Bank of America BAC

Reference table

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

Bank of America is a bank holding company headquartered in Charlotte, North Carolina, with 213,000 full-time employees, $3.410 trillion of total assets at 2025-12-31 and Brian Thomas Moynihan as chairman and chief executive since 2010. NYSE-listed; CUSIP 060505104.

Four reportable segments plus All Other

From the 10-Q's own segment discussion:

• Consumer Banking — lending, deposit and investment products for consumers and small businesses.
• Global Wealth & Investment Management (GWIM) — two businesses, Merrill Wealth Management and Bank of America Private Bank.
• Global Banking — Global Corporate Banking, Global Commercial Banking, Business Banking and Global Investment Banking.
• Global Markets — sales and trading, and research, across fixed income, credit, currency, commodity and equities.
• All Other — the residual, which is negative.

Second-quarter 2026 net income by segment, from the 10-Q: Consumer Banking $3.3B (+$308M), GWIM $1.4B (+$420M), Global Banking $2.0B (+$348M), Global Markets $2.6B (+$1.1B). Those four sum to $9.3B against total net income of $9.1B, implying All Other at approximately −$0.2B — consistent with the vendor's own annual segment table.

The single most important disclosure point: which "revenue" is real

A bank has two revenue numbers and only one of them means anything. Gross revenue — interest income plus noninterest income — includes the interest the bank pays depositors and bondholders on the way in. The figure banks actually report and manage to is "total revenue, net of interest expense."

The vendor reports the gross figure and labels it revenue. For FY2025 it prints $191.567B. The correct figure is $113.097B — $191.567B less $78.470B of interest expense. Every margin the vendor computes off the gross number is therefore meaningless: a "gross margin" of 56.1%, an "operating margin" of 19.7% and a "net margin" of 15.9% are all divided by a denominator 69% too large.

Two independent checks confirm the $113.097B figure and they are worth showing, because they also prove the correct interpretation of the geographic table:

1. FY2025: $191.567B − $78.470B = $113.097B. The vendor's seg_geo FY2025 rows, once the double-count is removed (below), sum to exactly $113.097B.

2. FY2024: $192.434B − $90.547B = $101.887B. The vendor's seg_geo FY2024 rows sum to exactly $101.887B.

And the 10-K names the basis directly: the geographic disclosure presents "revenue, net of interest expense, income before income taxes and net income by geographic area." The vendor's geographic table is on the correct bank basis while its income statement is not.

Read on the correct basis, Bank of America's revenue grew 11.0% in FY2025 and its net margin is 27.0%, not 15.9%. The vendor's presentation makes the bank look both larger and far less profitable than it is.

Geography — and a $15.410B double-count

The vendor's FY2025 seg_geo block contains five lines that sum to $128.507B against a true total of $113.097B. The reason is arithmetic and unambiguous:

Asia + EMEA + Latin America = $6.004B + $7.561B + $1.845B = $15.410B — exactly the "Non-US" line. The vendor lists the subtotal AND its three components. Removing the double-count gives $113.097B, which is precisely revenue net of interest expense. The FY2024 row has no "Non-US" line and sums correctly to $101.887B, confirming the defect is a FY2025 presentation artefact rather than a data error.

Corrected geographic mix, FY2025: United States $97.687B, 86.4%; non-US $15.410B, 13.6%, of which EMEA $7.561B (6.7%), Asia $6.004B (5.3%) and Latin America $1.845B (1.6%). Bank of America is a domestic bank with a wholesale international overlay, and that concentration has been stable for a decade — the US share was 86.8% in FY2024 and 86.8% in FY2023.

The segment revenue table changes basis between years

The vendor's seg_prod block is not comparable year to year and must not be read as a time series. FY2024 and earlier use a revenue-line taxonomy (Loans and Leases, Debt securities, Card income, Investment Banking Fees and so on) summing to $192.434B — the GROSS revenue figure. FY2025 switches to a segment taxonomy (Consumer Banking, GWIM, Global Banking, Global Markets) summing to $116.760B, which is neither gross revenue nor net revenue but the four named segments' net revenue excluding All Other. The residual of −$3.663B against the $113.097B total is All Other, and it is consistent with the segment net-income arithmetic above.

FY2025 segment revenue, net of interest expense (vendor, basis verified against the segment structure in the 10-K):

Consumer Banking is the largest single business and the deposit franchise underneath it is the company's real asset: average Consumer Banking deposits were $957.0 billion in the second quarter, up $5.0 billion, with net inflows of $11.5 billion in checking and $3.8 billion in time deposits, partially offset by $10.4 billion of money-market and savings outflow. Money moving from savings into checking is a margin event as much as a balance event.

2. The quarter — what actually happened, from the filing

The 10-Q filed 2026-07-31, covering the quarter ended 2026-06-30, states:

> "Net income was $9.1 billion and $17.7 billion, or $1.21 and $2.31 per diluted share, for the three and six months ended June 30, 2026 compared to $7.2 billion and $14.5 billion, or $0.90 and $1.79 per diluted share, for the same periods in 2025. The increase in net income was due to higher noninterest income and net interest income, as well as lower provision for credit losses, partially offset by higher noninterest expense."

That is +34.4% on second-quarter EPS and +29.1% on the half. And on the margin:

> "Net interest income increased $1.3 billion to $16.0 billion, and $2.6 billion to $31.7 billion for the three and six months ended June 30, 2026... Net interest yield on a fully taxable-equivalent (FTE) basis increased 14 basis points (bps) and 12 bps to 2.08 percent."

Segment by segment, second quarter versus prior-year quarter, all from the 10-Q:

Total Corporation investment banking fees were $2.1 billion in the quarter, up 50%, and $4.0 billion in the half, up 35% — "due to higher debt issuance, advisory and equity issuance fees."

Here is the honest caveat on that table, and it is the reason the growth score is a 6 rather than an 8. Of the $1.9 billion increase in quarterly net income, $1.1 billion — 58% — came from Global Markets, and a further $420 million from GWIM whose fee growth the filing itself attributes to "higher market valuations." Roughly three-quarters of the earnings improvement came from the two businesses most directly levered to asset prices that are themselves at record levels. Consumer Banking, the largest and most durable segment, contributed $308 million, or 16%.

Credit is the genuinely reassuring part, and it is not a market-dependent line:

> "During the six months ended June 30, 2026, net charge-offs decreased $79 million to $2.1 billion compared to the same period in 2025, primarily due to improvement in the credit card portfolio."

> "Commercial reservable criticized exposure decreased $2.7 [billion]... our net charge-off ratio decreased compared to the same period in 2025 primarily driven by lower commercial real estate office charge-offs."

Commercial real estate office charge-offs falling and credit card asset quality improving, simultaneously, is the cleanest possible refutation of the 2023-vintage bear case that dominates this name's knowledge-base lane. One caveat the filing discloses in the other direction: Global Banking's six-month provision included "a qualitative reserve build related to uncertainties associated with the ongoing conflicts in the Middle East."

