investment managers, brokerage houses, private equity and private credit firms, and financial technology companies… Many of these competitors enjoy fewer regulatory constraints and some may have lower cost structures."

And the 10-K names the specific structural threat, unprompted: "digital assets and alternative payment methods, such as cryptocurrencies, stablecoins, and tokens, as well as distributed ledger-based payment, clearing, and settlement processes have the potential to reduce reliance on traditional depository institutions and other financial intermediaries, could lead to a reduction in deposits at banks, and could lead to changes in how financial services are accessed, offered, and delivered." That is the company itself naming deposit disintermediation as a live risk — and it is the same mechanism that two of the knowledge-base claims describe from the other direction (Section 7).

The vendor peer set is unusually sensible for once: JPMorgan ($958.0B), Bank of America ($446.4B), Royal Bank of Canada ($289.8B), Citigroup ($234.7B), Toronto-Dominion ($201.7B), UBS ($175.2B), plus BMO, Bank of Nova Scotia, CIBC, Barclays, ING and Nu Holdings. Citigroup is in this same batch at $136.82, +2.43% today. No peer earnings, book values or multiples are supplied in the file, so no peer-multiple comparison is drawn — only the observation that Wells Fargo is the fourth-largest US bank by assets and fifth by market value of common stock, per its own 10-Q.

2. The single most important thing to understand about these numbers: this is a bank, and half the toolkit is invalid

Everything in this section is a decision about which lenses to reject, and it matters more than any number in the dive.

What we reject, and why

The vendor payload supplies a complete enterprise-value framework. All of it is discarded.

The reason, stated once and applied throughout. Enterprise value asks: what would it cost to buy the whole business free of its capital structure? The arithmetic is market capitalisation plus debt minus cash, on the logic that an acquirer inherits the debt and gets the cash. For a deposit-funded bank that question has no answer, because the debt IS the business. Wells Fargo's $425.7 billion of "total debt" at 2025-12-31 — $251.0B short-term and $174.7B long-term — plus roughly $1.47 trillion sitting in the vendor's otherCurrentLiabilities line (overwhelmingly customer deposits) is not a financing decision made by a treasurer. It is the raw material the bank buys in order to have something to lend. Netting it against the bank's cash produces a number that describes nothing. A bank that repaid all its "debt" would have no deposits and no earning assets and would earn nothing.

The same logic kills interest coverage. The vendor's interestCoverageRatioTTM of 0.650x looks like a company that cannot service its debt. It is nothing of the kind: interest expense is Wells Fargo's cost of goods sold. In 2025 it paid $39.830 billion of interest to earn $87.314 billion of interest — a gross spread of $47.484 billion. Dividing operating income by cost of goods sold is not a solvency test in any industry.

And the same logic kills the cash-flow multiples. For a bank, "operating cash flow" is dominated by changes in trading assets, loans held for sale and deposit balances — flows that reflect balance-sheet management rather than earnings quality. The evidence in this very file is that the vendor cannot decide what the number is: FY2025 operating cash flow is reported as NEGATIVE $19,001 million, while ratios_ttm.priceToOperatingCashFlowRatioTTM of 14.061x implies a trailing figure of POSITIVE $19,010 million. Same magnitude to within 0.05%, opposite sign. We cannot resolve it from the archive and it does not matter, because both figures are being used to answer a question that does not apply.

What we use instead

And a second, equally consequential rejection: the vendor's revenue definition

This is the most damaging single error in the file and it propagates everywhere.

inc_a[0].revenue for fiscal 2025 reads $123,529 million. Wells Fargo's reported total revenue for 2025 was $83,699 million. The vendor is summing total interest income ($87.314B) plus noninterest income ($36.215B) and calling the result revenue. A bank reports net interest income plus noninterest income, and the 10-Q says so explicitly when it defines the efficiency ratio as "noninterest expense divided by total revenue (net interest income and noninterest income)."

The gap is 47.6%, and the vendor contradicts itself in two places within its own file:

• seg_prod FY2025 sums to $83,699M — the correct basis, to the dollar.
• est.revenueAvg for FY2025 is $84,028M — the correct basis, within 0.4%.
• earn_cal.revenueActual for the June 2026 quarter is $22,622M — the correct basis, and it matches our rebuild exactly.
• inc_a.revenue and inc_q.revenue are on the gross basis.

So three blocks of one payload use one definition and two blocks use another, differing by nearly half. Any price-to-sales, margin or revenue-growth figure computed naively off the income statement is wrong. ratios_ttm.priceToSalesRatioTTM of 2.070x and revenuePerShareTTM of $42.34 are both on the gross basis and are rejected; on the correct basis price-to-sales is 3.08x and revenue per share is $28.79.

We rebuilt the entire revenue, expense and provision series from first principles and the rebuild reconciles exactly. For every one of the six annual periods and eight of the nine quarterly periods:

> total revenue (vendor revenue less interest expense) − noninterest expense (vendor operatingExpenses) − provision for credit losses (vendor costOfRevenue less interest expense) = reported pre-tax income, to the dollar.

The single exception is the March 2026 quarter, where a $130 million totalOtherIncomeExpensesNet line sits between operating income and pre-tax income. The derived provision series independently reproduces Wells Fargo's known credit history — a $14.129 billion build in 2020 and a $4.155 billion RELEASE in 2021 — which is the strongest possible confirmation that the decomposition is right.

3. Growth and margin trajectory — rebuilt on the correct basis

Annual, on the bank basis (net interest income plus noninterest income), rebuilt from the vendor income statements and verified to reported pre-tax income:

Read that table and three things are true at once, and they are all important.

First, revenue is not growing. From $74.264B in 2020 to $83.699B in 2025 is a 2.4% compound annual rate across five years. That is the mathematical signature of a bank operating under a hard asset-growth cap, and it is the reason this business has scored a 5 on Growth Quality and not higher. The trailing twelve months at $87.049B is the first genuine acceleration, and it post-dates the cap's removal.

Second, the efficiency ratio is the entire operating story. From 77.60% in 2020 to 65.52% in 2025 to 64.27% trailing is roughly 1,330 basis points of improvement — and, because revenue barely moved across that span, essentially all of the earnings growth from $0.41 to $6.93 of diluted EPS came from cost reduction and provision normalisation, not from selling more.

Third, the provision line is the volatility. Look at 2020 and 2021 side by side: a $14.129 billion build followed by a $4.155 billion release, an $18.3 billion pre-tax swing in twelve months on a revenue base that moved $4.9 billion. This is the defect-class-3 exposure in a bank and it is structural, not occasional. It is precisely why the JPMorgan Visa Class B-2 tender fooled a screen, and it is why Section 3's final subsection strips the June quarter down to a clean number.

Quarterly, rebuilt on the same basis and cross-checked against earn_cal:

The Q2 2025 row is shown grossed up by $253 million on both the noninterest-income and noninterest-expense lines. This is not a smoothing adjustment — it is a correction we derived and then verified against the filing. The 10-Q's own year-on-year bridge states that June-quarter net income rose on "a $1.2 billion increase in noninterest income and a $609 million increase in net interest income, partially offset by a $486 million increase in income tax expense and a $282 million increase in noninterest expense." On the raw vendor figures, net interest income (+$609M) and tax (+$486M) match the filing to the dollar, but noninterest income comes out at +$1,444M against the filing's $1.2B and noninterest expense at +$535M against the filing's $282M — both overstated by approximately $253 million. Grossing up Q2 2025 by $253M on both sides closes both gaps exactly (+$1,191M and +$282M) and leaves pre-tax income unchanged at $6,438M. And the same $253M closes the gap between earn_cal's Q2 2025 revenue of $20,822M and our raw rebuild of $20,569M. Three independent checks, one number. The Q4 2025 row shows the same $253M in the opposite direction against earn_cal and we have flagged it in Section 8 rather than adjusting it, because the annual FY2025 rebuild ties exactly to seg_prod on the unadjusted basis.

The June 2026 quarter is the best operating quarter in this file and it is not close.

• Total revenue $22.622B, up 8.6% year on year on the adjusted comparison — the highest quarterly revenue in the series.
• Noninterest income $10.305B, up 13.1%, and now 45.6% of total revenue — the highest fee share in the file. The 10-Q attributes it to higher deposit-related fees "due to account repricing", higher investment advisory fees "reflecting higher market valuations", higher brokerage transaction activity, higher debt and equity underwriting fees, and higher net gains from equity securities.
• Efficiency ratio 60.39%, against 64.25% a year earlier — 386 basis points. On the unadjusted comparison it is 342 basis points. Either way this is the first sub-62% quarter in the file.
• Diluted EPS $2.00 against $1.60, up 25.0%, confirmed verbatim by the 10-Q: "In second quarter 2026, we generated $6.4 billion of net income and diluted earnings per common share (EPS) of $2.00, compared with $5.5 billion of net income and diluted EPS of $1.60 in the same period a year ago."

And the margin headwind, in the company's own words:

> "Net interest margin decreased in both the second quarter and first half of 2026, compared with the same periods a year ago, driven by growth in lower-yielding assets in our Markets business as well as growth in interest-bearing deposits."

That sentence is the price of the growth. Net interest income rose $609 million, but the assets that produced it earn less than the assets already on the book, and the deposits funding them cost more. The 10-Q quantifies the funding side directly: "Our average deposit cost in second quarter 2026 increased to 1.51%, compared with 1.44% in fourth quarter 2025." Seven basis points in two quarters, and rising.

The exact level of net interest margin is NOT recoverable from this archive and we will not estimate it. Table 2 of the 10-Q — "Average Balances, Yields and Rates Paid (Taxable-Equivalent Basis)" — is a numeric table, and numeric tables are stripped from the current filing extraction. What can be said with precision: net interest income of $12.317 billion in the quarter, against total assets of approximately $2.3 trillion, is a net interest income yield on TOTAL assets of about 2.16% annualised. The reported net interest margin, struck on average earning assets on a taxable-equivalent basis (the 10-Q discloses taxable-equivalent adjustments of $75 million for the quarter), would be meaningfully higher. We state the direction from the filing — it fell — and decline to state the level.

Stripping the one-off items: what the June quarter is really worth

Bank quarters are routinely flattered or depressed by reserve movements, securities repositioning, one-time gains and discrete tax items. We went looking for all of them in the 10-Q. Here is the complete filing-sourced inventory.

Items that flatter the 2026 comparison by depressing the 2025 base:

• Q1 2025 investment securities repositioning loss. The 10-Q states first-half net gains from debt securities "increased driven by the impact of a repositioning of our investment securities portfolio in first quarter 2025." The size is in a stripped table.
• Q1 2025 gain of $263 million on the sale of the non-agency portion of the commercial mortgage third-party servicing business — this one reduces the flattery, being a 2025 gain.
• Q2 2025 gain of $253 million associated with the merchant services joint venture acquisition — again a 2025 gain.
• H1 2025 favourable mortgage servicing rights valuation adjustments, "including for higher expected escrow balances." Size not disclosed in prose.

