Bank of New York Mellon BNY

Reference table

1. What the business is

BNY is a Delaware corporation, NYSE-listed, with approximately 48,100 employees, led by Chairman and Chief Executive Robin Vince and Chief Financial Officer Dermot McDonogh. It is the world's largest custodian by assets under custody and administration.

Three reportable segments plus Other, from the second-quarter earnings release (8-K filed 2026-07-15, $M):

Read that and three things follow.

Every segment expanded its margin, and the two big ones did so on double-digit revenue growth. Securities Services added 400 basis points; Market and Wealth Services 360; Investment and Wealth Management 260. This is what operating leverage on a fixed platform looks like when volumes rise, and it is the entire investment case.

The margin dispersion is extreme and it names the weak link. Market and Wealth Services earns 52.0% pre-tax; Securities Services 39.3%; Investment and Wealth Management 21.1%. The asset-management business is 15.1% of revenue and 8.0% of pre-tax income, and the release attributes its AUM growth to "higher market values and improved seed capital results, partially offset by the mix of AUM flows" — a formulation that, together with the consolidated note on "cumulative net outflows", means the segment is still losing assets on a net basis and being rescued by markets. The other two segments together produce 92% of pre-tax earnings.

The fastest-growing lines are the plumbing, not the asset gathering. Issuer Services revenue +17%, Payments and Trade +17%, Clearance and Collateral Management +16%, securities-lending revenue +39% on securities on loan of $645 billion (+25%). These are volume businesses tied to the functioning of the capital markets, and they are running hot.

Scale and the balance sheet

Two observations that cut in opposite directions.

Tangible book value per share compounded 11.0% year on year while the share count fell 3.8% and 87% of earnings were paid out. That is a very high-quality combination: the company grew intrinsic value per share while distributing almost everything it earned.

But the Tier 1 leverage ratio has fallen in each of the last three periods — 6.2% to 6.0% to 5.9% — because total assets grew 11.2% in six months. The 10-Q states plainly: "The Tier 1 leverage ratio was 5.9% at June 30, 2026 and 6.0% at Dec. 31, 2025. The decrease primarily reflects higher average assets." For a custodian, the leverage ratio rather than the risk-based CET1 ratio is usually the binding capital constraint, because custody assets carry low risk weights but full leverage weight. At 11.0% CET1 and 5.9% leverage, the second number is the one to watch, and it is the one moving.

The 8-Ks — checked, and there is no acquisition

The brief flagged BNY as M&A-active. The check was run and the answer is no transaction in this window. The three 8-Ks in the archive are:

• 2026-06-24 (Item 8.01): BNY "announced its intention to increase its quarterly cash dividend on its common shares by 19% from $0.53 to $0.63 per share, commencing as early as the third quarter of 2026, subject to approval by the Company's Board of Directors." The same filing states the Stress Capital Buffer "has consistently remained at the 2.5% floor since the introduction of the SCB requirement in 2020" and confirms the existing repurchase authorisation approved in April 2026.
• 2026-07-15: second-quarter results, carrying the complete earnings release used throughout this dive.
• 2026-07-23 (Items 3.03, 5.03, 8.01): issuance of Series N Noncumulative Perpetual Preferred Stock, liquidation preference $100,000 per share, with the standard restriction on common dividends and repurchases should Series N dividends go unpaid. This is a preferred issuance, not an acquisition, and it postdates the June 30 balance sheet — so the preferred balance in every statement in this file is understated by the Series N amount, which the 8-K text in the archive does not size.

No merger, acquisition, divestiture or reverse Morris Trust appears anywhere in the archive. The 10-Q's only acquisition references are to a legacy German "cum/ex" matter concerning "pre-acquisition activity" in entities BNY bought years ago, for which "substantially all of the secondary liability notices have been resolved" and "we obtained an indemnity for liabilities from the sellers." We record it and do not weight it.

2. The revenue line, and a 101% gross-up

This is the largest vendor revenue gross-up this programme has documented, and it is essential to understand before any multiple is computed.

BNY publishes "total revenue" as total fee and other revenue plus NET interest income. The vendor's inc_a reports gross interest income plus fee revenue, before deducting interest expense.

The derivation of the FY2025 figure, and the honest caveat attached to it. BNY's 10-K does not contain an income statement. The annual report on Form 10-K filed 2026-02-25 incorporates the MD&A and financial statements by reference from the Annual Report (Exhibit 13), which is not in this archive — the extracted 10-K text contains 24 tables, none of them a statement of income. So the exact FY2025 published revenue figure is not obtainable from a filing here, and this dive does not pretend otherwise. What is obtainable: the 10-Q of 2026-07-31 gives H1 2025 total revenue of $9,820M, and the vendor's own earn_cal block — which, unlike inc_a, reports on the published net basis — gives Q3 2025 revenue of $5,069M and Q4 2025 of $5,179M. Those sum to $20,068M, which sits 0.2% from the FY2025 consensus est.revenueAvg of $20,024.8M. We use $20,068M and label it derived.

Note the internal contradiction this exposes. Within a single vendor file, inc_a reports FY2025 revenue of $40,441M, earn_cal reports the same year's quarters on a basis summing to $20,068M, and est.revenueAvg reports $20,024.8M. The file disagrees with itself by 101%. Every analyst estimate in the payload is struck on the published net basis; every historical revenue figure in inc_a is struck on the gross basis; and the two are used side by side by anyone who does not check.

inc_q is worse: it changes basis mid-series.

The June 2026 quarter is reported on the NET basis with an interest expense of zero, while every prior quarter in the same block is on the gross basis. A reader computing quarter-on-quarter revenue growth from inc_q would find revenue falling 42% in a quarter it actually rose 5%.