3. Trailing earnings, and one broken quarter in the vendor series

Trailing-twelve-month diluted EPS is $4.36 (the four reported quarters), giving a trailing P/E of 14.43x at $62.915. Note that the vendor's own ratios_ttm disagrees with itself here: netIncomePerShareTTM of $4.706 implies 13.37x, priceToEarningsRatioTTM of 14.266 implies $4.41, and priceToEarningsDilutedRatioTTM of 14.497 implies $4.34. The $4.706 figure does not deduct preferred dividends; the sum of the four reported quarters does. We use $4.36 and the 14.43x that follows from it.

Five consecutive beats, three of them by more than 7%, and the trailing revenue line has risen every single quarter — $26.463B, $28.088B, $28.367B, $30.272B, $31.558B. That is a 19.3% increase in five quarters. Consensus for the September quarter, $31.249B, implies a 1.0% sequential DECLINE, so the street has already stopped extrapolating.

The vendor's inc_q revenue series is internally inconsistent and must not be used. Eight of the nine quarters report GROSS revenue (Q2 2026: $49.393B). One — Q1 2026 — reports $30.272B, which is the NET figure and matches earn_cal exactly. A series that switches basis mid-stream will produce a fabricated 39% sequential collapse and a fabricated 63% rebound. Every quarterly revenue figure in this dive comes from earn_cal, which is consistently on the net basis, and is cross-checked against the filing.

4. Capital, not "balance sheet" — and every leverage metric in the vendor file is rejected

A deposit-funded bank does not have an enterprise value, a net debt position or an EV/EBITDA multiple, because its liabilities are its raw material rather than its financing. The vendor computes all three anyway, and they are the single largest category error in this file:

All nine are rejected. None appears in any conclusion in this dive. The vendor's own composite rating of C+ / 2, with a 1 out of 5 on debt-to-equity and 1 out of 5 on discounted cash flow, is produced by exactly these metrics and is likewise rejected as a lens.

What matters instead: regulatory capital, from the 10-Q

> "At June 30, 2026, the Corporation's CET1 capital ratio was 11.2 percent under the Standardized approach and 12.5 percent under the Advanced approaches."

> "At June 30, 2026, CET1 capital was $201.6 billion, an increase of $171 million from December 31, 2025, primarily due to earnings, largely offset by capital distributions and accumulated OCI."

> "RWA under the Standardized approach, which drove the lower CET1 capital ratio at June 30, 2026, increased $19.3 billion during the second quarter of 2026 to $1,792 billion."

> "At June 30, 2026 and December 31, 2025, the Corporation's minimum CET1 requirement was 10.0 percent under both the Standardized approach and the Advanced approaches."

> "The Corporation's global systemically important bank (G-SIB) surcharge is expected to increase on January 1, 2027."

Read the four facts together and the picture is a bank with a real but not generous cushion, shrinking on both sides. CET1 capital rose only $171 million over six months — earnings almost exactly offset by distributions and accumulated other comprehensive income — while risk-weighted assets grew $19.3 billion in a single quarter. The binding constraint, the filing says, is the Tier 1 capital ratio under the Standardized approach. Supplementary leverage exposure rose $135.8 billion in the quarter.

One regulatory tailwind, already banked: effective 2026-01-01 the corporation early-adopted the modified enhanced supplementary leverage ratio rule, which cut its leverage buffer from 2.0% to 0.75%. That is a genuine loosening — and it is already reflected in the current buyback pace, so it is not available again.

Tangible book value — the vendor's number includes preferred stock

For a bank, price to tangible book value per COMMON share is the primary valuation anchor, and the vendor's version is overstated.

bookValuePerShareTTM of $42.104 multiplied by the ~7.096B shares implied by market cap gives approximately $298.7B of equity — which is TOTAL equity including preferred stock. Bank of America's book value per COMMON share must exclude preferred. The 10-Q confirms preferred exists and was partially redeemed in the period: the change in Tier 1 capital is attributed to "a preferred stock redemption."

We derive tangible common book value two independent ways and they agree:

1. From the vendor's own figures: total equity ≈$298.7B, less preferred of roughly $28B, less goodwill of $69.021B and modest other intangibles ≈ $200B of tangible common equity, or ≈$28.2 per share.

2. From the filing's CET1 capital of $201.6 billion, which for an Advanced-approaches institution is common equity net of goodwill and intangibles with accumulated other comprehensive income included: $201.6B / 7.096B shares = $28.41 per share.

Two derivations, $28.2 and $28.4. We use ≈$28. That gives:

That difference is not cosmetic. At 1.94x the stock looks like a normally-valued large bank; at 2.24x it is at the upper end of where this security has historically been priced. Both figures are labelled as derived, and the derivation is shown so a reader can disagree with the preferred-stock estimate.

Return on tangible common equity, on the same basis: trailing net income to common of approximately $31.8 billion against roughly $200 billion of tangible common equity is ≈15.9% ROTCE — against the vendor's reported returnOnEquityTTM of 11.13% (which divides by total equity including preferred) and returnOnAssetsTTM of 0.96%. 15.9% ROTCE at 2.24x tangible book is close to the level a Gordon-growth frame would justify, and that observation is the entire valuation argument in Section 6.

5. Capital returns

FY2025 buybacks of $24.102 billion are 5.4% of the current market capitalisation and a 5.3x increase on FY2023. The authorisation behind it is stated in the 10-Q: "The Board authorized a $40 billion common stock repurchase program, effective August 1, 2025."

The dividend, and a labelling correction. The vendor reports lastDividend: 1.12 and dividendPerShareTTM: 1.12, a 1.78% yield. That is the trailing twelve-month total. The 10-Q gives the actual current rate:

> "On July 23, 2026, the Corporation's Board of Directors (Board) declared a quarterly common stock dividend of $0.32 per share, an increase of 14 percent compared to the prior quarterly dividend, payable on September 25, 2026."

The forward run-rate is therefore $1.28, a 2.03% forward yield, and the prior rate was $0.28. Payout ratio on trailing EPS of $4.36 is 29.4% — the vendor's dividendPayoutRatioTTM of 29.0% corroborates.

Total shareholder yield is approximately 7.4% — 2.03% dividend plus roughly 5.4% buyback. That is the strongest single fact in favour of owning this security, and it is the reason the verdict is Hold rather than Watch. It is also the fact most directly threatened by the January 2027 G-SIB increase, because a higher required CET1 ratio against a 120 basis point buffer is capital that cannot be distributed.

6. Valuation — priced in or room?

At $62.915 (market cap $446.482B, ~7.096B shares, trailing EPS $4.36, derived tangible common book ≈$28.2):

Estimate coverage is good on the near years and collapses after: 15 analysts on FY2026 and FY2027 EPS, 7 on FY2028, and 2 on FY2029. The FY2029 row is rejected outright — it shows revenue FALLING from $135.905B to $133.400B while EPS RISES 16.3%, which no coherent model produces, on two analysts.