Items that depress the 2026 quarter:

• The rail car leasing business was sold in Q1 2026. Lease income fell $148 million in the June quarter and $297 million in the half; lease expense fell $93 million and $180 million. Net pre-tax drag of $55 million in the quarter and $117 million in the half. This is a permanent reduction in the revenue base, not a one-off, and we do not add it back — but it means the underlying revenue growth is slightly better than reported.

Items that flatter the 2026 quarter and that we CANNOT size:

• Net gains from equity securities "increased reflecting improved results from our venture capital investments, including higher realized and UNREALIZED gains, partially offset by higher impairment losses." This is the defect-class-3 candidate and it is precisely the pattern that produced the fake 94% net margin at another name in this programme: unrealised marks on a private portfolio flowing through the income statement as earnings. The magnitude sits in Table 3, which is stripped. We state plainly that we cannot size it, and we treat the fee-income strength as partially unverified rather than pretending otherwise.

Items that turn out NOT to be a distortion — and this is the useful finding:

• There was no reserve release at the total-company level. June-quarter provision was $914 million against net charge-offs of $876 million (commercial $156M plus consumer $720M) — a $38 million BUILD. The year-ago quarter was the same shape: $1,005 million of provision against $997 million of charge-offs, an $8 million build. The allowance for credit losses on loans actually ROSE $70 million over the first half. So while the 10-Q does attribute part of the improvement to "a lower allowance for commercial real estate loans", that release was more than offset by builds on commercial-and-industrial and auto loans driven by higher balances. The quarter was not manufactured by a reserve release, and we checked specifically because bank quarters usually are.
• There is no FDIC special assessment in either period. The term does not appear in either 10-Q.

The tax line, which IS a real distortion. The effective rate was 17.42% in the June quarter against 14.23% a year earlier, and the 10-Q says the increase was "driven by higher pre-tax income and lower discrete tax benefits related to the resolution of prior period tax matters." The trailing-twelve-month rate is 16.51% (confirmed exactly by ratios_ttm.effectiveTaxRateTTM of 16.506%) and the 2025 full-year rate was 15.24% — both well below the 21% federal statutory rate before state taxes.

Clean operating EPS for the June 2026 quarter. Normalising the tax rate to 21% on the reported pre-tax income of $8,047 million gives net income of $6,357 million, less approximately $247 million of preferred dividends, over 3,080.0 million diluted shares:

> Tax-normalised diluted EPS ≈ $1.98, against a reported $2.00 — a 1.0% difference.

That is the honest answer and it is a boring one: the June quarter was not materially flattered by tax, and it was not flattered by a reserve release at all. The only unquantified flattery is the venture-capital gains, which we cannot size and therefore do not net out. Annualising is where the caution belongs: $2.00 × 4 = $8.00 would put the run-rate above the FY2027 consensus of $7.972 already, which is exactly the kind of arithmetic that leads a screen astray. The first half is the better base — $3.60 doubled is $7.20, essentially the FY2026 consensus of $7.264. We use the half-year, not the quarter.

4. Balance sheet — and why the enterprise-value rebuild is refused, not performed

The exemplar structure calls for an enterprise-value rebuild in this section. For Wells Fargo we are deliberately not performing one, and the refusal is the analysis. Section 2 set out why. What follows instead is the balance sheet a bank should be judged on: capital, liquidity, credit and tangible book.

The vendor balance sheet is TWO QUARTERS STALE — and the filing updates it

bal_a[0] is dated 2025-12-31, filed 2026-02-24 with the 10-K. There is no balance sheet in this payload for either 2026 quarter. The 10-Q filed 2026-07-28 states:

> "At June 30, 2026, our assets totaled $2.3 trillion, up $133.6 billion from December 31, 2025."

The vendor reports total assets of $2,148.631 billion; the filing shows approximately $2,282 billion; we use the filing. That is a 6.2% understatement — and, more importantly, it is the single most economically significant fact in the file, because a $133.6 billion balance-sheet expansion in six months by a bank that was legally forbidden to grow for seven years is the whole medium-term thesis in one number.

Three further post-balance-sheet events the vendor cannot see:

1. The 8-K of 2026-05-20 records the issuance of $6.0 billion of Medium-Term Notes, Series Y: $2.25 billion of fixed-to-floating senior redeemable notes due 2029, $500 million of floating-rate notes due 2029, and $3.25 billion of fixed-to-floating notes due 2032. Every balance sheet in this file predates it. (For a bank this is routine term funding rather than a leverage event, and we treat it as such — but it is exactly the class of omission that made another name's debt 23% understated, and it is reported.)

2. 84 million shares were repurchased for $7.1 billion during the first half, against $875 million of stock issued.

3. $3.3 billion of common and preferred dividends were paid in the first half.

Capital — the constraint that actually binds

From the 10-Q filed 2026-07-28, all filing-verified:

The 8.50% requirement decomposes, per Note 21, into a 4.50% minimum, a 2.50% stress capital buffer, a 1.50% G-SIB surcharge and a 0.00% countercyclical buffer. The stress capital buffer sits at 2.50%, which is the regulatory FLOOR — a bank cannot be assigned less. The 10-Q also notes that federal banking regulators issued a proposed rule in March 2026 to implement the final Basel III components, replacing both current approaches with an "Expanded Risk-Based Approach", and separately that the Company "does not expect a significant change to its current G-SIB capital surcharge based on an assessment of the proposed rule."

Sizing the excess capital. Common equity Tier 1 capital is not disclosed in the extracted prose (Table 29 is stripped). Working backwards from the 10.26% ratio and an estimated CET1 capital base of roughly $135 billion (common stockholders' equity of $164.5 billion at 2025-12-31 less goodwill of $25.0 billion and other regulatory deductions), risk-weighted assets are on the order of $1.32 trillion, and 176 basis points of excess CET1 is therefore roughly $23 billion. This is an estimate we have labelled as such, derived from a disclosed ratio and an inferred denominator — it is not a filing figure. What it is good for is one comparison: $23 billion of excess capital against a capital-return run-rate of roughly $20 billion a year. That is the tension in Section 6.

Liquidity

> "As of June 30, 2026, we had approximately $713.4 billion of borrowing capacity at the Federal Reserve Discount Window and Federal Home Loan Banks (FHLB). This borrowing capacity included $277.4 billion related to pledged high-quality liquid debt securities within our primary sources of liquidity and $436.0 billion related to pledged loans and other debt securities within our contingent sources of liquidity."

And the securities portfolio quality: "At June 30, 2026, 99% of the combined AFS and HTM debt securities portfolio was rated AA- or above."

With one adverse note, stated plainly: "The total net unrealized losses on AFS and HTM debt securities increased from December 31, 2025, primarily due to changes in interest rates." That is a direct headwind to tangible book value through accumulated other comprehensive income, which stood at −$6.673 billion at 2025-12-31 (improved from −$12.176 billion a year earlier). The June 2026 level is not disclosed in the extracted prose and we do not estimate it.

Tangible common equity — and the vendor error that matters most for a bank

ratios_ttm.tangibleBookValuePerShareTTM reads $51.59. It is wrong, and for a bank this is the most consequential error in the payload, because price-to-tangible-book is the primary valuation lens.

The vendor's figure equals km_ttm.tangibleAssetValueTTM of $157.332 billion divided by roughly 3,049.5 million shares. That numerator is total equity less goodwill — and total equity includes $16.608 billion of preferred stock and $1.921 billion of noncontrolling interests. Tangible COMMON equity must exclude both. The 10-Q's own definition is unambiguous: "Book value per common share is common stockholders' equity divided by common shares outstanding. Tangible book value per common share is tangible common equity divided by common shares outstanding."

Rebuilt from the 2025-12-31 balance sheet:

The vendor reports tangible book value per share of $51.59, implying 1.71x; the correct figure, excluding the preferred stock and noncontrolling interests the 10-Q's own definition excludes, is between $44.05 and $46.14, implying 1.92x to 2.01x. We use $46.14 and 1.92x as the base and disclose the range.

Two honest caveats on that number. First, Wells Fargo's own reconciliation (Table 33 of the 10-Q) deducts goodwill and certain identifiable intangibles and goodwill-related deferred taxes, and the precise deduction set is not recoverable from the extracted text — which is why we present both a goodwill-only and a goodwill-plus-intangibles version rather than claiming a single figure. Second, and more importantly, this is the DECEMBER 2025 tangible book, not the June 2026 one. Since then: retained earnings have risen by roughly $11.7 billion of first-half net income; $7.1 billion of buybacks at an average price of about $84.52 — well above tangible book — have reduced tangible book per share; $3.3 billion of dividends have been paid; and accumulated other comprehensive income has moved by an undisclosed amount as unrealized securities losses widened. The net effect is plausibly a small increase in tangible book per share, but we cannot compute it, and Table 33 is stripped. Every price-to-tangible-book figure in this dive is struck on the 2025-12-31 balance sheet and labelled accordingly.

Return on tangible common equity

A bank earning roughly 16% on tangible common equity trailing and heading toward 18% is earning well above any plausible cost of equity, and that — not the earnings multiple — is the analytical justification for paying nearly 2x tangible book.

Credit quality — the line that decides everything

All figures from the 10-Q filed 2026-07-28:

Every one of those lines is moving the right way, and the filing is explicit about the driver: "The allowance coverage for total loans was 1.40% at June 30, 2026, compared with 1.45% at December 31, 2025, reflecting a decrease in the allowance for our commercial real estate portfolio driven by improved credit performance." And: "Nonperforming assets (NPAs) of $7.9 billion at June 30, 2026, decreased $559 million from December 31, 2025, driven by lower commercial real estate nonaccrual loans."

And this is the answer to the office question, which is the reason most people avoided this stock in 2023 and 2024. The 10-Q states:

> "The portfolio is diversified both geographically and by property type. At June 30, 2026, the five states with the largest geographic concentrations of CRE loans, as shown in Table 13, represented a combined 51% of the total CRE portfolio. The largest property type concentrations were apartments at 27% and industrial/warehouse at 22% of the portfolio at June 30, 2026."

> "CRE mortgage loans classified as criticized were $11.8 billion and $13.4 billion at June 30, 2026, and December 31, 2025, respectively. CRE construction loans classified as criticized were $1.2 billion and $1.7 billion… The decrease in criticized CRE loans was primarily driven by the office, apartments, and industrial/warehouse property types."