Consequential rejections. priceToSalesRatioTTM of 2.99x (correct on the published basis: approximately 5.0x); evToSalesTTM of 3.76x; netProfitMarginTTM of 17.50% (correct: approximately 35%); ebitdaMarginTTM of 26.08%; grossProfitMarginTTM of 59.06%; revenuePerShareTTM of $52.07. Every revenue and margin figure in this dive uses the published total-revenue basis, from the filings.

3. The June quarter, in the company's own numbers

Four observations.

First, the composition of the 13% is not evenly good. Fee revenue — the durable, capital-light part — grew 11%, "primarily reflecting net new business, higher market values and higher client activity, partially offset by the mix of AUM flows." Net interest income grew 20%, and its stated driver is "the reinvestment of investment securities at higher yields and balance sheet growth, partially offset by deposit margin compression." A securities-reinvestment cycle is a catch-up, not a franchise improvement: the portfolio rolls into higher-yielding paper once and then stops. And the offsetting term — deposit-margin compression — is already disclosed and already running. The 11% fee line is what should be capitalised at a high multiple. The 20% NII line should not be.

Second, the expense discipline is the real achievement. Noninterest expense grew 7% "primarily reflecting higher revenue-related expenses, investments and employee salary increases, partially offset by efficiency savings." Revenue-related expenses grow with revenue by definition, so a 7% total against 13% revenue means the non-variable base grew slower still. 419 basis points of operating leverage, and 401 on an adjusted basis, is genuinely excellent and is the reason the pre-tax margin moved 320 basis points in a year.

Third, credit is not a variable here. A benefit of $8 million "primarily reflecting improvements in commercial real estate exposure." Average loans are $85.6 billion against $525.0 billion of total assets and $370.5 billion of deposits — BNY lends 16% of its balance sheet, and most of that against collateral to its own custody clients. This is the structural reason a custodian trades at a higher multiple of book than a commercial bank, and it is entirely legitimate.

Fourth, the beat record is consistent and large:

Five consecutive beats averaging 10.1%, on a line that has risen sequentially every quarter. That pattern is directly relevant to Section 6, because the FY2026 consensus embedded in today's price implies it stops immediately.

4. Capital return, and the arithmetic of an 87% payout

From the earnings release and the 10-Q:

Three things this makes true.

The buyback is large and is doing visible work. Treasury stock acquired of $2,086M in the first half, against $1,641M a year earlier, took the share count from 705,241 thousand to 678,504 thousand — a 3.8% reduction year on year. At the current pace, roughly 3.9% of the market capitalisation annually.

The dividend increase is the largest single forward-looking disclosure in the archive. From the 8-K of 2026-06-24: "an intention to increase its quarterly cash dividend on its common shares by 19% from $0.53 to $0.63 per share, commencing as early as the third quarter of 2026, subject to approval by the Company's Board of Directors." At $0.63 quarterly the annualised rate is $2.52, a 1.61% yield on $156.80. This is management guidance and is half-weighted under house rules; it is also board-conditional and explicitly stated as such. (The vendor's dividendPerShareTTM of $2.22 and dividendYieldTTM of 1.42% are trailing figures and are labelled as such.)

Total shareholder yield is therefore approximately 3.9% buyback plus 1.6% forward dividend ≈ 5.5%, consistent with the disclosed 87% total payout ratio. At 15.4x FY2027 consensus that is a material component of the return — but it is a smaller component than at COF, because BNY's earnings multiple is 66% higher.

A capital caveat that belongs here rather than in the risk section. The 87% payout is only possible because the Stress Capital Buffer sits at the 2.5% floor and CET1 is 11.0%. But the Tier 1 leverage ratio has fallen from 6.2% to 6.0% to 5.9% across three periods while total assets grew 11.2% in a half-year. If deposit growth continues at 11.6% a half-year, the leverage ratio constrains the payout before the risk-based ratio does. This is the mechanism by which a good problem — clients depositing more money — becomes a capital problem for a custodian, and it is worth watching precisely because it is not a credit event.

Cash flow is not a useful lens for a custodian and is not used as one. The vendor reports FY2022 operating cash flow of $15,068M, FY2023 $5,912M, FY2024 $687M and FY2025 $6,730M, with FY2024 free cash flow of −$782M. Those swings are changes in trading assets, receivables and payables, not in earning power. freeCashFlowYieldTTM of 2.51%, priceToFreeCashFlowRatioTTM of 39.9x, evToFreeCashFlowTTM of 50.2x and incomeQualityTTM of 0.564 are all rejected. The measures used are ROTCE, tangible book value per share, the pre-tax operating margin, operating leverage and the capital ratios.

5. Where the multiple actually comes from

This section is short because the argument is one equation.

A bank's price-to-tangible-book multiple is, to a first approximation, its return on tangible common equity divided by the earnings multiple the market will pay. BNY earns 31.3% on tangible common equity and trades at 4.78x tangible book value of $32.81. 31.3% ÷ 4.78 = 6.5%, an earnings yield equivalent to 15.3x — which is within a rounding error of the 15.4x FY2027 consensus multiple. The valuation is internally consistent, and it is entirely a bet on the durability of the 31.3%.

What that sensitivity looks like, holding the earnings multiple at 15.4x:

A 20% ROTCE — which is roughly what this company earned before the current rate and volume cycle — implies $101, which is within a dollar of the 52-week low of $100.66. That is not a forecast; it is the observation that the entire twelve-month 57% move is the market re-rating the durability of a return on equity, and the stock has already traded at both ends of the plausible range within the year.

The three things that determine whether 31.3% holds are named in the filings, not invented here: net interest income (+20%, from a securities-reinvestment cycle against disclosed deposit-margin compression), market values (AUC/A +12% "primarily reflecting higher market values and net client inflows"), and operating leverage (+419 bps, from "efficiency savings" and the platforms operating model). The first is finite by construction, the second is beyond management's control, and the third is the only one that is genuinely a management achievement.