The consensus revenue figures are on the correct net-of-interest-expense basis, and we verified it. First-half 2026 actual net revenue was $30.272B + $31.558B = $61.830B, annualising to $123.7B against the FY2026 consensus of $123.872B — a 0.1% difference. The estimate block uses the right revenue definition even though the income statement does not.

The est EBITDA and EBIT rows are REJECTED in full, and this file trips the corruption signature hard. In FY2026 through FY2029, netIncomeAvg EXCEEDS ebitdaAvg in every single year — FY2029 shows net income of $51.470B against EBITDA of $32.488B. Net income above EBITDA is arithmetically impossible for a taxpaying, interest-paying company. The historical rows are worse: FY2023 shows ebitdaAvg of $49.496B against ebitAvg of $8.993B, a 5.5x gap that no depreciation schedule explains. Every forward valuation in this dive runs on epsAvg.

A second estimate defect, less severe but worth stating: netIncomeAvg is inconsistent with epsAvg. FY2026's $35.820B divided by $4.633 implies 7.73 billion shares, against 7.294 billion diluted in the second quarter and falling at roughly 5% a year on the buyback. The niAvg rows are struck on a share count roughly 6% too high and are not used.

Peer context. The vendor peer set is genuinely useful here — Citigroup ($234.6B), Goldman Sachs ($310.8B), Wells Fargo ($267.1B), Royal Bank of Canada ($289.8B), HSBC ($364.5B), Bank of Montreal ($126.0B), Mitsubishi UFJ ($246.0B). Bank of America at $446.5B is the largest of them. No peer multiples are supplied in the file, so no peer-multiple comparison is drawn — which is a real limitation on a name where relative valuation is the standard method, and we state it rather than substitute recall.

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

At $62.915 — 12.0x FY2027 consensus EPS and roughly 2.24x derived tangible common book — the price embeds the following falsifiable claims:

• EPS reaches $4.633 in calendar 2026 and $5.242 in calendar 2027 — roughly 13% growth in the outer year. (Consensus-derived; 15 analysts each.) The near test: first-half 2026 EPS was $2.31, so the second half must deliver $2.32. With Q3 consensus at $1.19 that is nearly banked. The FY2027 step is the assumption doing the work, and it requires the current capital-markets environment to persist for another eighteen months.
• Revenue net of interest expense rises 9.5% in 2026 to $123.872B and a further 5.1% in 2027. (Consensus; 9 analysts.) Note that consensus for the September quarter, $31.249B, is 1.0% BELOW the June quarter actual of $31.558B. The street is modelling a plateau starting now.
• Return on tangible common equity holds near 15.9% and the market keeps paying roughly 2.2x tangible book for it. (Our derivation on both.) A Gordon-growth frame — justified P/TBV = (ROTCE − g) / (COE − g) — gives 1.98x at 15.9% ROTCE, 4% growth and a 10% cost of equity. The stock trades at 2.24x. The price already assumes either a higher sustainable ROTCE, a lower cost of equity, or both. That is the most fragile assumption on this page.
• The G-SIB surcharge increase on 2027-01-01 does not force a reduction in the buyback. (Filing-stated event; our assumption on the consequence.) The 10-Q says the increase is expected. CET1 is 11.2% against a 10.0% minimum, CET1 capital grew only $171 million in six months, and RWA grew $19.3 billion in one quarter. Roughly 5.4% of the 7.4% shareholder yield depends on this.
• Credit stays benign. (Filing-stated for the last two quarters; an assumption beyond that.) Charge-offs are falling and criticized exposure is down $2.7 billion. The provision is a swing factor of several billion dollars a year in either direction and nothing in the estimate rows visible to us models a turn.

Where our own signal has a reading: the knowledge base's single usable claim (compound_and_friends, 2026-01-20) is that "bank earnings are the economic source of truth; net charge-offs declining and credit-card stress absent — the awaited consumer rollover isn't here yet." The 10-Q filed six months later independently confirms it. That is corroboration of the credit assumption, which is the only assumption in 6a our library speaks to.

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

Expected return over the next twelve months decomposes as: EPS growth (+13.1%, FY2026E $4.633 to FY2027E $5.242) + multiple drift (COMPRESSION, from 12.0x to roughly 11.5x on the respective forward year, −4%) + shareholder yield (+7.4% — but note the buyback component is already inside the EPS growth, so counting only the 2.03% dividend avoids double-counting) ≈ +11%, against a base fair value that is 3.3% above spot.

Our base assumes mild multiple compression, and the reason is regulatory rather than fundamental. The bank's earnings power is not in question; its ability to keep converting that earnings power into per-share growth at the current rate is, because the January 2027 G-SIB increase raises the capital that must be retained. A bank returning 7.4% of its market value annually deserves a higher multiple than one returning 5%, and the difference between those two states is a regulatory decision that has already been announced.

To be explicit about the guardrail: 13.6x FY2026 becoming 12.0x FY2027 at a constant price is mechanical rolldown. The compression we assume is a change in what the market pays at each date — 12.0x forward today to roughly 11.5x forward in a year — and it is modest.

Most of the expected return here is earnings growth and dividend, which is the healthiest possible composition, and it is why this is not an Avoid. If the multiple simply HELD at 12.0x on FY2028E of $5.865, the price would be $70.4 (+11.9%); at 10.5x, $61.6 (−2.1%). The spread is narrow. This is a low-variance security and the valuation debate is correspondingly small.

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

• We do not differ from the street on the business, and we say so. Consensus is $66 (+4.9%) with 35 buys and 18 holds across 54 analysts; our base is $65 (+3.3%), 1.5% below. With no name-level knowledge-base conviction and a filing that corroborates the consensus narrative in every particular, we have no edge on this name and no-edge names default toward Hold. Manufacturing a differentiated view on the fourth-largest bank in the world, covered by 54 analysts, would be dishonest.
• We differ on the tangible book multiple, and it is the one number we think is misread. Every screen shows Bank of America at 1.94x tangible book using a per-share figure that includes preferred stock. On common tangible book we derive ≈2.24x, cross-checked two ways including against the filing's own $201.6 billion of CET1 capital. At 15.9% ROTCE, a Gordon frame justifies about 1.98x. The stock is roughly 13% above the multiple its own returns support, which is a materially different starting point from "cheap large bank."
• We differ on how to read the quarter's composition. The consensus framing is "broad-based strength across all four segments." The arithmetic is that 58% of the incremental net income came from Global Markets and a further 22% from a wealth business whose fee growth the filing attributes to "higher market valuations." Three-quarters of the improvement is levered to asset prices at highs. Watchable number: Global Markets revenue in the 2026-10-14 print against $8.0 billion. A reversion to $6.5 billion — the year-ago level — removes roughly $1.1 billion of quarterly net income, about 12% of the company's earnings.
• We differ on the capital picture, mildly but concretely. The headline is 11.2% CET1, comfortably well capitalized. The detail is a 120 basis point buffer, CET1 capital that grew $171 million in six months, RWA that grew $19.3 billion in three months, a leverage buffer already reduced from 2.0% to 0.75% by an early adoption that cannot be repeated, and a surcharge increase the company itself says is coming on 2027-01-01. Watchable number: the CET1 ratio in the 2026-10-14 print. A third consecutive decline, or any print below 11.0%, is a direct constraint on the buyback that supplies most of the shareholder yield.
• Positive surprise that would force a re-rate: a September quarter above $31.249B of revenue with the CET1 ratio stable or higher, which would show growth and capital generation coexisting; net interest yield extending materially above 2.08%; or a G-SIB surcharge increase that proves smaller than feared with the buyback pace maintained.
• Negative surprise that would break the thesis: Global Markets revenue reverting toward $6.5 billion; net charge-offs turning up, particularly in credit card where the current improvement is concentrated; the CET1 ratio falling below 11.0%; an announced reduction in the repurchase pace; or the Middle East qualitative reserve build growing into a genuine commercial credit event. Any of these hits a stock at its 52-week high with a 3.3% base-case return.