Office is no longer among the two largest property-type concentrations — apartments at 27% and industrial/warehouse at 22% are — and office is now named as one of the property types DRIVING the decline in criticized loans. (The March 2026 10-Q gave apartments 28% and industrial/warehouse 21% on the same measure, so the mix is stable.) The exact office balance and the office-specific reserve ratio sit in Table 13, which is stripped from the extraction. We can establish the direction and the ranking from the filing; we cannot establish the office dollar exposure, and we do not estimate it. The one number we can put on the whole portfolio: the total allowance of $14.4 billion is 1.40% of approximately $1.029 trillion of loans, and it FELL because of commercial real estate.

Unfunded CRE commitments were $6.2 billion for mortgage loans at both dates and $11.8 billion for construction loans against $9.2 billion at year-end — construction commitments up $2.6 billion, the one CRE line moving the other way.

5. Cash flow and capital returns

The cash-flow statement for a bank is not a cash-generation statement, and we treat it accordingly.

Operating cash flow swings from +$40.4 billion to −$19.0 billion across three years on an earnings stream that never fell below $19 billion. That is what happens when trading assets, loans held for sale and deposit balances dominate the operating section. capitalExpenditure is $0 in all four years, which is also wrong — banks capitalise premises, equipment and internally developed software, and the 10-Q names "amortization of internally developed software" as a rising expense line — but the outlay is folded into "other investing" and cannot be separated. Free cash flow, free-cash-flow yield and every price-to-cash-flow multiple are rejected. Nothing in this dive depends on them.

What DOES matter, and it is the loudest number in the file, is the capital return — and it comes from the 10-Q, not from the vendor's lastDividend field:

> "During the first half of 2026, we issued $875 million of common stock, substantially all of which was issued in connection with employee compensation and benefits, and we repurchased 84 million shares of common stock at a cost of $7.1 billion. We paid $3.3 billion of common and preferred stock dividends during the first half of 2026."

> "On July 28, 2026, the Board approved an increase to the Company's third quarter 2026 common stock dividend to $0.50 per share."

> "On April 29, 2025, we announced that the Board authorized the repurchase of up to $40 billion of common stock. Unless modified or revoked by the Board, this authorization does not expire. At June 30, 2026, we had remaining Board authority to repurchase up to approximately $22.7 billion of common stock."

The arithmetic:

Net share shrinkage. 84 million repurchased less roughly 10 million issued ($875M at ~$85) is about 74 million net shares in six months, roughly 2.4%, or 4.8% annualised. The vendor's diluted-share series is corrupted for three of the last four quarters (Section 8) and cannot corroborate this, but the June quarter's diluted count of 3,080.0 million against a March 2025 count of 3,321.6 million is a 7.3% reduction in five quarters, which is consistent.

Sustainability — the honest test. Trailing net income to common is $22.152 billion and the forward return is roughly $20.2 billion: a 91% payout of earnings. That is not funded from operations alone over time; it is funded from operations plus the roughly $23 billion of excess CET1 estimated above. And crucially, it competes directly with balance-sheet growth, because every dollar of new risk-weighted assets consumes capital that would otherwise fund a buyback. You cannot grow total assets $133.6 billion in six months AND return 7.6% of market capitalisation indefinitely. One of the two moderates. That is the single most important thing to watch in this name and it is where 6a's most fragile assumption sits.

Note the vendor's dividend trap, which is the same one seen at other names in this programme. profile.lastDividend reads 1.8 and ratios_ttm.dividendPerShareTTM reads 1.80. That $1.80 is the trailing-twelve-month total, not the declared rate. The filing gives the forward rate: $0.50 a quarter, $2.00 annualised, a 2.26% yield — not the vendor's implied 2.04%. The filing wins.

6. Valuation — priced in or room?

At $88.39 (market cap $267.291B, implied 3,024.0 million shares):

A GAAP/non-GAAP note that is unusually simple here. Wells Fargo does not report a non-GAAP earnings measure. inc_a/inc_q are GAAP, the 10-Q's $2.00 diluted EPS is GAAP, and consensus epsAvg is struck on the same GAAP basis — which we verified: est.epsAvg for FY2025 is $6.31563 against a reported GAAP diluted EPS of $6.32, a 0.06% difference. Similar for FY2024 ($5.333 vs $5.37) and FY2023 ($4.848 vs $4.83). There is no GAAP-to-non-GAAP gap to disclose, no basis argument, and no reconciliation needed. That is rarer than it sounds and it is worth recording. The company does designate three items as non-GAAP — tangible common equity, tangible book value per common share, return on average tangible common equity, and net interest income excluding Markets — and we label them as such wherever used.

Per the data contract, est.ebitdaAvg and est.ebitAvg are not used for forward valuation. We tested them anyway and report the result: all eight years pass every measured defect signature — ebitAvg is below ebitdaAvg in every year, there are no negative EBITDA rows, and there is no 50%/33%-of-revenue fabrication pattern. The rows are nonetheless rejected, on the different ground that EBITDA is not a meaningful measure for an institution whose largest expense is interest.

The FY2029 estimate row is discarded entirely. Its revenueAvg of $129.241 billion is a 32.7% jump from FY2028's $97.398 billion while EPS rises only 8.5% — because that single row appears to have been struck on the gross interest income basis (it sits close to the vendor's gross-basis 2025 revenue of $123.5 billion) rather than the bank basis used for FY2026 to FY2028. It rests on ONE revenue analyst and three EPS analysts. No conclusion in this dive touches it.

Estimate coverage, stated plainly: FY2026 rests on 15 EPS analysts and 9 revenue analysts; FY2027 on 17 and 9 — good, and comparable to the best coverage in this programme. FY2028 EPS rests on 9 analysts but FY2028 REVENUE on only TWO, so the FY2028 revenue line is used nowhere. FY2029 is excluded outright. All fair-value anchors are built on FY2027, where coverage is deepest.

Peer context. Vendor peers include four genuine US comparables — JPMorgan ($958.0B market cap, $357.52), Bank of America ($446.4B, $62.90), Citigroup ($234.7B, $136.82) — plus six Canadian and European banks and one Latin American digital bank (Nu Holdings) that is not a comparable at all. No peer earnings, book values or multiples are supplied in the file, so NO peer-multiple comparison is drawn. What can be said is filing-sourced and useful: the 10-Q states Wells Fargo ranked fourth in assets and fifth in the market value of its common stock among all US banks at June 30, 2026 — i.e. it is the large US bank that carries the lowest market value relative to its asset rank. Citigroup, in this same batch at $136.82 and +2.43% today, is the closest analogue in the "large US bank that trades below its franchise" category, and a reader holding both should understand that they are two expressions of the same re-rating trade.

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

At $88.39 — 12.75x trailing, 11.09x FY2027 consensus, roughly 1.92x tangible common equity — the price embeds the following falsifiable claims, each with a number and a date.

• EPS reaches $7.264 in calendar 2026 and $7.972 in 2027. (Consensus-derived; 15 and 17 analysts.) The first half delivered $3.60, which doubled is $7.20 — so 2026 requires the second half to run about 1% ahead of the first, and 2027 requires +9.7% on top. The nearest check is the 2026-10-13 print, where consensus is $1.84 — BELOW the $2.00 just delivered. (That is not a typo and it is the most interesting single number in the estimate set: the street is modelling a sequential DECLINE from the June quarter.)
• The efficiency ratio does not give back the June quarter. (Our number, derived from the filings.) It printed 60.39% against 64.25% a year earlier and 64.16% on a trailing basis. The price assumes the 386 basis points are structural. Falsifiable at the October print: any efficiency ratio above 64% says the June quarter was a fee-income event, not an expense event. This is where we think the market is being too conservative — see 6c.
• Credit stays approximately where it is. (Filing-confirmed levels; our assumption on the direction.) Allowance coverage 1.40% of loans, criticized CRE mortgage $11.8B, NPAs 0.77% of loans. A move back to 1.55% coverage on the ~$1.029 trillion loan book is roughly $1.5 billion of incremental pre-tax provision — about $0.39 of EPS, or 4.9% of the FY2027 consensus, in a single quarter.
• The capital return continues at roughly $20 billion a year. (Filing-sourced run-rate: $7.1B of buyback in the first half plus a $2.00 forward dividend.) Against $22.7 billion of remaining authority and roughly $23 billion of estimated excess CET1. This is the most fragile assumption in the price, because it is in direct competition with the risk-weighted-asset growth the removal of the asset cap now permits — total assets rose $133.6 billion in six months. You can have the buyback or you can have the growth; the price currently assumes both.
• The market keeps paying about 11x forward and roughly 1.9x tangible book for a bank earning ~16% on tangible common equity — i.e. NO re-rating for the removal of a seven-year regulatory constraint. (Our number.) At 10x FY2027E the stock is $79.72; at 13x it is $103.64. The multiple, not the earnings, is where the variance in the outcome lives, and this bullet is where we disagree with the crowd.

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

Expected return decomposes as: EPS growth + multiple drift + shareholder yield. For Wells Fargo the decomposition is unusually lopsided and we will not disguise it.

Our base fair value of $99 is 12.4x the FY2027 consensus of $7.972, applied to today's estimate. Today the market pays 11.09x the same estimate. So the entire +12.0% base-case gap is multiple re-rating — 11.09x to 12.4x — with zero contribution from estimate revision. Earnings growth accrues separately, as the market's forward year rolls from FY2027 to FY2028, and that rolldown is mechanical and is explicitly NOT counted as re-rating.

Stated as arithmetic over the next twelve months:

• If the multiple simply HOLDS at 11.09x and the market's forward year rolls to FY2028E of $9.158, the price is $101.6 (+14.9%), plus 2.26% of dividend received. That is the mechanical case and it requires no opinion from us at all.
• If the multiple holds at 11.09x on the SAME FY2027 estimate, the price is unchanged at $88.39. Flat.
• Our base assumes a genuine re-rating to 12.4x, which is +11.8% on the same estimate.

Why we assume expansion, and it is the one judgement in this dive that is not arithmetic. Wells Fargo traded at a persistent discount to JPMorgan and Bank of America for seven years for one specific, identifiable, disclosed reason: it was the only large US bank operating under a Federal Reserve order capping its total assets. That order's growth limitation was removed on 2025-06-03 and the order itself was terminated on 2026-03-05, both filing-verified. A discount applied for a reason should narrow when the reason is removed. The balance sheet has already grown $133.6 billion in the six months since, and the operating evidence has arrived alongside it — a 60.39% efficiency ratio, ~17.7% annualised return on tangible common equity, and credit improving on every disclosed measure.

But we name it as the fragile leg, without qualification. A re-rating is a judgement about what other people will pay, not about what the company will earn. If we are wrong about the multiple and right about everything else, the return is the mechanical rolldown — roughly +15% over a year, plus the dividend — which is still acceptable but is not our thesis. If we are wrong about both, the shareholder yield of ~7.6% is the floor under the outcome, and it is a real floor: at the current pace roughly 5.3% of the shares outstanding are being retired every year, which raises per-share earnings mechanically regardless of what the multiple does.