6. Valuation — priced in or room?

At $156.80 (678,504,000 shares outstanding at 2026-06-30; true market capitalisation ≈ $106.39B, not the vendor's $107.62B — see Section 8):

Estimate coverage is adequate on FY2026 and FY2027 and collapses beyond. 9 analysts on FY2026 EPS, 10 on FY2027 — the anchor — and only 3 on FY2028. The FY2027 EPS range is $9.728 to $10.477, a 7.7% spread. The FY2029 row is not thin, it is CORRUPT: epsAvg, epsHigh, epsLow and netIncomeAvg are all exactly ZERO on one analyst, while revenueAvg of $24,476.2M, ebitdaAvg and ebitAvg are populated. It is excluded entirely and flagged in Section 8.

est.ebitdaAvg and est.ebitAvg are REJECTED on two independent grounds. First, the fixed-ratio fabrication signature: ebitdaAvg is exactly 25.45% of revenueAvg and ebitAvg exactly 18.36% in every forward year, FY2026 through FY2029. Second, arithmetic impossibility: netIncomeAvg EXCEEDS ebitAvg in every forward year — FY2027 $7,119.5M against $4,245.4M, FY2028 $7,995.8M against $4,366.5M. And the level is wrong by more than half: an 18.36% implied EBIT margin against a reported June-quarter pre-tax operating margin of 39.8%. All forward valuation uses epsAvg.

Enterprise value is rejected as a concept and the field is internally contradictory. enterpriseValueTTM of $135,325M exceeds market capitalisation by $27,701M, implying net debt of that amount. The same file's bal_a reports netDebt of −$97,641M — that is, $97.6 billion of net CASH. The two fields disagree by $125.3 billion. For a bank holding $370.5 billion of customer deposits, neither figure means anything: deposits are the raw material, not leverage. netDebtToEBITDATTM of 2.95x (computed against a negative net-debt figure), evToSalesTTM, evToEBITDATTM of 14.42x, interestCoverageRatioTTM of 0.540x, currentRatioTTM of 40.6x and daysOfSalesOutstandingTTM of 329 days are all discarded as a class.

Peer context is partially usable, which is unusual. The vendor peer set is AMG, AMP, ARCC, BEN, BLK, BX, IVZ, JHG, MAIN, NTRS, PFG, SEIC, STT and TROW. Two of the fourteen are genuine peers — State Street ($51.2B) and Northern Trust ($33.4B), the other two US trust banks — and their presence is worth noting. Nine are traditional or alternative asset managers, and two (ARCC, MAIN) are business development companies, which are not comparable to anything here. We draw no peer-multiple comparison, because the payload contains no multiples for the peers — only their market capitalisations — and inventing them is not permitted.

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

At $156.80 — 17.1x FY2026 consensus, 15.4x FY2027, 4.78x tangible book — the price embeds:

• A return on tangible common equity that stays near 31%. (Our derivation; the release supplies the 31.3%.) This is the most fragile assumption in the price, and Section 5 sizes it: holding the earnings multiple constant, a drift to 25% implies $126 and to 20% implies $101, which is the 52-week low. Nothing else in this file has a comparable coefficient.
• EPS reaches $9.194 in FY2026 and $10.187 in FY2027. (Consensus; 9 and 10 analysts.) The FY2026 number is demonstrably conservative: first-half adjusted EPS was $4.71 ($2.25 + $2.46), so the consensus requires second-half EPS of $4.48 — BELOW the first half — on a line that has risen sequentially for five consecutive quarters. Simply repeating the June quarter twice gives $9.63, 4.7% above consensus. The September consensus of $2.25 against a June actual of $2.46 is the same point at quarterly resolution.
• Revenue growth decelerates from 10.4% to 4.4% to 2.8% while EPS grows 10.8% then 11.2%. (Consensus; 9, 9 and 5 analysts on revenue.) That gap has to be closed by expense discipline and buyback: roughly 4% a year from the share count and the rest from operating leverage. The June quarter delivered 419 basis points of leverage, so this is achievable — but it means the consensus earnings path does NOT depend on revenue and therefore does not depend on markets going up. That is a more robust assumption than it first appears.
• The market keeps paying roughly 15-17x forward earnings and 4.8x tangible book. (Our number.) At 13x FY2027E the stock is $132; at 18x it is $183. Custody banks have historically been valued on tangible book because their earnings are volume-driven, which is why the tangible-book cross-check governs the bear case here.
• Net interest income does not reverse. (Our number, from the release's own attribution.) NII grew 20% on securities reinvestment against disclosed deposit-margin compression, and average deposits FELL 1% sequentially to $314,036M. NII is roughly 25% of total revenue and close to 100% incremental margin; a flat NII line rather than a growing one removes about 5 points of revenue growth and rather more of earnings growth.

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

Expected return over the next twelve months decomposes as: EPS growth (+10.8%, from FY2026E $9.194 to FY2027E $10.187) + multiple drift (COMPRESSION of roughly 6%, from today's 17.1x on the forward year toward about 16.0x) + shareholder yield (+5.5%) ≈ +10%.

Our base assumes mild multiple compression, and the reason is the tangible-book cross-check rather than any view on the earnings. At $172 the stock would be 5.24x the current tangible book value of $32.81 — and tangible book itself is compounding at 11% a year, so by mid-2027 it would be roughly 4.7x the then-current figure. In other words the base case holds the price-to-tangible-book multiple roughly flat and lets book value growth plus earnings growth do the work, which is the correct way to value a custodian and is deliberately more conservative than holding the P/E flat.