Synthos fair values

All three anchors are multiples of the FY2027 consensus EPS distribution (mean $5.242, low $5.112, high $5.381, 15 analysts — the deepest coverage in the file), cross-checked against a price-to-tangible-book frame on our derived ≈$28.2 of current tangible common book.

• Bear ~$49 — 9.5x the FY2027 consensus LOW of $5.112. Cross-check: roughly 1.6x current tangible common book, and close to the 52-week low of $44.92. The scenario: capital markets revenue reverts, the G-SIB increase forces a buyback cut, credit turns, and the multiple returns to where large banks trade in a normal rather than an exceptional year. −22.1%.
• Base ~$65 — 12.4x the FY2027 consensus MEAN of $5.242. Cross-check 1: 2.0x tangible common book — the multiple a 15.9% ROTCE supports on a Gordon frame at a 10% cost of equity and 4% growth. Cross-check 2: ≈2.0x forward tangible book of roughly $32.4 by end-2027, assuming retained earnings of about $2.2 per share per year. The two cross-checks bracket $63–$65 and the EPS anchor gives $65. Sensitivity, stated openly: 12x gives $62.9 — exactly spot — and 13x gives $68.1. The whole answer lives in a 12–13x band and we claim no more precision than that. +3.3%.
• Bull ~$75 — 13.9x the FY2027 consensus HIGH of $5.381, which is also the street's target high. Cross-check: ≈2.7x current tangible common book. The scenario: the capital-markets cycle extends, net interest yield keeps expanding past 2.08%, credit stays benign, the G-SIB increase is absorbed without cutting the buyback, and the market pays a premium multiple for a 16%-plus ROTCE. +19.2%.

Base is 3.3% above spot; asymmetry is roughly 0.87:1 — slightly unfavourable (22.1% down against 19.2% up). Add the 2.03% dividend and the skew is close to neutral. That neutrality, at a 52-week high, with no company-specific catalyst for 71 days and no knowledge-base edge, is the arithmetic behind the Hold.

7. Knowledge base — 38 hits, 18 of them a sentiment survey, and a speaker who reversed twice

Raw hits: 38. Entity matches: 18. Text matches: 20. Used as sector-level colour: 17. Used as name-level conviction: 0. Discarded: 21. Search terms: BAC, Bank of America, BofA, Merrill Lynch, Brian Moynihan, word-boundary matched across entity lists and claim text over all 51,928 distilled claims.

Discard class 1 — the BofA Global Fund Manager Survey (18 claims). This is the largest single collision found in this batch and it deserves naming. Bank of America publishes a monthly institutional positioning survey and a "sell-side indicator" that macro commentators cite constantly as a contrarian sentiment gauge. Every one of these references produces a text match on "BofA" and not one is a claim about the security. Representative examples, all discarded:

• "Remains bullish; this is a hated, under-owned rally — BofA survey most bearish since June '25." (2026-04-20, Andreas Steno) — a claim about the Nasdaq.
• "Long global semis is by far the most crowded trade in the BofA fund-manager survey." (2026-06-26, Sonu Varghese) — a claim about semiconductors.
• "Three of four institutions in BofA's survey have zero direct crypto exposure." (2025-08-18, Andreas Steno) — a claim about Bitcoin.
• "Bank of America's sell-side indicator near a sell signal is noise." (2026-01-06, compound_and_friends) — a claim about the S&P 500.
• "BofA's three-hikes call is hilarious — fade the hawkishness." (2026-06-26, forward_guidance) — a claim about the Fed, and about BofA's ECONOMISTS rather than its shareholders.

A naive entity or text search would report "38 claims" of coverage on Bank of America. Roughly half of them are about Treasuries, semiconductors, Bitcoin, the Nasdaq and the Fed.

Discard class 2 — already quarantined by the store (2 claims). Two records carry thinker: quarantine_misattributed with quarantine_class: NOT_VISSER:

• 2023-06-02, bearish, conviction 90: "Banks are technically insolvent, propped up by government guarantees."
• 2023-08-24, bearish, conviction 65: "Long-term bank charts look like the 'cliff of death' (Citi, BofA, regionals)."

Both carry skill: 2.0 — the highest weight in the store — because they were originally attributed to a speaker whose channel they were not. Had the attribution fix not caught them, the two most bearish claims in this lane would have been weighted double and credited to the highest-skill voice available. We honour the quarantine flags and exclude both. This is the attribution policy doing exactly the job it exists to do, and it is worth showing rather than silently filtering.

Discard class 3 — genuine name collisions and off-subject entity tags (1 claim). The 2024-06-08 we_study_billionaires claim is about Li Lu's concentrated 13F, which happens to include Bank of America among five names. It assigns no view to BAC.

What survives: 17 sector claims, zero name-level views

The surviving lane divides cleanly into three temporal groups, and the shape of that division is the most informative thing in it.

Group A — the 2023-2024 solvency panic (5 claims, all bearish, all wrong).

> 2023-10-05 · bearish · conviction 85 · thinker: luke_gromen · skill 1.0

> "Banks were regulated into buying $4.6T of low-yield Treasuries unhedged; at higher rates those losses wipe out capital, so stocks price the insolvency."

> 2023-11-01 · bearish · conviction 85 · entities: Bank of America, Morgan Stanley, Goldman Sachs · thinker: arthur_hayes · skill 1.2

> "Big banks are functionally insolvent on unrealized HTM treasury losses; state-backed but unable to lend, and banks #9-3000 hold the same problem in CRE."

> 2024-11-07 · bearish · conviction 70 · thinker: natalie_brunell · skill 1.0

> "Bank failures are just beginning — after First Republic, larger institutions like Bank of America look very bad; the number-one near-term concern."

These claims are dated, testable and were tested by events. Bank of America closed 2026-08-04 at its 52-week high with a CET1 ratio of 11.2%, falling charge-offs and $24 billion of annual buybacks. We report them because honesty requires reporting the record, not only the parts that worked, and because the same speakers still contribute to this lane.