Contrast with the alternative structure: in the highest-quality names in this batch, most of the return is earnings growth and the multiple is fighting it. Here it is the reverse: the earnings are pedestrian and the multiple is the whole trade. That asymmetry is exactly why this is Buy — Tactical rather than Buy — Core.

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

• We think the regulatory release is worth a multiple and the price does not yet pay one. The market has had five months since the 2026-03-05 consent-order termination and fourteen months since the 2025-06-03 asset-cap removal, and the stock has returned +13.6% over twelve months against SPY's +24.3% — it has underperformed through the entire period in which its defining constraint was lifted. The evidence that the release is economically real is in the filings: total assets +$133.6 billion in six months to $2.3 trillion. The knowledge base's single name-level claim corroborates the mechanism independently: a 2026-04-01 observation that Wells Fargo "deployed $200B+ into repo after its asset cap removal, materially easing funding-market stress and dampening quarter-end volatility." That claim is channel-attributed with no named speaker and we weight it accordingly — but it describes exactly the balance-sheet behaviour the 10-Q's $133.6 billion confirms.
• We think the June quarter's 60.39% efficiency ratio is more repeatable than the street's own estimate implies, and this is our clearest divergence. Consensus for 2026-10-13 is $1.84 of EPS on $22.369 billion of revenue — both below the $2.00 and $22.622 billion just delivered. That is the street explicitly declining to extrapolate. The street may well be right, because the 10-Q attributes part of the fee strength to venture-capital gains "including higher realized and unrealized gains" whose size we cannot recover from the archive. But the expense side is not a gain: noninterest expense of $13.661 billion was the LOWEST in five quarters and fell sequentially from $14.460 billion, while technology, telecommunications and equipment expense ROSE. That is cost reduction, not a mark. Watchable number: the efficiency ratio on 2026-10-13. Below 62% validates us; above 64% validates the street.
• Negative surprise that would break the thesis — a credit re-build, and specifically a commercial real estate re-build. The allowance fell to 1.40% of loans because the CRE allowance fell. That is the line with the shortest fuse. Watchable number: allowance coverage of total loans on 2026-10-13. Any move above 1.45% reverses the direction that produced two consecutive quarters of favourable provision, and each 15 basis points is roughly $1.5 billion of pre-tax provision.
• Negative surprise on funding — deposit cost. The 10-Q discloses average deposit cost rising to 1.51% in the June quarter from 1.44% in the fourth quarter of 2025, and states net interest margin decreased year on year because growth is concentrated in "lower-yielding assets in our Markets business as well as growth in interest-bearing deposits." The growth the market wants and the margin the market wants are in tension, and that tension is not in consensus. Watchable: average deposit cost above 1.60%, or a quarter in which net interest income falls sequentially.
• Negative surprise on capital return — a slowdown in the buyback. At $7.1 billion a half against $22.7 billion of authority and roughly $23 billion of estimated excess CET1, the current pace has about six quarters of runway before it must be funded entirely from earnings while the balance sheet is also growing. Watchable: the third-quarter repurchase disclosure. Anything materially below $3.5 billion is the first sign that the growth is winning.

Synthos fair values

All three anchors are multiples of the FY2027 consensus EPS distribution (mean $7.97181, low $7.45912, high $8.39402; 17 analysts — the deepest coverage in the estimate set), each cross-checked against the December-2025 tangible common equity of $46.14 a share.

• Bear ~$76 — 10.2x the FY2027 consensus LOW of $7.459. Cross-check: 1.65x tangible common equity of $46.14 gives $76.13 — the two methods land 0.1% apart. Second cross-check: 10.5x the FY2026 consensus of $7.264 gives $76.27. The scenario: the commercial real estate improvement stops and reverses, allowance coverage rebuilds toward 1.55%, deposit costs keep climbing past 1.60% while the Markets-driven asset growth keeps diluting margin, the efficiency ratio reverts to the 64-66% band, and the buyback is throttled to fund risk-weighted-asset growth. The multiple stays at today's 11x and applies to a lower number. −14.0%.
• Base ~$99 — 12.4x the FY2027 consensus MEAN of $7.972, which is $98.85. Cross-check one: 2.15x tangible common equity of $46.14 gives $99.20 — 0.4% apart. Cross-check two: 13.6x the FY2026 consensus of $7.264 gives $98.79. Three independent constructions inside a 0.4% band. Sensitivity, stated openly: 11.5x FY2027E gives $91.68 — barely above spot — and 13.5x gives $107.62. The entire answer lives in an 11.5x to 13.5x band, and we claim no more precision than that. The 2.15x tangible-book anchor is itself justified by the standard identity: a bank earning 16% on tangible common equity with a 10% cost of equity and 4% growth supports (16−4)/(10−4) = 2.0x; at 17% it supports 2.17x. We are asking for the low end of what the returns support. The scenario: efficiency holds near 62%, credit stays benign, the capital return continues at a moderated pace, and the market pays a slightly higher multiple because the regulatory constraint is gone. +12.0%, plus 2.26% of dividend.
• Bull ~$118 — 14.05x the FY2027 consensus HIGH of $8.394, which is $117.94. Cross-check: 12.9x the FY2028 consensus of $9.158 gives $118.14; 2.56x tangible common equity. The scenario: the June quarter's 60.39% efficiency ratio proves to be the new floor rather than a peak, the post-asset-cap balance-sheet growth converts into net interest income without further margin dilution, credit stays where it is, the full $22.7 billion of buyback authority is executed near current prices, and Wells Fargo closes the valuation gap to the two US banks ranked above it. +33.5%.

Base is 12.0% above spot; asymmetry is 2.39:1 to the upside (14.0% down, 33.5% up). That is the best risk/reward ratio produced by any anchor set in this batch, and it is the arithmetic reason the verdict is a Buy rather than a Watch. What keeps it Tactical rather than Core is that the entire base-case gap is a multiple assumption, our base sits only 0.9% above the street's mean, and the knowledge base contributes essentially nothing.

7. Knowledge base — thirteen claims, zero named speakers, and one that is actually about this company

Raw hits: 13. Entity matches: 9. Text-only matches: 4. Used: 6. Discarded: 7. Named speakers across the entire lane: ZERO.

The search covered WFC, Wells Fargo, Wells Fargo & Company and Wachovia across all 51,928 distilled claims. Entity matches are claims whose entities list contains the company; text-only matches are claims where the name appears in the thesis or test text but not in the entity list, and every one of those four was inspected individually for collision before use.

The attribution problem, stated first because it governs everything else

Not one of the thirteen claims carries a speaker field. All thirteen are attributed only to a CHANNEL — business_breakdowns, compound_and_friends, darius_dale, we_study_billionaires, arthur_hayes, empire. Under the 4-lane attribution policy this is the weakest form of sourcing we accept, and it is exactly the failure mode that policy exists to prevent: a channel label tells you where a claim was said, not who said it or what their record is. For a name where the channel-level lane is 100% of the evidence, that is a material limitation and it is the primary reason the conviction rating is Low.

Source and document concentration is also severe. Thirteen claims come from only eight distinct documents. compound_and_friends supplies 4 of 13 (30.8%), and three of those four come from a single episode (doc compound_and_friends-UhpU91veGZg, 2025-12-12) — of which two are the SAME claim distilled twice. arthur_hayes supplies 2 (15.4%), darius_dale 2 (both from one document), empire 2 (both from one document), business_breakdowns 2, we_study_billionaires 1. No position reversals are visible in the dates. No claim carries a speaker_role of management, so the half-weight management discount does not apply anywhere in this lane.

What we discarded, and why

Seven of thirteen discarded. Each reason stated.

What we kept, verbatim

The one claim that is actually about Wells Fargo:

> 2026-04-01 · bullish · conviction 70 · horizon: fact · skill 0.8 · entities: Wells Fargo (sole entity) · channel: darius_dale · NO NAMED SPEAKER

> "Wells Fargo deployed $200B+ into repo after its asset cap removal, materially easing funding-market stress and dampening quarter-end volatility."

This is the only claim in thirteen whose subject is Wells Fargo, and it is the only one that carries the company as its SOLE entity. It is a factual claim rather than a valuation view, its horizon is tagged fact, and — critically — it is independently corroborated by the filings. The Federal Reserve removed the asset-growth cap on 2025-06-03; the 10-Q filed 2026-07-28 reports total assets up $133.6 billion in six months to $2.3 trillion, and the 10-Q attributes part of the balance-sheet growth to "an increased volume of resale agreements" — which is the repo activity the claim describes. A channel-attributed claim with no named speaker that is verifiable against a filing is the best case for this class of sourcing, and this one verifies.

The sector claims, kept as supporting colour and nothing more:

> 2025-12-12 · bullish · conviction 60-70 · horizon: thesis · entities: Wells Fargo, The Hartford, Microsoft · channel: compound_and_friends · NO NAMED SPEAKER

> "The next AI trade is stodgy incumbents (banks, insurers, accounting, law, drug discovery) gaining operating leverage and margins as AI penetrates — driving the small/midcap rally."

> (Retained once; the same claim appears twice in the same episode and the duplicate is discarded.)

> 2024-10-14 · bullish · conviction 70 · horizon: thesis · entities: AGL, Barclays, Wells Fargo, Centerbridge, Beach Point, KeyBank · channel: business_breakdowns · NO NAMED SPEAKER

> "Bank/private-credit joint ventures (bank originates, private firm funds off-balance-sheet) are proliferating and will keep growing."

> 2025-10-15 · bullish · conviction 72 · horizon: thesis · entities: JPMorgan, Bank of America, Wells Fargo, Citibank · channel: arthur_hayes · NO NAMED SPEAKER

> "Bessent's push to shift money creation to commercial banks (JPMorgan's $1.5T initiative) means large banks expand credit, adding to money printing and asset inflation."

> 2024-02-17 · bullish · conviction 65 · horizon: thesis · entities: Wells Fargo, Berkshire · channel: we_study_billionaires · NO NAMED SPEAKER

> "Buy misunderstood financials on distributable-cash analysis when banks get hit; Berkshire and Wells Fargo were large winning positions."

And the only bearish claim in the lane, retained at zero weight and reported for honesty:

> 2020-06-04 · BEARISH · conviction 70 · horizon: thesis · entities: Goldman Sachs, Citigroup, Wells Fargo, Western Union · channel: empire · NO NAMED SPEAKER

> "Old archaic financial providers are declining (multi-year lows) as value migrates to digital-native finance."

What this lane is, stated without inflation

One claim in thirteen is about Wells Fargo. It is a factual observation, not a valuation view, it has no named speaker, and it is corroborated by the filings. Everything else uses the company as a member of a set — an AI-operating-leverage illustration alongside The Hartford and Microsoft, a private-credit-joint-venture list of six institutions, a policy claim naming the four largest US banks, a retrospective about a past position, and a research-house index citation.