Note what this makes the return: roughly half earnings growth and half shareholder yield, with the multiple a small negative. That is a respectable decomposition. It is not, at 9.7%, a compelling one — and the difference between this and the COF case in the same batch is instructive: there, the same shape produced a 14.9% base with 2.46:1 asymmetry at 9.3x earnings; here it produces 9.7% with 1.67:1 at 15.4x.

If the multiple held at 17.1x FY2027E the price would be $174 (+11.0%). If it compressed to 13x, $132 (−15.8%).

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

• On the level we have almost NO variant perception and we say so. Our base of $172 is 2.1% above the street's $168.50 and 0.3% below the median target of $172.50, with 10 of 22 analysts at buy and none at sell. We are not seeing anything the market is missing about this company's earnings power.
• Where we do differ is on the near-term estimate, and the difference is arithmetic rather than interpretive. The FY2026 consensus of $9.194 requires second-half EPS of $4.48 against a first half of $4.71, and the September consensus of $2.25 requires a sequential decline from $2.46 — on a line that has risen sequentially in each of the last five quarters and beaten consensus by an average of 10.1% over the same period. Watchable number: the 2026-10-15 print against $2.25. A print at or above $2.46 puts FY2026 on track for roughly $9.63 and drags FY2027 with it — but note carefully that this is a point about consensus being set low, not about the business accelerating, and a low bar cleared is worth less to a stock at 15.4x than to one at 9x.
• We think the market is capitalising a 20% net interest income growth rate that the company has itself explained away. The release attributes it to "the reinvestment of investment securities at higher yields and balance sheet growth, partially offset by deposit margin compression." A reinvestment catch-up ends; the compression does not. The 11% fee-revenue growth deserves a high multiple and the 20% NII growth does not, and the blended 13% is being valued as though it were uniform.
• We think the Tier 1 leverage ratio is the under-watched constraint. It has fallen 6.2% → 6.0% → 5.9% across three periods as total assets grew 11.2% in six months and 87% of earnings were paid out. Watchable number: the Tier 1 leverage ratio in the October print. Below roughly 5.5% and the buyback — which is 3.9% of the market capitalisation and roughly 40% of the expected return — comes under pressure, without any credit or earnings event.
• Positive surprise that would force a re-rate: a September quarter above $2.46 with operating leverage still positive and the Tier 1 leverage ratio stable, which would make the FY2027 consensus of $10.187 look conservative and support $180+; or a first disclosure of digital-asset and tokenisation custody economics, which is the one place a custodian's growth rate could step-change and is currently a paragraph in a forward-looking-statement list.
• Negative surprise that would break the thesis: net interest income flat or down sequentially, confirming the reinvestment cycle has run; operating leverage turning negative; the Tier 1 leverage ratio below 5.5% with the buyback slowed; a market drawdown taking AUC/A down (12% of the current growth is attributed partly to "higher market values"); or the board declining to ratify the announced 19% dividend increase, which the 8-K explicitly makes conditional.

Synthos fair values

All three anchors are multiples of the FY2027 consensus EPS distribution (mean $10.187, low $9.728, high $10.477, 10 analysts), each cross-checked against the filing's tangible book value of $32.81 per share at 2026-06-30.

• Bear ~$128 — 13.2x the FY2027 consensus LOW of $9.728, and 3.90x current tangible book. Cross-check: 27.2% above the 52-week low of $100.66; 1.0% above the 200-day average of $126.82. The scenario: the securities-reinvestment cycle completes, net interest income flattens, markets stop rising so AUC/A growth stalls, ROTCE drifts to 25%, and the market pays a normal trust-bank multiple of tangible book. −18.4%.
• Base ~$172 — 16.9x the FY2027 consensus MEAN of $10.187, and 5.24x current tangible book (roughly 4.7x the tangible book implied by another year of 11% growth). Cross-check: 2.1% above the street's $168.50; 15.2x FY2028E. Sensitivity, stated openly: 15x FY2027E gives $153 — below spot — and 18x gives $183. The whole answer lives in a 15-18x band on one estimate row. The scenario: operating leverage persists, the buyback continues at 4% a year, tangible book compounds at 10-11%, ROTCE holds near 30%, and the multiple stays roughly where it is on a growing book. +9.7%.
• Bull ~$205 — 19.6x the FY2027 consensus HIGH of $10.477, and 6.25x current tangible book. Cross-check: 13.9% above the 52-week high of $162.35; 13.9% above the highest street target of $180. The scenario: the September quarter beats decisively, FY2026 lands near $9.63 and FY2027 near $10.80, digital-asset custody acquires disclosed economics, and the market accepts that a platform monopoly earning 31% on tangible equity deserves a growth multiple. +30.7%.

Base is 9.7% above spot; asymmetry roughly 1.67:1 to the upside (18.4% down, 30.7% up), before a shareholder yield of approximately 5.5%. That combination — a base under 10%, a payoff ratio under 2:1, a base within 2% of the street's, and an entry 23.6% above the 200-day average at 4.78x tangible book — is a Watch, not a Buy. What would make it a Buy is set out in Section 11 and is mostly a price.

7. Knowledge base — eight hits, one survivor, and it is about crypto

Raw hits: 8. After a case-sensitive entity re-run: 1. Entity matches on BNY: 1. Name-level claims: 1, and it is BEARISH. Discarded: 7.

The primary sweep ran entity terms BNY, BNY Mellon, Bank of New York Mellon and BK, plus the custody-bank peers State Street, STT and Northern Trust, with free text on BNY, Bank of New York, custody bank and custodian bank, across all 51,928 distilled claims. It returned 8 hits across 7 channels — one of the thinnest raw lanes in this batch, which for a $106 billion systemically important bank is itself a finding.