Group B — the 2025-2026 payments and tokenisation debate (9 claims, split). The most substantive claims in the lane and the only genuinely forward-looking ones. Bullish for banks:

> 2025-07-02 · neutral · conviction 65 · thinker: arthur_hayes — "Banks' superior distribution beats crypto exchanges; real threat to Coinbase is a bank offering zero-fee Bitcoin trading." (A near-duplicate of the same speaker's 2025-06-11 claim; the two are collapsed to one.)

> 2025-03-21 · bullish · conviction 60 · thinker: empire — "Centralized issuers (brokerages, banks, large asset managers) hold the forward product-market fit for stablecoins over decentralized ones."

> 2025-09-10 · bullish · conviction 75 · thinker: forward_guidance — "Banks resisted blockchain to protect fees but are now investing billions; any bank not adopting will be left behind in 10 years."

> 2025-10-15 · bullish · conviction 72 · thinker: arthur_hayes — "Bessent's push to shift money creation to commercial banks means large banks expand credit."

Bearish for banks, from the same debate:

> 2025-03-21 · conviction 55 · thinker: forward_guidance — "Stablecoins are a narrow-banking middle ground that could meaningfully disrupt US commercial banking over 5-10 years."

> 2025-11-05 · conviction 80 · thinker: jordi_visser · skill 2.0 — "Stablecoins collapse the payment 'vig'/float; Genius Act deregulation triggers a race." Float is what a deposit franchise monetises.

The library cannot decide whether Bank of America is the disruptor or the disrupted, and we do not pretend it can. What is worth noting is that nothing in Bank of America's own filings quantifies its stablecoin or tokenisation position at all — so this optionality is unpriced in both directions and unmeasurable from here.

Group C — the one claim that is both relevant and verified (1 claim).

> 2026-01-20 · bullish · conviction 72 · entities: Bank of America, JP Morgan · thinker: compound_and_friends · skill 1.0

> "Bank earnings are the economic source of truth; net charge-offs declining and credit-card stress absent — the awaited consumer rollover isn't here yet."

This is the only claim in 38 that made a checkable statement about Bank of America's business and can now be checked. The 10-Q filed 2026-07-31 states that first-half net charge-offs decreased $79 million to $2.1 billion "primarily due to improvement in the credit card portfolio." The claim was correct, six months in advance, and it is worth more than the other sixteen combined.

The concentration and reversal problem, stated plainly

Arthur Hayes supplies 6 of the 17 surviving claims — 35% of the lane — and reverses across them:

From "losses don't exceed capital" to "functionally insolvent" to "banks are healthy" in three years, with one claim duplicated three weeks apart. Weighting that as a signal would be indefensible. It is disclosed and discounted to zero.

Net conviction: neutral. Breadth after collapsing by speaker: 9. Claim count used: 17. Name-level conviction: ZERO. The Synthos knowledge base has no view on Bank of America Corporation as a security. It has a stale and falsified bearish sector view, an unresolved argument about stablecoins, and one correct macro observation about consumer credit. That is stated as a finding, not padded, and it is a direct input to the Hold verdict.

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

Fifteen findings. This is the most defective vendor file in this batch, and the defects are structural rather than incidental: most of them stem from applying an industrial-company template to a bank.

1. revenue is GROSS of interest expense — every derived margin is REJECTED. The vendor prints FY2025 revenue of $191.567B. The figure banks report and manage to is total revenue, net of interest expense, which is $191.567B − $78.470B = $113.097B. The vendor's "net margin" of 15.9% is really 27.0%; its "gross margin" and "operating margin" are meaningless. Confirmed two ways: the vendor's own seg_geo sums to $113.097B (FY2025) and $101.887B (FY2024), exactly matching the derivation in both years, and the 10-K describes the geographic table as presenting "revenue, net of interest expense, income before income taxes and net income by geographic area."

2. The inc_q revenue series switches basis mid-stream — REJECTED as a series. Eight of nine quarters print gross revenue (Q2 2026: $49.393B). Q1 2026 prints $30.272B, the NET figure, matching earn_cal exactly. Read as a series this fabricates a 39% sequential collapse followed by a 63% rebound. All quarterly revenue in this dive comes from earn_cal, which is consistently net.

3. seg_geo FY2025 DOUBLE-COUNTS $15.410B — corrected. The block lists Asia $6.004B + EMEA $7.561B + Latin America $1.845B = $15.410B and then lists "Non-US: $15.410B" as a fifth line. The stated sum of $128.507B overstates by exactly the subtotal. Corrected total $113.097B. The FY2024 row contains no "Non-US" line and sums correctly, confirming this is a FY2025 artefact.

4. seg_prod changes taxonomy between FY2024 and FY2025 and is not a time series. FY2024 and earlier use revenue-line categories summing to $192.434B (gross revenue); FY2025 uses segment names summing to $116.760B (four segments, net revenue, excluding All Other). Only FY2025 is used, and only as a segment-mix snapshot with All Other derived at −$3.663B.

5. Enterprise value, net debt and every EV multiple are CATEGORY ERRORS for a deposit-funded bank — REJECTED in full. enterpriseValueTTM $1,149.922B; netDebtToEBITDATTM 15.41x; evToEBITDATTM 25.19x; evToSalesTTM 6.48x. A bank's liabilities are its raw material. None of these appears in any conclusion.

6. Liquidity and coverage ratios are likewise category errors — REJECTED. currentRatioTTM 0.073; workingCapitalTTM −$363.577B; netCurrentAssetValueTTM −$3,169.242B; interestCoverageRatioTTM 0.592x; grahamNetNetTTM −$443.18. Every solvent bank in the world fails these tests.

7. Cash-flow-derived metrics are noise — REJECTED. Operating cash flow was −$8.805B (FY2024), +$44.982B (FY2023), −$6.327B (FY2022) — swings driven by trading-asset and loan-balance movements, not by profitability. freeCashFlowYieldTTM of 21.2% and priceToFreeCashFlowRatioTTM of 4.71x are artefacts of that noise and are not used.

8. est.ebitdaAvg and est.ebitAvg are IMPOSSIBLE — REJECTED in full. In every forward year FY2026-FY2029, netIncomeAvg EXCEEDS ebitdaAvg — FY2029 shows net income of $51.470B against EBITDA of $32.488B. Historically, FY2023 shows ebitdaAvg of $49.496B against ebitAvg of $8.993B. This file trips the known corruption signature. All forward valuation runs on epsAvg.

9. est.netIncomeAvg is inconsistent with est.epsAvg — NOT USED. FY2026's $35.820B / $4.633 implies 7.73 billion shares, against 7.294 billion diluted in Q2 2026 and falling ~5% a year on the buyback. A ~6% share-count error. epsAvg is used throughout.

10. The FY2029 est row is incoherent — REJECTED. Revenue FALLS from $135.905B (FY2028) to $133.400B while EPS RISES 16.3% from $5.865 to $6.82, on 2 analysts. Not used for any conclusion.