The bearish side of the lane is a single claim from 2020-06-04 and it has been falsified by six years of events. Struck near the pandemic low, it argued that legacy banks were in structural decline. Wells Fargo's diluted EPS went from $0.41 in 2020 to $6.93 trailing. We report it because it is the only bearish claim in the lane and suppressing it would be dishonest; we weight it at zero because it is six years old and wrong.

The one thing this lane genuinely adds — and it is worth something. The darius_dale claim and the business_breakdowns claim, taken together, describe the two structural forces acting on Wells Fargo's balance sheet in opposite directions: the bank re-entering funding markets at scale now that it is legally permitted to grow, and origination migrating off bank balance sheets into private-credit joint ventures. The 10-K names the same disintermediation risk in its own words — "digital assets and alternative payment methods… could lead to a reduction in deposits at banks." That is a real tension and the knowledge base surfaced it. It is not name-level conviction and it does not justify a conviction premium in the fair value.

Conclusion: conviction rating Low, breadth 6, net conviction positive-low. The knowledge base did not tell us to own this stock and we are not pretending it did. The valuation, the capital return and the filing-verified removal of the regulatory constraint did.

8. Data integrity — what we rejected from the vendor file and why

Twenty findings. Wells Fargo produced the largest defect count of any name in this programme so far, and the reason is structural rather than sloppy: the vendor's data model is built for an industrial company and a bank breaks most of it. Several findings below are not vendor errors at all — they are correct arithmetic applied to a question that does not exist for this issuer.

1. Enterprise value, net debt and every EV multiple — REJECTED AS A CLASS, not rebuilt. enterpriseValueTTM $523.055B, implied net debt $255.764B, bal_a[0].netDebt $251.512B, netDebtToEBITDATTM 7.720x, evToEBITDATTM 15.787x, evToSalesTTM 4.051x, evToOperatingCashFlowTTM and evToFreeCashFlowTTM 27.285x. For a deposit-funded institution, debt is the raw material of the business, not a financing decision, and netting it against cash produces a number with no economic meaning. The exemplar-standard EV rebuild is deliberately NOT performed here; performing it would be the error. Also rejected on the same ground: interestCoverageRatioTTM 0.650x (interest expense is cost of goods sold, not a solvency burden), currentRatioTTM/quickRatioTTM 0.512x, workingCapitalTTM −$895.279B, investedCapitalTTM −$858.685B, grahamNetNetTTM −$380.44, netCurrentAssetValueTTM −$1,160.149B.

2. inc_a/inc_q revenue is on a GROSS INTEREST INCOME basis, not the bank's reported total revenue — REJECTED and the entire series rebuilt. FY2025 vendor revenue $123,529M; Wells Fargo's total revenue (net interest income plus noninterest income) $83,699M. A 47.6% gap. The 10-Q defines the efficiency ratio as "noninterest expense divided by total revenue (net interest income and noninterest income)", settling the definition. The vendor contradicts itself in three places within its own file: seg_prod FY2025 sums to $83,699M (exact match to our rebuild), est.revenueAvg FY2025 is $84,028M (0.4%), and earn_cal.revenueActual for the June 2026 quarter is $22,622M (exact match). We rebuilt total revenue, noninterest expense and the provision for credit losses from first principles for all six annual and nine quarterly periods, and the rebuild reproduces reported pre-tax income to the dollar in fourteen of fifteen periods (the exception being March 2026, where a $130M other-income line intervenes). Consequential rejections: priceToSalesRatioTTM 2.070x → 3.07x; revenuePerShareTTM $42.34 → $28.79; every margin ratio in ratios_ttm computed on the gross base.

3. The balance sheet is TWO QUARTERS STALE, and post-dated by a $6.0B bond issue. bal_a[0] is 2025-12-31, filed with the 10-K on 2026-02-24. There is no 2026 balance sheet in this payload. The 10-Q filed 2026-07-28 states total assets "totaled $2.3 trillion, up $133.6 billion from December 31, 2025" — against the vendor's $2,148.631B, a 6.2% understatement. The vendor reports total assets of $2,148.6B; the filing shows approximately $2,282B; we use the filing. Additionally, the 8-K of 2026-05-20 records $6.0 billion of senior Medium-Term Notes issued after every balance sheet in the file, and the 10-Q records 84 million shares repurchased for $7.1 billion plus $875 million issued — none of it reflected.

4. The 10-K is a WRAPPER and the annual financial statements are ABSENT from this archive. The file is 64,639 characters against 473,441 and 430,417 for the two 10-Qs. It contains ZERO occurrences of "total revenue" and ZERO of "net interest income". Part IV states: "The Company's consolidated financial statements, including the Notes thereto, and the report of the independent registered public accounting firm thereon, are set forth in the 2025 Annual Report to Shareholders, and are incorporated into this item by reference." That Annual Report exhibit was not extracted. The 10-K in this archive contains Item 1 (Business), Regulation and Supervision, and Part IV administrative matter — and no financial statements at all. Consequence: every annual figure in this dive comes from the vendor payload; the filing-verified anchor throughout is the 10-Q filed 2026-07-28, which covers the June 2026 quarter and the first half. Stated so no reader mistakes a vendor annual figure for a filing-verified one.

5. tangibleBookValuePerShareTTM of $51.59 does not deduct preferred stock or noncontrolling interests — REJECTED and rebuilt. For a bank this is the most consequential error in the file. The figure equals tangibleAssetValueTTM of $157.332B ÷ ~3,049.5M shares, where the numerator is TOTAL equity less goodwill only. Tangible COMMON equity must exclude the $16.608B of preferred stock and the $1.921B of noncontrolling interests, per the 10-Q's own definition: "Tangible book value per common share is tangible common equity divided by common shares outstanding." The vendor implies P/TBV of 1.71x; the correct figure is 1.92x on a goodwill-only basis ($46.14 a share) or 2.01x deducting other intangibles as well ($44.05). We use 1.92x and disclose the range. Related rejection: bookValuePerShareTTM $59.78 is total stockholders' equity per share including preferred — arithmetically consistent but the wrong numerator; common book value is $54.40. And returnOnEquityTTM of 12.57% uses the same inflated denominator; ROTCE of ~15.9% trailing is the correct measure.

6. weightedAverageShsOutDil carries the FY2025 ANNUAL figure into three consecutive quarters — SHARE-COUNT DEFECT. Q3 2025, Q4 2025 and Q1 2026 all read exactly 3,217,400,000, which is also the FY2025 annual diluted count. Only Q2 2026 (3,080,000,000) and the 2024–Q2-2025 quarters (3,486.2M, 3,425.1M, 3,360.7M, 3,321.6M, 3,267.0M — a smooth decline) look like genuine quarterly figures. Any per-share figure derived from those three quarters is unreliable and we do not derive one. epsDiluted is independently corroborated by the filing — $2.00 for Q2 2026, $3.60 for the first half, $1.60 for Q2 2025, $2.98 for H1 2025, all verbatim from the 10-Q — and is used instead throughout. (Note the June-quarter diluted count of 3,080.0M against a basic count of 3,049.5M: the 30.5M gap is substantially the Series L convertible preferred, which the 10-Q states converts at 6.3814 common shares each.)

7. bottomLineNetIncome equals netIncome in Q4 2025 and Q1 2026 — preferred dividends are not deducted. Elsewhere the two differ correctly (Q2 2026: $6,407M against $6,160M, a $247M preferred deduction). The 10-Q confirms Series L preferred alone carries $74 million per quarter and the 2025 cash-flow statement shows $1,050 million of preferred dividends paid. Net income to common for those two quarters is overstated by roughly $250M each and is not used.

8. earn_cal EPS actuals CONTRADICT the filing — the filing wins. For the June 2026 quarter earn_cal reports $1.96 against the 10-Q's stated $2.00; for March 2026, $1.56 against $1.60; for June 2025, $1.54 against the 10-Q's $1.60. The vendor reports $1.96 for the June quarter; the 10-Q says $2.00; we use $2.00 and every beat/miss figure in this dive is restruck on it (the June-quarter beat is 15.6% against consensus of $1.73, not the 13.3% earn_cal implies). Revenue actuals, by contrast, are reliable in four of six rows and match our rebuild to the dollar for Q2 2026 ($22,622M), Q1 2026 ($21,446M) and Q3 2025 ($21,436M).

9. A recurring ~$253 million presentation difference between the vendor's quarterly build and both earn_cal and the filing — DIAGNOSED, not smoothed. For Q2 2025 the vendor's noninterest income and noninterest expense are each approximately $253M LOW relative to the filing; for Q4 2025 the vendor's revenue is approximately $253M HIGH relative to earn_cal. Pre-tax income and EPS are unaffected in both cases because the offset is symmetric — but the efficiency ratio for those two quarters moves by roughly 45 to 80 basis points. The diagnosis is verified: the 10-Q's own June-quarter bridge ("a $1.2 billion increase in noninterest income… a $282 million increase in noninterest expense") only reconciles once Q2 2025 is grossed up by $253M on both lines, at which point both the fee and the expense deltas match the filing exactly (+$1,191M and +$282M) alongside net interest income (+$609M) and tax (+$486M), which already matched to the dollar. We show the adjusted Q2 2025 row in Section 3 and flag Q4 2025 rather than adjusting it, because the annual FY2025 rebuild ties exactly to seg_prod on the unadjusted basis and we will not break a verified tie to fix an unverified one.

10. seg_prod stacks the CURRENT four-segment taxonomy on top of a SUPERSEDED one — DEFECT CLASS 7, confirmed. FY2025 uses Consumer Banking and Lending / Commercial Banking / Corporate and Investment Banking / Wealth and Investment Management, which the 10-Q confirms verbatim as the four current reportable segments. FY2024 and every year back to FY2020 uses "Community Banking" and "Wholesale Banking" — a structure Wells Fargo abandoned. FY2019 is not a segment table at all: two lines reading "Credit Card Rewards and Rebates" $1.5B and "Trailing Commission" $1.2B, which are revenue-disaggregation footnote fragments. The FY2025 rows sum to $83,699M and tie to our rebuilt total revenue to the dollar; the FY2024 rows sum to $83,759M against a rebuilt $82,296M — a $1,463M (1.8%) overstatement, because the FY2024 vintage omits the Corporate and Reconciling Items lines the FY2025 vintage carries. ONLY FY2025 is used and NO year-on-year segment comparison is drawn. Compounding this, the 10-Q discloses that certain business customers were transferred from Commercial Banking to Consumer Banking and Lending in the third quarter of 2025, which contaminates segment comparisons even within the filings.

11. seg_geo is EMPTY — stated as a gap, not estimated around. No geographic revenue disaggregation exists in any source in this file. The only geographic disclosure recoverable is on the lending side: non-US commercial and industrial loans of $81.9 billion at 2026-06-30. No geographic revenue figure appears anywhere in this dive.