The single case-sensitive survivor, verbatim:

> 2024-06-18 · BEARISH · conviction 65 · horizon: thesis · entities: Coinbase, BNY Mellon · channel: raoul_pal · no named speaker · skill 1.0

> "Biggest 2-3 year crypto risk: a major custodian holding concentrated ETF assets (Coinbase, TradFi banks) gets hacked — no bailout possible in crypto."

What this is and is not. It is a claim about operational and concentration risk in digital-asset custody, naming BNY Mellon as one instance of "TradFi banks" alongside Coinbase. It is not a view on BNY's securities-servicing franchise, its earnings, its capital or its valuation. It is twenty-six months old, it names a two-to-three-year window which is now largely elapsed, and the event it describes has not occurred. We record it as the only name-level claim in the store, we grade it as unresolved-trending-wrong, and we weight it at a conviction of 65 on a lane of one — which is to say, barely at all.

The seven discards, and they split cleanly into two groups.

Five are peer-term collisions. The entity term "State Street" pulled claims about the passive-investing complex that name BlackRock, Vanguard and State Street and have nothing to do with BNY: mike_green 2022-05-30 (bearish 90, "Passive index funds are not truly passive — they transact daily on inflows, violating Sharpe's definition"), natalie_brunell 2026-04-14 (bearish 80, on a "financial-industrial complex"), compound_and_friends 2025-01-25 (bearish 60, on ESG and fiduciary duty), invest_like_the_best 2021-05-12 (neutral 70) and real_vision 2026-01-27 (bullish 78, on Galaxy tokenising a State Street money-market fund on Solana). All discarded: none names BNY and none concerns custody economics.

Two are free-text mentions of BNY inside crypto arguments:

> 2022-12-12 · bullish · conviction 72 · horizon: thesis · entities: FTX · channel: arthur_hayes · no named speaker · skill 1.2

> "FTX collapse is a cleansing forest fire; proof-of-reserves, custodial solutions, and institutions (BNY, Fidelity) entering will heal the market."

> 2023-04-03 · bearish · conviction 75 · horizon: principle · entities: Federal Reserve · channel: arthur_hayes · no named speaker · skill 1.2

> "The Fed quietly blocks full-reserve/narrow banks (TNB, Custodia) by endless delay — the fractional-reserve system requires banks that lend, not safe custody banks."

The first names BNY as a beneficiary of institutional crypto adoption; the second is a claim about Federal Reserve policy toward narrow banking that touches BNY's business model only obliquely. Neither is entity-tagged to BNY and both are discarded from the conviction pool.

The pattern is worth stating plainly, because it is the most interesting thing this lane contains. Every single knowledge-base mention of BNY in 51,928 distilled claims concerns digital-asset custody. Not one concerns the $62.6 trillion securities-servicing franchise that produces essentially all of the company's earnings. The credible-thinker universe in this store discusses BNY exclusively as a piece of crypto infrastructure, and BNY's own filings in this archive give that business no revenue line at all. That is a complete disconnect between what the outside voices find interesting and what the company actually earns, and it means the knowledge base contributes nothing to the valuation question here.

Attribution note. All 8 hits carry speaker: null and are attributed only to a channel — the weakest sourcing the four-lane policy admits. No claim carries a management speaker_role. No concentration sensitivity test is meaningful across eight claims from seven channels, though arthur_hayes supplies two of the eight.

Conclusion. Breadth 1, claim count 1, net conviction negative-low. The Synthos knowledge base's only view on BNY is a two-year-old warning about custodian hacking risk in crypto ETFs, which has not materialised and which addresses a business line the filings do not size. On the question this dive is actually asking — whether a 31.3% return on tangible common equity is durable — the store is silent.

8. Data integrity — what we rejected and why

Eight findings. BNY's vendor file contains the largest revenue gross-up and the largest internal contradiction this programme has documented, plus a corrupt estimate row, a mid-series basis change, a stale share count in the market-capitalisation field, and a segment-geography block that double-counts its own subtotal.

1. revenue is a 101% gross-up the company does not publish — REJECTED and rebuilt, and the file disagrees with itself by more than 2:1. The vendor reports FY2025 revenue of $40,441M; the published total revenue is approximately $20,068M. The reconciliation runs through the vendor's own interestExpense of $15,818M... of $20,682M: $40,441M − $20,682M = $19,759M, within 1.5% of the derived published figure, the residual being noncontrolling interests in consolidated investment funds. The same file reports the same year on the published basis in two other places: earn_cal revenueActual sums to $20,068M and est.revenueAvg reads $20,024.8M. Consequential rejections: priceToSalesRatioTTM 2.99x (correct ≈ 5.0x), netProfitMarginTTM 17.50% (correct ≈ 35%), evToSalesTTM 3.76x, ebitdaMarginTTM 26.08%, grossProfitMarginTTM 59.06%, revenuePerShareTTM $52.07. FY2024's gross-up is larger still at +114.6%. This is the IBKR/COF/SCHW gross-up defect class at the largest magnitude yet recorded.

2. inc_q CHANGES BASIS between the March and June 2026 quarters. Every quarter from Q2 2024 through Q1 2026 is reported gross, with an interest-expense line of $4.4–5.6 billion. The June 2026 quarter reports revenue of $5,698M — the exact published figure — with interestExpense of ZERO. A quarter-on-quarter growth rate computed from this block shows revenue falling 42% in a quarter it actually rose 5%. The block is unusable as a series and is used in this dive only for net income and share counts, where it reconciles.