11. bookValuePerShareTTM and tangibleBookValuePerShareTTM INCLUDE preferred stock — corrected. $42.104 × ~7.096B shares ≈ $298.7B, which is total equity including preferred. We derive tangible COMMON book of ≈$28.2 per share two ways — from the vendor's balance sheet less an estimated $28B of preferred and $69.021B of goodwill, and independently from the filing's CET1 capital of $201.6B divided by shares — and the two agree to within 0.7%. The consequence is material: price to tangible book is ≈2.24x, not the vendor's 1.94x. Both figures are labelled derived.

12. lastDividend: 1.12 is the TTM total, not the current rate — corrected from the filing. The 10-Q states: "On July 23, 2026, the Corporation's Board of Directors declared a quarterly common stock dividend of $0.32 per share, an increase of 14 percent compared to the prior quarterly dividend." Forward run-rate is $1.28, a 2.03% yield, not $1.12 / 1.78%.

13. The FY2024-to-FY2025 debt comparison in bal_a is not meaningful. shortTermDebt reads $375.149B (FY2024) and $48.088B (FY2025) — a $327 billion swing — with totalDebt moving from $658.428B to $365.904B. A bank does not repay $292 billion of debt in a year; this is a reclassification of repurchase agreements and federal funds purchased. The year-on-year debt comparison is flagged as broken and not used. bal_a also ends at 2025-12-31, two quarters before the latest filing.

14. The vendor composite rating of C+ / 2 is an artefact — REJECTED as a lens. debtToEquityScore 1 and discountedCashFlowScore 1 are generated by exactly the industrial-template metrics rejected above. It is a rating of the template, not of the bank.

15. quote and tech disagree on the 52-week range — we use tech. The quote block gives a year high of $63.54 (which is today's intraday high) and a low of $44.75; the computed technical block gives $62.915 and $44.92. We use tech throughout. Note that tech.hi52 of $62.915 equals today's close and pct_from_hi is exactly 0.0 — that is not a defect; it means today's close is the highest close of the trailing 252 sessions, which is the finding, not an error.

A minor filing-versus-vendor discrepancy, resolved in the filing's favour: earn_cal reports Q2 2025 EPS actual of $0.89; the 10-Q states $0.90. The filing wins; the difference does not affect any conclusion, and we use the filing's stated year-on-year comparison (+34.4%) rather than one computed from the vendor's figure.

Not defects, correctly reported and independently confirmed: the earn_cal revenue series matching the net-of-interest-expense basis exactly; the FY2025 and FY2024 seg_geo totals reconciling to derived net revenue in both years; dividendPayoutRatioTTM of 29.0% against our 29.4% computed on the four reported quarters; returnOnEquityTTM of 11.13% being arithmetically correct on total equity even though ROTCE is the right measure; the consensus FY2026 revenue of $123.872B annualising within 0.1% of first-half actuals; and the employee count and CEO name matching the filings.

Non-equity tripwire — checked and passed, and it matters here. Bank of America has a large public preferred-stock complex, several series of which trade under BAC-prefixed tickers and would inherit parent fundamentals in this feed. This file is the COMMON stock: CUSIP 060505104, ISIN US0605051046, NYSE, price $62.915 (not the $25 or $1,000 par-like level of a preferred or baby bond), beta 1.175, a variable dividend just raised 14% rather than a fixed coupon, a 52-week range of $44.92 to $62.915 (a 40% span, not the narrow band a fixed-income-like instrument shows), and volume of 23.08 million shares. This is common equity.

Extraction limits, stated rather than estimated around. Numeric tables are stripped from the extracted filings, so the actual balance-sheet, capital-table and segment-revenue figures exist in the 10-Q only where the MD&A states them in prose. The efficiency ratio, the exact preferred-stock balance, the loan and deposit totals and the allowance for credit losses are therefore NOT available from this archive and no figure for them appears in this dive. Earnings-call transcripts are not available on this data plan, so management Q&A — including any commentary on the G-SIB increase — is absent. The three 8-Ks in the archive (2026-05-06, 2026-07-14, 2026-07-24) are cover documents; the second-quarter earnings release itself sits in an unextracted Exhibit 99.

9. Technicals

• Price $62.915 — the 52-week HIGH. pct_from_hi is 0.0 and max_dd_from_peak is 0.0: today's close is the highest close of the trailing year. +40.1% above the 52-week low of $44.92. Position within the annual range: 100th percentile.
• Moving averages, both supportive and both well below: +9.0% above a 50-day average of $57.70; +17.2% above a 200-day average of $53.68. Both rising.
• RSI 58.89 — firm but NOT overbought, which is unusual for a stock at its high and suggests the advance has been orderly rather than parabolic. MACD +1.30, positive and meaningful at 2.1% of the share price.
• Relative performance: 3-month +20.5% vs SPY +7.6% and QQQ +7.7%; 6-month +16.4% vs SPY +11.1% and QQQ +15.7%; 12-month +37.8% vs SPY +24.3% and QQQ +30.8%. BAC has beaten the S&P by 13.5 points over twelve months and by 13 points over the last three months alone — the recent acceleration is the striking part.
• Sentiment: 35 buy, 18 hold, 1 sell across 54 analysts. Consensus target $66 (+4.9%), median $65, high $75 (+19.2%), low $59 — 6.2% BELOW spot. A well-covered, tightly-clustered distribution with limited implied upside.

Today's move and what it does to the entry

BAC closed 2026-08-04 at $62.915, up 0.70% or $0.435 from a previous close of $62.48. It opened at $62.47, traded $62.35 to $63.54, and closed 1.0% below the day's high on 23.08M shares against a 35.89M average — light volume. A new 52-week closing high on 64% of average volume is a weak-participation high, and that is worth noting even though a single session proves nothing.

The honest read. This is not a stock gapping out of a drawdown into value; it is a stock making a quiet new high after a 20.5% three-month run, at roughly 2.24x tangible common book, with our base fair value 3.3% above the close and the street's target 4.9% above it. There is no company-specific catalyst for 71 days, and the next dated event that matters is a regulatory one on 2027-01-01 that works against the buyback.

What we are telling the reader plainly: hold, collect the 7.4% shareholder yield, and do not chase a new high on light volume 3% below your own fair value.

10. Insiders — eight transactions, all one person, all the same mechanical event

Every transaction in this file belongs to one person and describes one mechanism, repeated monthly on the 15th at exactly 18,083 shares. The pattern is: cash-settled restricted stock units convert (M-Exempt), an equal number of common shares are acquired, and an identical 18,083 shares are returned to the issuer (D-Return). Holdings before and after are constant at 2,717,695 and 2,699,612 shares in every month. This is a scheduled, formulaic settlement of a cash-settled RSU programme, not a discretionary trade.

What the file contains is therefore: zero open-market purchases and zero discretionary sales. That is neutral and is reported as neutral. The three D-Return prices — $49.77 in May, $55.87 in June, $61.59 in July — are, incidentally, a clean 23.7% record of the share price advance over ten weeks, and nothing more.