12. Operating cash flow: FY2025 is −$19,001M while the trailing figure implied by ratios_ttm is +$19,010M. The magnitudes agree to 0.05% and the signs are opposite. We cannot resolve which is right from the archive and we do not need to, because operating cash flow for a bank is dominated by trading-asset, loans-held-for-sale and deposit flows and is not a measure of cash generation. priceToFreeCashFlowRatioTTM (13.943x), freeCashFlowYieldTTM (7.17%), priceToOperatingCashFlowRatioTTM (14.061x) and freeCashFlowToEquityTTM ($102.295B) are all REJECTED as lenses.

13. capitalExpenditure is $0 in all four cash-flow years and every capex-derived ratio is zero. capexToOperatingCashFlow, capexToDepreciation, capexToRevenue, capexPerShare all read 0. Banks do capitalise premises, equipment and internally developed software — the 10-Q explicitly names "amortization of internally developed software" as a rising expense line — but the outlay is folded into "other investing" and cannot be separated. Free cash flow therefore equals operating cash flow by construction in this file, which is why freeCashFlowOperatingCashFlowRatioTTM reads exactly 1.

14. Zeroed and unstable income-statement fields. depreciationAndAmortization is $0 in the June 2026 quarter against $1,836M, $2,123M and $1,839M in the three preceding quarters — the annual figure was $7,713M in 2025. stockBasedCompensation is $0 in all four cash-flow years, which is demonstrably wrong: the 8-K of 2026-04-30 records shareholder approval of an amended and restated 2022 Long-Term Incentive Plan, and the 10-Q records $875 million of common stock issued in the first half "substantially all of which was issued in connection with employee compensation and benefits." researchAndDevelopmentExpenses is $0 throughout, which is correct for a bank. And the sellingGeneralAndAdministrativeExpenses / otherExpenses split is not on a stable basis: $11,472M / $2,189M in the June 2026 quarter against $1,766M / $12,694M in the March quarter — the two lines effectively swap. No conclusion in this dive uses either; only the operatingExpenses total, which reconciles exactly to reported pre-tax income in every period.

15. The FY2029 estimate row is on a DIFFERENT REVENUE BASIS — discarded entirely. revenueAvg of $129.241 billion is a 32.7% jump from FY2028's $97.398 billion while epsAvg rises only 8.5%. That row appears to be struck on the gross-interest-income basis (it sits close to the vendor's gross-basis FY2025 revenue of $123.5 billion) while FY2026-FY2028 are on the bank basis. It rests on ONE revenue analyst and three EPS analysts. Discarded in full; no conclusion touches it.

16. est.ebitdaAvg/ebitAvg pass every measured defect signature — and are rejected anyway, on a different ground. Tested across all eight years: ebitAvg is below ebitdaAvg in every year, no negative EBITDA rows, and no 50%-of-revenue / 33%-of-revenue fabrication pattern. This file is clean on the programme's measured signatures. The rows are nonetheless not used, because EBITDA is not a meaningful measure for an institution whose principal expense is interest. epsAvg was verified against actuals and is RELIABLE: FY2025 $6.31563 against a reported $6.32 (0.06% off), FY2024 $5.33341 against $5.37, FY2023 $4.84767 against $4.83. All forward valuation in this dive runs on epsAvg.

17. Estimate coverage collapses after FY2027 — the thin rows are named and excluded. FY2026: 15 EPS, 9 revenue. FY2027: 17 EPS, 9 revenue — the deepest in the set and the basis of every fair-value anchor. FY2028: 9 EPS but only TWO revenue analysts, so the FY2028 revenue line is used nowhere. FY2029: 3 EPS and ONE revenue analyst — excluded outright (see finding 15).

18. quote.yearHigh/yearLow disagree with tech.hi52/lo52 — we use tech and say why. The quote block reports $97.76 / $72.78; the computed technical block reports $96.39 / $73.42. The gaps are 1.4% and 0.9%. We use the tech figures throughout, because they come from the same six-year close series that produces the moving averages, drawdown and relative-return figures, and mixing sources inside the technical section would make the percentages inconsistent. (Note: tech.max_dd_from_peak of −8.30% equals tech.pct_from_hi exactly. That is NOT a defect — it correctly says the current price IS the worst point of the trailing-year drawdown.)

19. lastDividend: 1.8 is the trailing-twelve-month total, not the declared rate — corrected from the filing. dividendPerShareTTM reads the same 1.80, giving a 2.04% yield. The 10-Q filed 2026-07-28 states: "On July 28, 2026, the Board approved an increase to the Company's third quarter 2026 common stock dividend to $0.50 per share." The forward annualised rate is $2.00, a 2.26% yield. The vendor reports $1.80; the filing shows a $0.50 quarterly declaration; we use $2.00. (A note for completeness: the archive's three 8-Ks are dated 2026-04-30 — the annual meeting results — 2026-05-20 — the $6.0B note issuance — and 2026-07-14 — the earnings-release cover page. NONE of them is a CCAR or stress-test capital-return announcement, so the dividend and buyback figures used here come from the 10-Q's Capital Management section rather than from an 8-K.)

20. rating sub-scores are non-bank templates applied to a bank — REJECTED as lenses. The composite reads B− / 2 overall, dragged there by debtToEquityScore 1 and discountedCashFlowScore 1. A debt-to-equity ratio of 2.548x is a funding structure, not a solvency signal, and a discounted-cash-flow score built on a −$19 billion operating cash flow is meaningless. priceToBookScore 2 is struck on the inflated book value of finding 5. The only sub-score that carries information is returnOnEquityScore 4, and even that uses the wrong denominator.

Currency test — PASSED and reported. inc_a[0].reportedCurrency is USD, profile.currency is USD, and the NYSE quote is in USD. No mixing anywhere in the file. Every multiple in this dive is single-currency.

Not defects, correctly reported and independently confirmed. ratios_ttm.priceToEarningsDilutedRatioTTM of 12.7547x reproduces exactly on our rebuild ($88.39 ÷ $6.93 = 12.75x). effectiveTaxRateTTM of 16.506% reproduces exactly ($4,533M of tax on $27,463M of pre-tax across the four quarters). netIncomePerShareTTM of $7.4265 reproduces ($22,647M ÷ 3,049.5M). The FY2025 seg_prod tie to $83,699M is exact. est.epsAvg is verified against three years of actuals to within 0.7%. The inc_a/inc_q pre-tax reconciliation holds to the dollar in fourteen of fifteen periods, and the derived provision series independently reproduces Wells Fargo's known credit history — a $14.129 billion build in 2020 and a $4.155 billion RELEASE in 2021 — which is the strongest available evidence that the decomposition is correct. And the June-quarter net income, first-half net income, and both years' diluted EPS all match the 10-Q verbatim, which is a genuinely clean vendor-to-filing reconciliation on the numbers that matter most.

Share count — a gap, stated rather than papered over. Market capitalisation of $267,291,360,000 ÷ $88.39 = exactly 3,024,000,000, a suspiciously round number. The 10-Q cover page's "Indicate the number of shares outstanding of each of the issuer's classes of common stock" line is PRESENT in the extraction but the value itself sits in a stripped table, so there is NO filing-verified share count in this archive. Cross-checks that make 3,024.0M plausible without confirming it: the June quarter's basic weighted average is 3,049.5M and diluted 3,080.0M; 84 million shares were repurchased during the first half against $875M of issuance; a period-end count slightly below the quarterly average is exactly what that implies. Every per-share figure in this dive that depends on the count is labelled, and the two that matter most — trailing P/E and forward P/E — do not depend on it at all, because they use reported diluted EPS.

Non-equity tripwire — checked and passed. WFC is common stock. The Form 4 filings in the insider block name the security as "Common Stock, $1 2/3 Par Value"; CUSIP 949746101, ISIN US9497461015, NYSE-listed. The price of $88.39 is not par-like ($25 or $1,000); beta is 0.917, far above the sub-0.3 typical of a baby bond; the dividend is variable and was just raised to $0.50 from a lower rate, rather than fixed; the 52-week band of $73.42 to $96.39 is a 31.3% range, the opposite of the narrow band a fixed-income-like instrument shows; and volume was 12.45 million shares, roughly $1.10 billion of turnover. One point requiring explicit care: the 8-K cover pages list six separate NYSE-listed preferred series (Series L, Y, Z, AA, CC, DD) plus a guarantee of Wells Fargo Finance LLC medium-term notes, all registered under Section 12(b). This dive covers the common stock and none of those instruments. The Series L convertible preferred is the reason the diluted share count exceeds the basic count by 30.5 million.

9. Technicals

• Price $88.39, +0.57% (+$0.505) on the day. Previous close $87.885, opened $87.82, day range $87.37–$89.43, closed 3.3% below the day high and 1.2% above the low.
• Volume 12,450,483 against a 19,933,645 average — 62.5% of normal. This was a quiet drift, not a participation event. The rest of this batch saw a violent semiconductor melt-up; Wells Fargo did not take part, and neither did the financials broadly (Citigroup +2.43%, American Express +0.62%). No company-specific news appears in this file for 2026-08-04; the last company event was the 10-Q filing on 2026-07-28, seven days earlier.
• −8.30% from the 52-week high of $96.39, and +20.39% above the 52-week low of $73.42. Position within the annual range: 65th percentile — the upper third, but not extended.
• Moving averages, both reclaimed and both supportive: +5.32% above a 50-day average of $83.93 and +4.18% above a 200-day average of $84.84. The 50-day sits above the 200-day, which is the constructive configuration.
• RSI 53.1 — neutral, with no overbought condition to work off and no oversold bounce to fade. MACD +0.77 — positive.
• Maximum drawdown from peak over the trailing year: −8.30%, i.e. the current price is the worst point of a shallow drawdown. For context, that is a quarter of the drawdown carried by the semiconductor names in this batch.
• Relative performance — a laggard that has just turned:

That table is the whole technical argument. Wells Fargo has underperformed the S&P by 10.6 points over twelve months and by 15.3 points over six — through the exact period in which the Federal Reserve removed its asset cap (2025-06-03) and terminated its consent order (2026-03-05). The three-month column is the first period in which it leads, and it leads by four points. A stock that has lagged badly, has reclaimed both moving averages, sits at a neutral RSI, and has just delivered its best operating quarter in two years is the setup this section exists to identify.

• Sentiment is oddly split and worth reading carefully. 0 strong buy, 27 buy, 29 hold, 4 sell across 60 analysts, composite rating "Hold". Consensus target $98.14 (+11.0%), median $94 (+6.3%), high $108 (+22.2%) — and low $92, which is 4.1% ABOVE the current price. Not one analyst in the set carries a target below spot. A street that will not call it a buy in aggregate but cannot construct a bear target is telling you it sees limited downside and limited conviction — which is close to our own reading.