3. The balance block and the enterprise-value field disagree by $125.3 billion. bal_a at 2025-12-31 reports netDebt of −$97,641M — $97.6 billion of net CASH. km_ttm reports enterpriseValueTTM of $135,325M against a market capitalisation of $107,624M, implying $27,701M of net DEBT. The two cannot both be true and neither is meaningful for a bank holding $370.5 billion of customer deposits. netDebtToEBITDATTM of 2.95x is computed against the negative net-debt figure and is sign-incoherent. This is the ETN/ABT internal-contradiction defect class at roughly six times the largest previously recorded magnitude ($21B on ABT). All enterprise-value-derived metrics are discarded as a class, along with currentRatioTTM of 40.6x, interestCoverageRatioTTM of 0.540x, daysOfSalesOutstandingTTM of 329 days and assetTurnoverTTM of 0.069.

4. est FY2029 is a CORRUPT ROW — epsAvg, epsHigh, epsLow and netIncomeAvg are all exactly ZERO. The row carries revenueAvg of $24,476.2M, ebitdaAvg of $6,230.1M and ebitAvg of $4,493.6M on 2 revenue analysts and 1 EPS analyst, with every earnings field set to zero rather than null. A naive forward-P/E calculation on this row divides by zero; a naive growth calculation shows EPS falling 100%. The row is excluded entirely.

5. est.ebitdaAvg and est.ebitAvg carry BOTH the fixed-ratio fabrication signature AND arithmetic impossibility — REJECTED. ebitdaAvg is exactly 25.45% of revenueAvg and ebitAvg exactly 18.36% in every forward year FY2026–FY2029. netIncomeAvg exceeds ebitAvg in every forward year — FY2027 $7,119.5M against $4,245.4M, FY2028 $7,995.8M against $4,366.5M — which is impossible. And the level is wrong by more than half: an implied 18.36% EBIT margin against a reported June-quarter pre-tax operating margin of 39.8%. All forward valuation uses epsAvg.

6. marketCap is struck on a STALE share count — the vendor overstates market capitalisation by 1.16%. The vendor's $107,624,227,200 at $156.80 implies exactly 686,379,000 shares. The 10-Q reports 686,379 thousand shares outstanding at MARCH 31, 2026 and 678,504 thousand at JUNE 30, 2026 — the vendor is using the prior quarter's count. True market capitalisation is approximately $106,389M, $1,235M lower. The error is small but it flows into evToSalesTTM, enterpriseValueTTM and every per-share cross-check, and it is exactly the kind of error that is invisible without the filing. (All per-share figures in this dive use the filing's 678,504,000 June 30 count.)

7. bookValuePerShareTTM and tangibleBookValuePerShareTTM do not deduct preferred equity — overstated by 11.5% and 13.8%. The vendor reports $65.61 and $37.35. The 10-Q reports book value per common share of $58.82 and tangible book value per common share of $32.81 at 2026-06-30. The difference — roughly $4.66 billion, or $6.79 a share — is the preferred stock, which BNY has just added to via the Series N issuance of 2026-07-23. shareholdersEquityPerShareTTM of $64.62 sits alongside bookValuePerShareTTM of $65.61, confirming no preferred deduction is applied to either. We use the filing's $58.82 and $32.81 throughout. The difference matters: price-to-tangible-book is 4.78x on the filing and 4.20x on the vendor's figure — a 14% understatement of the multiple that carries the entire valuation argument.

8. seg_geo FY2025 contains a DUPLICATE SUBTOTAL that inflates the total by 35% if summed, and seg_prod covers a third of revenue and is missing FY2023 entirely. The FY2025 seg_geo entry reads: United States $12,996M, EMEA $4,565M, Asia Pacific $1,393M, Other Countries $1,126M — and "Non-US" $7,084M, which is exactly EMEA + Asia Pacific + Other Countries. Summing all five gives $27,164M against a published total near $20,068M. The FY2024 entry has no "Non-US" line and sums to $18,619M, consistent with the FY2024 consensus of $18,431.9M. On the correct basis the United States is 64.7% of FY2025 revenue. Separately, seg_prod FY2025 reports only three categories — Financial Service $10,097M, Investment Advisory/Management/Administrative $3,083M, Distribution and Shareholder Service $146M — summing to $13,326M, 33.6% short of published revenue, and the FY2023 entry is absent altogether (the series runs 2025, 2024, 2022, 2021...). Neither block corresponds to BNY's three reportable segments. All segment analysis in this dive comes from the earnings release and the 10-Q.

Where vendor and filing AGREED — recorded, because verified-clean is a real finding:

• tech.max_dd_from_peak of −3.41% is CORRECT here, and this was checked rather than assumed. It is identical to pct_from_hi, which is the signature of the defect found on ADP in this same batch. But ret_12m of +57.1% on a last price of $156.81 implies a price near $99.81 twelve months ago, which is within 1% of lo52 of $100.66 — so the low PRECEDED the high and a 3.4% maximum drawdown is arithmetically consistent. Verified, not flagged.
• earn_cal is on the published net revenue basis and reconciles to the filings exactly — Q2 2026 revenueActual of $5,698,000,000 matches the earnings release to the dollar, and epsActual of $2.46 matches the disclosed adjusted diluted EPS.
• capitalExpenditure of −$1,553M for FY2025 against D&A of $1,852M is plausible for a bank capitalising software and shows no sign of the COP-class understatement. The capex-versus-D&A sanity check was run and passed.
• effectiveTaxRateTTM of 20.45% is consistent with the release's disclosed June-quarter effective rate of 21.0%.
• bal_a FY2025 totalAssets of $472,300M matches the 10-Q's December 31, 2025 comparative exactly.

Vendor composite rating — accepted with heavy caveats. The rating block gives B / 3 overall, with 5/5 on discounted cash flow, 4/5 on return on equity, 2/5 on price-to-earnings and 1/5 on price-to-book. The 5/5 discounted-cash-flow score rests on the free-cash-flow series rejected in Section 4 — the same series that shows FY2024 free cash flow of MINUS $782 million — and should carry no weight whatever. The 1/5 price-to-book is computed on the preferred-inflated book value of finding 7 and understates the true multiple.