Chief executive Brian Moynihan's retained holding of 2,699,612 shares is worth $169.9 million at today's close. It is substantial and long-tenured. The absence of any discretionary purchase or sale in a stock at its 52-week high is uninformative, and we say so rather than manufacture a reading.

11. Verdict, kill-criteria and flip conditions

Hold.

This is a very good bank having a very good year at a price that reflects it, and our job is to say so without inflation in either direction.

What is genuinely excellent, and all of it is filing-verified: second-quarter net income of $9.1 billion and diluted EPS of $1.21, up 34% year on year; net interest income up $1.3 billion to $16.0 billion with net interest yield up 14 basis points to 2.08%; all four segments growing net income; Global Markets revenue up $2.0 billion to $8.0 billion and investment banking fees up 50% to $2.1 billion; GWIM's operating margin at 27% against 22% with client balances up 12% to $4.9 trillion; net charge-offs FALLING $79 million to $2.1 billion on credit card improvement, and commercial real estate office charge-offs down; CET1 capital of $201.6 billion; a dividend raised 14% to $0.32 a quarter; a $40 billion repurchase authorisation; total shareholder yield of approximately 7.4%; and a derived return on tangible common equity of roughly 15.9%.

What we are declining to pay for: a close at the 52-week high with a drawdown from peak of exactly zero; roughly 2.24x tangible common book against a Gordon-frame justified multiple of about 1.98x at the current ROTCE; a base fair value of $65 that is 3.3% above spot and 1.5% below the street's $66; a 120 basis point CET1 buffer with CET1 capital up only $171 million in six months against $19.3 billion of quarterly RWA growth, and the filing's own statement that the G-SIB surcharge is expected to increase on 2027-01-01; earnings improvement that is 58% Global Markets and 22% market-valuation-driven wealth fees; and a knowledge-base lane with zero name-level conviction, 18 of 38 hits collided on a sentiment survey, and a dominant speaker who reversed twice.

The distinction that matters. We are not saying Bank of America is expensive in absolute terms — 12.0x forward earnings with a 7.4% shareholder yield is not expensive. We are saying that the specific things that would make it a buy — a discount to tangible book that reflects its returns, a capital buffer that is not about to be squeezed, an informational edge from our own library, or a price that is not at its high — are all absent simultaneously. In the Synthos frame, a well-run business at a fair price with no edge and a dated regulatory headwind is a Hold by construction.

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

• The CET1 ratio falling below 11.0% in the 2026-10-14 or subsequent print, or any announced reduction in the repurchase pace to absorb the January 2027 G-SIB increase. That removes roughly 5.4 points of the 7.4% shareholder yield, which is most of the reason to own it.
• Net charge-offs turning up for two consecutive quarters, particularly in credit card where the entire current improvement is concentrated. The provision is a multi-billion-dollar swing factor and nothing in consensus models a turn.
• Global Markets revenue reverting toward $6.5 billion — the year-ago level — which alone removes about 12% of quarterly earnings and would falsify the FY2027 EPS step in 6a.
• Net interest yield contracting below roughly 2.00%, which would say the deposit-repricing tailwind has reversed.
• The Middle East qualitative reserve build in Global Banking converting into realised commercial credit losses, or a material commercial real estate reversal after the current improvement.
• A price above roughly $75 without commensurate earnings revision, which would be roughly 2.7x tangible book and beyond any return-based justification.

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

• Price below approximately $55 — roughly 10.5x FY2027E and 1.95x tangible common book, the multiple our derived 15.9% ROTCE actually supports — with no deterioration in credit or capital. At $55 the base case is +18% and the skew inverts. That is 12.6% below today's close and inside the range this stock traversed in the last six months.
• The G-SIB surcharge increase being quantified as small, with the buyback pace explicitly maintained. That converts the single largest overhang into a resolved question and would justify paying today's multiple.
• Two consecutive quarters of CET1 accretion with the ratio above 11.5%, showing the bank can grow risk-weighted assets and capital simultaneously.
• Global Markets revenue holding at or above $8.0 billion for two more quarters, converting a cyclical spike into a level.
• A genuine name-level knowledge-base claim appearing — an argument about Bank of America the security rather than about the banking sector or the BofA survey. At present there is none in 38 hits, and that alone is worth roughly one tier.

Where BAC fits in the Synthos Framework Portfolio. The large-cap financials sleeve, core-yield expression. On batch overlap, which matters here: BAC, Morgan Stanley and Goldman Sachs are in this same batch and they are three expressions of one capital-markets cycle. Global Markets revenue up 33% and investment banking fees up 50% at Bank of America is the same trade that shows up in the other two, in more concentrated form. Bank of America is the LEAST levered of the three to that cycle — the capital-markets businesses are roughly 40% of its earnings against a far higher share at the other two — and correspondingly has the most durable deposit franchise and the lowest beta. If the capital-markets cycle is to be owned at all, BAC is the defensive expression; if it is to be avoided, BAC is the least urgent sale. Logged as a tracked Synthos call (Hold) as of 2026-08-04 at $62.915.

Single biggest risk: a capital-requirement increase against a 120 basis point buffer while returning $24 billion a year. The 10-Q states the G-SIB surcharge is expected to rise on 2027-01-01. CET1 is 11.2% against a 10.0% minimum. CET1 capital grew $171 million over six months while risk-weighted assets grew $19.3 billion in three. The bank is currently distributing approximately its entire earnings — dividends plus a buyback that consumed $24.102 billion last year — and a higher required ratio means some of that must be retained instead. At roughly 2.24x tangible book, a material part of the valuation rests on the distribution continuing at this rate. There is no cushion in the multiple to absorb a cut.

Most fragile assumption in the price: that the market keeps paying roughly 2.24x tangible common book for a 15.9% return on tangible common equity. Every other assumption on this page is about earnings, and earnings are visible quarter by quarter. This one is about what the market is willing to pay, and on a Gordon frame at a 10% cost of equity and 4% terminal growth, a 15.9% ROTCE justifies about 1.98x — roughly 12% below where the stock trades. The gap closes if ROTCE rises (which requires the capital-markets cycle to persist), if the cost of equity falls, or if the price falls. We are not confident which, and the honest position when the multiple is doing the work and we cannot explain why is to hold what we have and not add.