Today's move and what it does to the entry

A +0.57% day on 62% of average volume changes essentially nothing about the entry, and that is itself worth stating plainly. Elsewhere in this batch, several names gapped 5-17% on sector beta and a buyer today would be paying for a one-day move. Here you are not. The stock closed within the middle of a $2.06 range, on light volume, having sat out the melt-up next door.

What the chart says about where to want it. The 50-day moving average at $83.93 is 5.0% below the close and the 200-day at $84.84 is 4.0% below. Both have been reclaimed and both would be natural support on a pullback. A fill in the $84-85 zone would put our $99 base at +17% instead of +12% and would move the asymmetry from 2.39:1 toward 3:1 — but it also requires the stock to give back the entire post-earnings advance, which it has not done in three weeks. The honest read is that $88.39 is a workable entry rather than an exceptional one, and that the position is best built rather than taken at once.

What would invalidate the technical read: a close below the 200-day average of $84.84 on company-specific news rather than sector beta, which would put the 52-week low of $73.42 back in play and would coincide with the bear case's $76.

10. Insiders — eight transactions, and not one of them is a purchase or a sale

Every transaction in this file is administrative. Read carelessly, the table looks like a chief risk officer disposing of his entire direct holding; it is nothing of the kind.

The reading, stated carefully, because two of these rows are easy to misread.

The Flowers transaction is a GIFT, not a sale — and the paired rows are one event, not two. The two entries are the same 67,966 shares moving from direct ownership to indirect ownership on the same day at a price of $0, coded G-Gift. That is the signature of an estate-planning transfer into a trust or family entity. His total beneficial ownership did not fall: it went from 67,966 direct plus 273,773 indirect to zero direct and 341,739.57 indirect. No shares left his household and no cash was received. Reporting this as "the chief risk officer sold his entire stake" would be flatly wrong.

The Rosenberg transactions are a vesting event, not a sale. The M-Exempt rows record 17,217.80 Restricted Share Rights converting into 17,217.80 shares of common stock. The F-InKind row records 8,079.49 shares withheld BY THE ISSUER at $83.73 to satisfy the tax liability on that vesting. These are not open-market sales. The employee never receives those shares and never sells them; the company retains them and remits cash to the tax authority. He ended the day holding 21,569.86 more shares than he started with, after tax.

The three director rows are quarterly retainer awards of Phantom Stock Units on 1 July, the first day of the quarter — a scheduled compensation event, not a discretionary purchase.

What the file therefore contains is: ZERO open-market purchases and ZERO open-market sales. That is a neutral signal and it is reported as neutral. It is worth noting only what it does NOT contain: no executive sold a share on the open market in the run-up to, or the aftermath of, a quarter that delivered a 25% year-on-year EPS increase and the best efficiency ratio in two years. That is a mild positive and no more than that.

11. Verdict, kill-criteria and flip conditions

Buy — Tactical.

What is genuinely attractive, and every item is filing-verified:

The regulatory constraint that defined this company for seven years is gone — the Federal Reserve removed the asset-growth cap on 2025-06-03 and terminated the February 2018 consent order on 2026-03-05 — and the balance sheet has responded, up $133.6 billion in six months to $2.3 trillion. The June 2026 quarter was the best in the file: $2.00 of diluted EPS against $1.60, total revenue of $22.622 billion (+8.6%), and an efficiency ratio of 60.39% against 64.25%, a 386 basis point improvement. Credit is improving on every disclosed measure — allowance coverage down to 1.40% of loans, criticized CRE mortgage loans down from $13.4 billion to $11.8 billion, nonperforming assets down $559 million to 0.77% of loans, commercial charge-offs at 10 basis points against 18. Capital is comfortably above requirement — CET1 10.26% against 8.50%, TLAC 22.81% against 21.50%, LCR 119% against 100%. And the capital return is enormous: 84 million shares retired for $7.1 billion in the first half, $22.7 billion of authority remaining, and a dividend just raised to $0.50 a quarter — roughly 7.6% of market capitalisation a year. All of that is available at 12.75x trailing earnings, 11.09x the FY2027 consensus, and roughly 1.92x tangible common equity for a bank earning ~15.9% on it.

What we are honest about:

Revenue compounded 2.4% a year over five years. This is not a growth business and the Growth Quality score of 5 says so. Essentially all of our +12.0% base-case return is a multiple moving from 11.09x to 12.4x — a judgement about what other people will pay, not about what the company will earn. Our base of $99 sits only 0.9% above the street's mean of $98.14, so we have very little analytical edge on the number itself. The June quarter's fee income included venture-capital gains, realized and unrealized, that the 10-Q describes but does not size — and the archive's stripped tables mean we cannot size them either. The knowledge base contributed thirteen claims with ZERO named speakers and exactly one about this company. And the 10-K is a wrapper with no financial statements in it, so the entire annual series in this dive rests on the vendor payload rather than a filing.

The distinction that makes this a Buy rather than a Watch, and Tactical rather than Core. Unlike the highest-quality name in this batch — where our fair value sat below the street on an empty conviction lane, which is a Watch by construction — here the arithmetic is genuinely favourable: 14.0% of downside against 33.5% of upside, a 2.39:1 asymmetry, with a 7.6% shareholder yield underneath it that pays you while the re-rating either happens or does not. And unlike a Core position, the thesis rests on a multiple rather than on a compounding engine, which is exactly the kind of view that should be sized as a tactical allocation and reviewed at each print rather than held through anything.

Pre-registered KILL criteria — what would take this to Avoid:

• Allowance coverage of total loans back above 1.50% in any quarter, or a disclosed re-build in the commercial real estate allowance. The improvement in coverage from 1.45% to 1.40% is what made the last two quarters look good; reversing it is roughly $1.0-1.5 billion of pre-tax provision per 10-15 basis points.
• Criticized CRE mortgage loans back above $13.4 billion, the December 2025 level — that would say the office and apartment improvement the 10-Q names was a cycle low, not a trend.
• The efficiency ratio back above 66% in any quarter, which would reverse two years of expense progress and directly attack the FY2027 consensus.
• A quarterly buyback below roughly $2.5 billion while total assets continue to grow — the signal that risk-weighted-asset growth has won the competition for capital, and the point at which the 7.6% shareholder yield stops being the floor under the thesis.
• CET1 below 9.75%, which is 125 basis points of headroom and the level at which capital return would face genuine constraint.
• Any new Federal Reserve or OCC enforcement action, or an escalation of the outstanding 2024-09-12 OCC formal agreement on anti-money-laundering and sanctions risk management, which the 10-Q filed 2026-07-28 still describes as in force alongside continuing government inquiries. The entire re-rating thesis is that the regulator is finished with this bank; a new order falsifies it directly.
• Two consecutive quarters of declining net interest income, which would say the post-asset-cap growth is dilutive rather than accretive.

Pre-registered FLIP conditions — what would take this to Buy — Core:

• A 2026-10-13 print with an efficiency ratio below 62% and allowance coverage at or below 1.40%. That validates both legs of 6a simultaneously — the expense improvement is structural and the credit improvement is not a one-quarter release — at which point the re-rating stops being an assumption and becomes an observation.
• Two consecutive quarters of net interest income growth WITH a stable or rising net interest margin. That would resolve the tension the 10-Q names — growth in lower-yielding Markets assets diluting margin — in the favourable direction, and would convert the asset-cap removal from a step-change into a genuine growth story.
• The full $22.7 billion of buyback authority executed at or below $92 while CET1 stays above 10%. That is the proof that the capital return and the balance-sheet growth are not mutually exclusive after all, which is the single biggest open question in Section 6.
• Genuine name-level knowledge-base conviction appearing with a NAMED SPEAKER. At present the lane has one company-level claim and zero named speakers across thirteen hits. A single well-sourced, named, name-level view would be worth roughly one tier on its own.
• A price below approximately $82 with no deterioration in the disclosed fundamentals, which would put the base case at +21% and the asymmetry near 4:1.

Pre-registered downgrade to Watch — if the multiple re-rates to 12.4x without the operating evidence arriving (i.e. the stock reaches $99 on sentiment while the efficiency ratio drifts back toward 64%), the thesis has been paid out and the position should be closed rather than held. This is a valuation-and-catalyst position, and it has a natural exit.

Where WFC fits in the Synthos Framework Portfolio. The US large-cap financials sleeve, at 0% today, with a 2.0-2.5% target built in two tranches — one at $88.39 and one on any retest of the $84-85 moving-average zone. On the batch overlap question: Citigroup, in this same batch at $136.82 (+2.43%), is the closest analogue — another large US bank trading below its franchise value on a multi-year restructuring, and holding both is one trade with two expressions rather than diversification. Within that pair, Wells Fargo is the cleaner expression: its regulatory overhang is documented as terminated rather than in progress, its capital return is running at a higher percentage of market capitalisation, and its credit metrics are improving on every disclosed line. The argument against it is that a re-rating thesis on a $267 billion bank is a crowded and obvious idea, which is precisely why the street's 60 analysts produce 27 buys, 29 holds and a composite "Hold" — everyone can see it and nobody will commit to it. Logged as a tracked Synthos call (Buy — Tactical) as of 2026-08-04 at $88.39.

Single biggest risk: a credit re-build, and specifically a commercial real estate re-build. The allowance for credit losses fell from 1.45% to 1.40% of loans because the commercial real estate allowance fell, and criticized CRE mortgage loans fell $1.6 billion in six months. That is the line with the shortest fuse in any bank income statement, and it is the one that took Wells Fargo's 2020 provision to $14.129 billion on a revenue base of $74 billion. A move back to 1.55% coverage on the approximately $1.029 trillion loan book is roughly $1.5 billion of incremental pre-tax provision — about $0.39 of EPS, or 4.9% of the FY2027 consensus, in a single quarter — and it would arrive alongside a multiple contraction rather than instead of one. That combination is what the $76 bear case is.

Most fragile assumption in the price: that the capital return and the balance-sheet growth can both run at full speed. Every other assumption in this dive is observable quarter by quarter — the efficiency ratio, the provision, the deposit cost, the CET1 ratio. This one is an arithmetic constraint that has not bound yet. Total assets rose $133.6 billion in six months, which consumes risk-weighted assets and therefore capital. The capital return is running at roughly $20 billion a year against trailing earnings to common of $22.152 billion — a 91% payout — funded from earnings plus roughly $23 billion of estimated excess CET1. At the current pace that excess is spent in about six quarters if risk-weighted assets keep growing. Either the growth moderates, or the buyback moderates, or CET1 drifts toward the 8.50% requirement. The price currently assumes all three continue as they are, and they cannot. If the buyback is the variable that gives, the 7.6% shareholder yield that is the floor under this thesis becomes 4-5%, and the +12.0% base case loses its cushion.