Non-equity tripwire — checked and passed, with a note. BNY is common stock, $0.01 par value, NYSE-listed. The same registrant also has two other NYSE-listed securities that are NOT this one: the 6.244% Fixed-to-Floating Rate Normal Preferred Capital Securities of Mellon Capital IV (BNY/P) and Depositary Shares representing Series K preferred (BNY PRK), plus the newly issued Series N preferred. Price of $156.80 is not par-like; beta is 1.058; the dividend is variable and about to rise 19%; volume was 1.73M shares (~$271M of turnover); the 52-week band of $100.66 to $162.35 is a 61% range. This is common equity.

9. Technicals

• Price $156.80. −3.4% from the 52-week high of $162.35; +55.8% above the 52-week low of $100.66. Position within the annual range: 91st percentile.
• Both moving averages are below the price and the gap to the longer one is large. +5.4% above a 50-day average of $148.73; +23.6% above a 200-day average of $126.82. The 200-day gap is the second-widest in this batch and is the technical restatement of the valuation argument.
• RSI 37.0 against a positive MACD of +2.34 — and this tension is worth naming. An RSI in the mid-thirties on a stock 3.4% below its 52-week high describes a sharp, recent, short pullback inside an intact medium-term uptrend. It is the one piece of evidence in this section arguing the entry is not at a local top. (The chief executive's 16 July transfer was struck at $161.075, 2.7% above today's price.)
• Maximum drawdown from peak over the trailing year: −3.41% — verified consistent with the 52-week range and the twelve-month return, because the low preceded the high. This is the shallowest twelve-month drawdown in this batch by a wide margin.
• Relative performance is exceptional across every window. 3-month +18.9% against SPY +7.6% and QQQ +7.7%; 6-month +28.9% against SPY +11.1% and QQQ +15.7%; 12-month +57.1% against SPY +24.3% and QQQ +30.8% — a 33-point outperformance against the index and 26 against the Nasdaq. This is a stock that has been right for a year, and the question this dive is asking is what is left.
• Sentiment: 10 buy, 12 hold, 0 sell across 22 analysts; consensus target $168.50 (+7.5%), median $172.50 (+10.0%), high $180 (+14.8%), low $147 (−6.2%). No sell ratings, but a majority at hold and a consensus target under 8% away — the street's collective position is "good company, fair price," which is where we land too.

Today's move and what it does to the entry

BNY closed 2026-08-04 at $156.80, up 0.38% or $0.60 from $156.20. It opened at $157.29, traded $156.71 to $158.96, and closed near the low of the day on 1.73M shares. No company filing is dated 2026-08-04; the last was the 10-Q of 2026-07-31, four days earlier, and the last insider filing was the same day.

The honest read: the entry is 23.6% above the 200-day average and at 4.78x tangible book, in a name that has outperformed the index by 33 points over twelve months. The RSI of 37.0 says the last few sessions have been weak, which is the only argument that this is not the worst possible moment. That is not enough to build a position on. The pre-registered upgrade levels in Section 11 are prices, not narratives, precisely because the business is not the question here.

10. Insiders — eight transactions, one economic disposal, routed through a gift

The reading, and it requires care because the form of these transactions obscures their substance.

The chief executive's two entries are a single transfer, not a trade. 24,833 shares moved from direct to indirect ownership on 16 July at a stated price of $161.075. Robin Vince's total holding is unchanged at 650,627 shares. No economic disposition occurred. (The $161.075 reference price is itself informative: it is 2.7% above today's close and within 0.8% of the 52-week high.)

The chief financial officer's six entries are one economic transaction and it IS a disposal. On a single day, 31,800 shares were gifted from direct to indirect ownership — and the entire 31,800 was then sold from the indirect account across four tranches at $153.78 to $156.68, ending at a holding of zero in that account. Total proceeds approximately $4.93 million.

Two honest readings and we give both. The generous one: a gift-then-sell structure through an indirect entity is the standard mechanic of a charitable contribution or a family-trust funding, the transaction was executed fourteen days after the earnings release inside a normal window, and Dermot McDonogh retains 248,727 shares directly — worth approximately $39 million — an unchanged direct position. The unforgiving one: whatever the structure, 31,800 shares of economic exposure left the chief financial officer's control at an average of roughly $155, two weeks after a quarter that beat by 10.3% and near a 52-week high. We report it as a disposal of $4.93 million by the CFO, at the top of the twelve-month range, structured as a gift, and we do not claim to know which reading is correct.

What the file does not contain: a single open-market purchase, by anyone, at any price. On a stock up 57% in twelve months, that absence is unsurprising and carries no signal.

11. Verdict, kill-criteria and flip conditions

Watch.

This is a very good company that has already been priced as one, and the discipline being applied here is the same one that produced a Buy on COF in this batch: the arithmetic decides.

What is genuinely excellent. The June quarter delivered total revenue +13% against noninterest expense +7%, 419 basis points of operating leverage, a 39.8% pre-tax operating margin against 36.6%, diluted EPS +27%, and a 31.3% return on tangible common equity. Every segment expanded its margin. The provision for credit losses was a benefit. AUC/A reached $62.6 trillion, +12%. Tangible book value per share compounded 11.0% while the share count fell 3.8% and 87% of earnings were returned. The Stress Capital Buffer has been at the 2.5% regulatory floor since 2020, and the dividend is set to rise 19%. Five consecutive quarters have beaten consensus by an average of 10.1%.

What stops it being a Buy, stated as four specific reservations.

First, the base case is 9.7% and the payoff ratio is 1.67:1. That does not clear the bar this programme applies. COF, in the same batch, offered 14.9% and 2.46:1 at 9.3x earnings with a 7.4% shareholder yield. BNY offers less of everything at 15.4x.