Provenance & disclosures

• Traceability: 38 raw knowledge-base hits on BAC, Bank of America, BofA, Merrill Lynch, Brian Moynihan across 51,928 distilled claims (18 entity matches, 20 text matches). 17 used as sector-level colour; ZERO used as name-level conviction; 21 discarded. Discards by reason: 18 are collisions on the BofA Global Fund Manager Survey, the BofA sell-side indicator or BofA research — claims about Treasuries, semiconductors, Bitcoin, the Nasdaq and the Fed that merely cite a Bank of America sentiment product (speakers include Andreas Steno, Sonu Varghese, Michael Howell, Jesse Felder, Jim Bianco, Brent Johnson, Darius Dale and Josh Brown); 2 carry quarantine_class: NOT_VISSER and are excluded on the store's own flag — both are bearish bank claims dated 2023 that carried skill: 2.0 under a misattributed speaker and would have dominated the lane; 1 is an off-subject entity tag (Li Lu's 13F). Of the 17 retained, Arthur Hayes supplies 6 — 35% of the lane — and reverses across them, from "losses don't exceed capital" (2023-04-03) to "functionally insolvent" (2023-11-01) to "banks are healthy" (2026-04-24), with one claim duplicated three weeks apart (2025-06-11 and 2025-07-02, collapsed to one). The five bearish 2023-2024 solvency claims (Gromen, Hayes, Brunell) are reported despite being falsified by events, because the record includes what did not work. The single claim of genuine value is compound_and_friends 2026-01-20 on declining net charge-offs, independently confirmed by the 10-Q filed six months later. Breadth after collapsing by speaker: 9. All quotes are verbatim from the stored claim text.
• Data as-of: annual fundamentals to 2025-12-31; quarterly fundamentals and all capital data to 2026-06-30; most recent filing-verified financials 2026-06-30, from the 10-Q filed 2026-07-31 · estimates 2026-08-04 · prices 2026-08-04, quote timestamp 1785873598 = 2026-08-04T19:59:58Z ($62.915, +0.70%; 50-DMA $57.70; 200-DMA $53.68; RSI 58.9; MACD +1.30) · knowledge-base claims 2026-08-04. All figures come from the Synthos vendor data file for BAC or from the SEC filings in the BAC archive; no figure comes from memory, recall or external retrieval.
• Filing archive contents: 10-K filed 2026-02-25 (fiscal 2025); 10-Q filed 2026-05-01 (March 2026 quarter); 10-Q filed 2026-07-31 (June 2026 quarter — the primary source for this dive); 8-K filed 2026-05-06; 8-K filed 2026-07-14 (second-quarter results); 8-K filed 2026-07-24. The 8-Ks are cover documents; the earnings releases sit in unextracted Exhibit 99 files and contain no numbers in this archive. Numeric tables are stripped from all extracted filings, so the balance sheet, the regulatory capital tables, loan and deposit totals, the allowance for credit losses and the efficiency ratio are unavailable except where the MD&A states them in prose; every figure used is either prose-stated in the filing or explicitly labelled as derived.
• Where the filings contradicted or corrected the vendor (detailed in Section 8): the vendor's revenue of $191.567B is GROSS of interest expense, against the meaningful $113.097B net figure the 10-K's own geographic disclosure is struck on, invalidating every vendor-computed margin; the inc_q revenue series switches basis at Q1 2026 and is rejected as a series in favour of earn_cal; seg_geo FY2025 double-counts $15.410B by listing "Non-US" alongside its own three components; seg_prod changes taxonomy between FY2024 (gross revenue lines, $192.434B) and FY2025 (segment net revenue, $116.760B); enterprise value ($1,149.922B), net debt, EV/EBITDA (25.19x), net-debt/EBITDA (15.41x), current ratio (0.073), working capital (−$363.577B), interest coverage (0.592x) and free-cash-flow yield (21.2%) are all rejected as category errors for a deposit-funded bank; the entire est EBITDA/EBIT block is rejected because netIncomeAvg exceeds ebitdaAvg in every forward year; est.netIncomeAvg implies a share count ~6% too high and is not used; the FY2029 est row is rejected as incoherent (revenue falls, EPS rises 16.3%, 2 analysts); bookValuePerShareTTM and tangibleBookValuePerShareTTM include preferred stock, with tangible COMMON book derived at ≈$28.2 (cross-checked against the filing's $201.6B of CET1 capital, which gives $28.41) so that price/tangible book is ≈2.24x rather than 1.94x; lastDividend of $1.12 is the TTM total against a $1.28 forward run-rate confirmed by the 10-Q's 2026-07-23 declaration of $0.32 per quarter, "an increase of 14 percent"; the FY2024-to-FY2025 bal_a debt comparison is broken ($375.149B to $48.088B of short-term debt); the vendor composite rating of C+ / 2 is rejected as an artefact of the same industrial template; quote.yearHigh/yearLow ($63.54/$44.75) are rejected in favour of tech.hi52/lo52 ($62.915/$44.92); and earn_cal's Q2 2025 EPS of $0.89 is superseded by the filing's $0.90. Where vendor and filing AGREED — worth recording: the earn_cal revenue series being on the correct net-of-interest-expense basis and reconciling to the 10-Q; the FY2025 and FY2024 seg_geo totals matching derived net revenue exactly in both years; consensus FY2026 revenue annualising within 0.1% of first-half actuals; the four-segment net income summing to the reported total with a coherent All Other residual; the dividend payout ratio; the CEO and employee count.
• Basis note: analyst EPS estimates and reported EPS for a bank are both essentially GAAP and are directly comparable here — unlike most names in this batch, no GAAP-to-non-GAAP bridge is required. The trailing figure used throughout is $4.36, the sum of the four reported quarterly diluted EPS figures net of preferred dividends, in preference to the vendor's netIncomePerShareTTM of $4.706 which does not deduct preferred.
• Estimate coverage: 9 analysts on FY2026 and FY2027 revenue, 15 on FY2026 and FY2027 EPS, 7 on FY2028 EPS, and 2 on FY2029 — the FY2029 row is rejected outright. Only FY2027 is used as a valuation anchor; FY2028 is used as a cross-check.
• Peer note: the vendor peer set is genuinely relevant — Citigroup ($234.6B), Goldman Sachs ($310.8B), Wells Fargo ($267.1B), Royal Bank of Canada ($289.8B), HSBC ($364.5B), Mitsubishi UFJ ($246.0B), Bank of Montreal ($126.0B). BAC at $446.5B is the largest. No peer multiples are supplied in the file, so no peer-multiple comparison is drawn — a real limitation on a name where relative price-to-tangible-book is the standard method, stated rather than filled from recall.
• Fair-value caveat: the $49 / $65 / $75 anchors are multiples of the FY2027 consensus EPS distribution — 9.5x the low of $5.112, 12.4x the mean of $5.242, and 13.9x the high of $5.381 — cross-checked against a price-to-tangible-common-book frame on a derived ≈$28.2 of current tangible book. Stated arithmetic, not a discounted cash flow. The base is sensitivity-disclosed: 12x gives $62.9, exactly spot; 13x gives $68.1. The whole answer lives in a 12–13x band. The Gordon-growth cross-check — justified P/TBV = (ROTCE − g)/(COE − g) = (15.9% − 4%)/(10% − 4%) = 1.98x, or approximately $56 on current tangible book and $65 on forward tangible book — is shown so a reader can substitute their own cost-of-equity assumption.
• Timing: second-quarter results were released 2026-07-14, twenty-one days before this dive, and the 10-Q was filed 2026-07-31, four days before it, so the dive is fully informed by the most recent disclosure. The next print is 2026-10-14, 71 days away — no near-term company-specific catalyst, which is an explicit input to the Hold verdict. The dated event that matters is regulatory: 2027-01-01, when the filing states the G-SIB surcharge is expected to increase.
• 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.