Provenance & disclosures

• Traceability: 13 tagged knowledge-base claims name Wells Fargo (raw hits 13; entity matches 9, text-only matches 4; used 6, discarded 7; breadth 6, net conviction positive-low). ZERO claims in the entire lane carry a named speaker — all thirteen are attributed to a CHANNEL only, which under the 4-lane attribution policy is the weakest sourcing we accept, and this is flagged as the primary limitation of the lane rather than buried. Sources: compound_and_friends (4 claims, 30.8% of the lane, of which three come from a single episode and two of those three are the same claim distilled twice), arthur_hayes (2, promoter-class frame disclosed and weighted low), darius_dale (2, both from one document), empire (2, both from one document, both dated 2020-06-04), business_breakdowns (2), we_study_billionaires (1). Discarded with reasons: one research-house collision (Wells Fargo cited as the source of index commentary in a market-breadth claim whose entity list is three indices and one bank), one lender collision (Wells Fargo as the provider of a $600M senior facility inside a Sunex/Centerbridge capital-structure claim), one illustration collision (Wells Fargo as a historical analogy inside a Robinhood regulatory-fines claim), one exact duplicate, and three same-document duplicates. Exactly ONE of thirteen is a claim about Wells Fargo as a company — a 2026-04-01 observation that the bank deployed more than $200 billion into repo after its asset cap was removed, which the 10-Q's disclosed $133.6 billion of six-month asset growth and its reference to "an increased volume of resale agreements" independently corroborate. The only bearish claim in the lane is dated 2020-06-04, is six years old, and has been falsified by events; it is reported and weighted at zero. No claim carries a speaker_role of management, so the half-weight management discount does not apply anywhere. No position reversals are visible in the dates. All quotes are verbatim from the stored claim text.
• Data as-of: fundamentals — filing-verified through 2026-06-30 from the 10-Q filed 2026-07-28 (June-quarter net income $6.4 billion, diluted EPS $2.00; first-half net income $11.7 billion, diluted EPS $3.60; total assets $2.3 trillion, up $133.6 billion from December 31, 2025; CET1 10.26%; allowance for credit losses on loans $14.4 billion at 1.40% coverage). Annual figures come from the VENDOR payload, not from a filing — see the archive note below. Estimates 2026-08-04 · prices 2026-08-04, quote timestamp 1785873754 ≈ 2026-08-04T20:00:01Z ($88.39, +0.57%; 50-DMA $83.93; 200-DMA $84.84; RSI 53.1; MACD +0.77) · knowledge-base claims 2026-08-04. Fiscal/calendar note: Wells Fargo's fiscal year is the calendar year. "FY2025" means the twelve months to 2025-12-31 and "Q2 2026" means the June 2026 quarter. All figures come from the Synthos vendor data file for WFC or from the SEC filings in the WFC archive; no figure comes from memory, recall or external retrieval — including, and especially, the regulatory status, which was established from the filings in both directions rather than asserted.
• Filing archive contents and its central gap: 10-K filed 2026-02-24 (fiscal 2025); 10-Q filed 2026-04-29 (March 2026 quarter); 10-Q filed 2026-07-28 (June 2026 quarter); 8-K filed 2026-04-30 (annual meeting results and 2022 Long-Term Incentive Plan amendment); 8-K filed 2026-05-20 ($6.0 billion of Medium-Term Notes, Series Y); 8-K filed 2026-07-14 (second-quarter earnings release cover page). THE 10-K IS A 64,639-CHARACTER WRAPPER AND THE ANNUAL FINANCIAL STATEMENTS ARE NOT IN THIS ARCHIVE. Part IV states the financial statements "are set forth in the 2025 Annual Report to Shareholders, and are incorporated into this item by reference" — the Annual Report exhibit was not extracted. The 10-K contains zero occurrences of "total revenue" and zero of "net interest income". Every annual figure in this dive therefore comes from the vendor payload, and the filing-verified anchor throughout is the 10-Q filed 2026-07-28. The 2026-07-14 8-K is present as the COVER DOCUMENT ONLY — Exhibits 99.1 (the news release), 99.2 (the 2Q26 Quarterly Supplement) and 99.3 (the presentation materials) were not extracted and the file contains no financial figures. NONE of the three 8-Ks is a CCAR or stress-test capital-return announcement, so the dividend and buyback figures come from the 10-Q's Capital Management section. A further, pervasive extraction limitation: numeric tables are stripped. This removes Table 2 (average balances and yields — hence the exact net interest margin is not recoverable and is not estimated), Table 13 (CRE loans by state and property type — hence the office dollar exposure and office-specific reserve ratio are not recoverable), Table 29 (capital components), Table 33 (the tangible common equity reconciliation — hence June 2026 tangible book value per share is not recoverable, and every price-to-tangible-book figure here is struck on the 2025-12-31 balance sheet), and the cover-page share count.
• Where the filings contradicted or corrected the vendor (detailed in Section 8): the vendor's income-statement revenue of $123,529M for 2025 is on a gross-interest-income basis; the filing's basis gives $83,699M, a 47.6% gap, and the entire revenue, expense and provision series was rebuilt and verified to reported pre-tax income to the dollar in fourteen of fifteen periods; tangibleBookValuePerShareTTM of $51.59 fails to exclude $16.608B of preferred stock and $1.921B of noncontrolling interests, and the correct figure of $44.05-$46.14 moves price-to-tangible-book from the vendor's 1.71x to 1.92x-2.01x; earn_cal reports June-quarter EPS of $1.96 against the 10-Q's stated $2.00, and we use $2.00; bal_a[0] total assets of $2,148.6B against the filing's approximately $2,282B, a 6.2% understatement, plus a $6.0B note issuance on 2026-05-20 that post-dates every balance sheet in the file; lastDividend of $1.80 is the trailing-twelve-month total against the 10-Q's disclosed $0.50 quarterly declaration of 2026-07-28, giving a forward rate of $2.00 and a 2.26% yield rather than 2.04%; weightedAverageShsOutDil carries the FY2025 annual count into three consecutive quarters; bottomLineNetIncome omits preferred dividends in two quarters; seg_prod stacks the current four-segment taxonomy on a superseded Community/Wholesale Banking one and a FY2019 row that is not a segment table at all; the FY2029 estimate row is on a different revenue basis and is discarded; and a recurring $253 million presentation difference in two quarters was diagnosed against the 10-Q's own year-on-year bridge and disclosed rather than smoothed. Where vendor and filing AGREED — worth recording, and it is more than the defect list suggests: June-quarter diluted EPS ($2.00), June-quarter net income ($6,407M against the 10-Q's "$6.4 billion"), first-half net income ($11,697M against "$11.7 billion") and first-half diluted EPS ($3.60), June-2025 diluted EPS ($1.60) and first-half-2025 EPS ($2.98 against a summed $2.99); the net interest income increase of $609 million and the tax increase of $486 million in the 10-Q's year-on-year bridge, both matching our rebuild to the dollar; the FY2025 seg_prod sum of $83,699M, matching our independently rebuilt total revenue exactly; priceToEarningsDilutedRatioTTM of 12.7547x and effectiveTaxRateTTM of 16.506%, both reproducing exactly; and est.epsAvg verified against three years of reported actuals to within 0.7%.
• Basis note: Wells Fargo does not report a non-GAAP earnings measure, and consensus is struck on GAAP. We verified it: est.epsAvg for FY2025 is $6.31563 against a reported GAAP diluted EPS of $6.32 — a 0.06% difference. There is therefore no GAAP-to-non-GAAP gap to disclose and no basis argument in this dive. The company does designate four measures as non-GAAP — tangible common equity, tangible book value per common share, return on average tangible common equity, and net interest income excluding Markets — and each is labelled as non-GAAP wherever used here.
• Estimate coverage: 15 EPS analysts and 9 revenue analysts on FY2026; 17 and 9 on FY2027 — the deepest coverage in the set and the basis of every fair-value anchor. FY2028 rests on 9 EPS analysts but only TWO revenue analysts, so the FY2028 revenue line is used nowhere and the FY2028 EPS line only as a bull-case cross-check. FY2029 rests on 3 EPS analysts and ONE revenue analyst, and its revenue figure appears to be struck on a different (gross-interest) basis entirely — it is discarded outright.
• Peer note: the vendor peer set contains three genuine US comparables — JPMorgan ($958.0B market capitalisation, $357.52), Bank of America ($446.4B, $62.90) and Citigroup ($234.7B, $136.82) — alongside six Canadian and European banks (Royal Bank of Canada, Toronto-Dominion, UBS, Bank of Montreal, Bank of Nova Scotia, CIBC, Barclays, ING) and one Latin American digital bank (Nu Holdings) that is not a comparable. No peer earnings, book values or multiples are supplied in the file, so NO peer-multiple comparison is drawn. The filing-sourced substitute is the 10-Q's own statement that Wells Fargo "ranked fourth in assets and fifth in the market value of our common stock among all U.S. banks at June 30, 2026."
• Fair-value caveat: the $76 / $99 / $118 anchors are multiples of the FY2027 consensus EPS distribution — 10.2x the low of $7.459, 12.4x the mean of $7.972, and 14.05x the high of $8.394 — each cross-checked against the December 2025 tangible common equity of $46.14 a share (1.65x, 2.15x and 2.56x respectively), with the base landing within 0.4% on both methods. Stated arithmetic, not a discounted cash flow. The base is deliberately sensitivity-disclosed: 11.5x gives $91.68, essentially spot; 13.5x gives $107.62. The whole answer lives in an 11.5x to 13.5x band and no greater precision is claimed. Every price-to-tangible-book figure in this dive is struck on the 2025-12-31 balance sheet, because the June 2026 tangible common equity reconciliation is in a stripped table and is not recoverable from this archive. The base case assumes multiple EXPANSION from 11.09x to 12.4x and that assumption is named repeatedly as the fragile leg; the bear case assumes no re-rating at all, only a lower number.
• Timing: second-quarter 2026 results were released 2026-07-14, twenty-one days before this dive, and the 10-Q carrying them was filed 2026-07-28, seven days before, so the entire first half is filing-verified. The next print is 2026-10-13, 70 days away — and consensus for it, at $1.84 of EPS on $22.369 billion of revenue, is BELOW the $2.00 and $22.622 billion just delivered, which is an explicit input to the variant perception in 6c rather than an oversight. 2026-08-04 was a broad semiconductor and technology up-day (ARM +17.36%, MRVL +12.81%, INTC +10.92%, LRCX +7.85%, KLAC +6.95%, PANW +5.53%, TXN +5.42%); Wells Fargo's +0.57% on 62% of average volume was not part of it, and no company-specific news appears in this file for the date.
• Accessibility note: no information in this dive is conveyed by colour. All emphasis is carried by bold text, table structure and explicit labelling.
• Not investment advice. Independent research, educational and informational only, never personalised. No recommendation to buy, sell or hold any security is made to any person.
• Version: 2026-08-04-full.