Second, we have essentially no variant perception on the level. Our $172 sits 2.1% above the street's $168.50 and 0.3% below the median target. The one thing we do see — that the FY2026 consensus implies second-half EPS below the first half on a line that has risen for five straight quarters — is a point about a low bar, not about a mispriced business, and a low bar cleared is worth less at 15.4x than at 9x.

Third, the multiple carries the entire risk and Section 5 sizes it. 4.78x tangible book is the 31.3% ROTCE capitalised at 15.4x. At a 25% ROTCE the same multiple implies $126; at 20% — roughly where this company earned before the current cycle — it implies $101, which is the 52-week low. And 20% of the current revenue growth is a securities-reinvestment catch-up that the company itself has told us is offset by deposit-margin compression.

Fourth, the entry is 23.6% above the 200-day average after a 57% twelve-month run, the knowledge base's only claim is bearish and about a different business line, and the chief financial officer disposed of $4.93 million of economic exposure two weeks after the print.

Pre-registered UPGRADE conditions — what would take this to Buy — Tactical. Two are prices and two are numbers:

• A price near $137, which is 13.5x the FY2027 consensus of $10.187 and roughly 4.2x current tangible book. At that level the base case is +25%, the asymmetry exceeds 2.5:1, and nothing has to change at the company. This is the most likely route and it requires a 13% pullback — smaller than the drawdown this stock experienced twice in the last two years.
• A September quarter at or above $2.46 adjusted EPS (consensus $2.25) with operating leverage still positive, which would put FY2026 near $9.63 and make the FY2027 consensus of $10.187 conservative — and would also demonstrate that the June result was not the peak.
• Net interest income growing sequentially for a third consecutive quarter with average deposits rising, which would mean the reinvestment cycle is being replaced by balance-sheet growth rather than running out.
• The Tier 1 leverage ratio stabilising at or above 6.0% while the buyback continues, which would remove the one capital constraint visibly tightening.

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

• Return on tangible common equity below 26% in any quarter, which at a constant earnings multiple implies roughly $131.
• Operating leverage turning negative — expense growth exceeding revenue growth — which is the single number that has driven the entire re-rating.
• Net interest income flat or falling sequentially with average deposits declining, confirming the reinvestment cycle has completed.
• The Tier 1 leverage ratio below 5.5% with the buyback slowed or suspended, which removes roughly 40% of the expected return without any earnings event.
• The board declining to ratify the announced 19% dividend increase, which the 8-K of 2026-06-24 explicitly makes conditional on board approval.
• AUC/A declining year on year, which would mean the market-value tailwind has reversed and would take fee revenue with it.

Where BNY fits in the Synthos Framework Portfolio. No position today. It goes on the watch list at $137 (13.5x FY2027 consensus, ~4.2x tangible book) for a 2% initial sleeve in the financials bucket, with a 3.5% target on either a second confirming quarter of positive operating leverage or a print above $2.46. Sizing note: the temptation on a name like this is to pay up because the quality is obvious. The quality IS obvious — that is precisely why 4.78x tangible book is the price, and why the only edge available is patience. Logged as a tracked Synthos call (Watch) as of 2026-08-04 at $156.80, with the fair-value anchors, the kill criteria and the $137 upgrade trigger all gradeable.

Single biggest risk: that a 31.3% return on tangible common equity is a cyclical peak. BNY has just reported the best profitability in its modern history, produced by three things — a securities-reinvestment cycle lifting net interest income 20%, market values lifting assets under custody 12%, and genuine expense discipline delivering 419 basis points of operating leverage. The first is finite by construction and the company has disclosed its offset. The second is beyond management's control. Only the third is a durable management achievement, and it is the smallest of the three. At 4.78x tangible book the price assumes all three persist. If the return drifts to 25% — not a failure, a normal cycle — the arithmetic implies $126, a 20% decline, with no operational disappointment whatsoever. That is why this is a Watch with a price trigger rather than a Buy with a hope.

Provenance & disclosures

• Traceability: 1 name-level knowledge-base claim names BNY out of 51,928 distilled claims, and it is BEARISH (raw hits 8, surviving a case-sensitive entity re-run 1, entity matches 1, name-level 1, discarded 7; breadth 1, net conviction negative-low). The claim: raoul_pal, 2024-06-18, bearish, conviction 65, entities Coinbase and BNY Mellon, no named speaker — "Biggest 2-3 year crypto risk: a major custodian holding concentrated ETF assets (Coinbase, TradFi banks) gets hacked — no bailout possible in crypto." It concerns digital-asset custody operational risk, is twenty-six months old, names BNY only as an instance of "TradFi banks," and the event has not occurred; we grade it unresolved-trending-wrong and weight it at a conviction of 65 on a lane of one. Five discards are peer-term collisions pulled by the entity "State Street" and naming BlackRock, Vanguard and State Street rather than BNY: mike_green 2022-05-30 (bearish 90, on passive indexation), natalie_brunell 2026-04-14 (bearish 80), compound_and_friends 2025-01-25 (bearish 60, ESG), invest_like_the_best 2021-05-12 (neutral 70) and real_vision 2026-01-27 (bullish 78, Galaxy tokenising a State Street fund). Two discards are free-text mentions of BNY inside crypto arguments: arthur_hayes 2022-12-12 (bullish 72, "institutions (BNY, Fidelity) entering will heal the market") and arthur_hayes 2023-04-03 (bearish 75, on the Federal Reserve blocking narrow banks). The structural finding: every knowledge-base mention of BNY in the entire store concerns digital-asset custody, and not one concerns the $62.6 trillion securities-servicing franchise that produces essentially all of the earnings — a complete disconnect between what the outside voices discuss and what the company earns. All 8 hits carry speaker: null and are channel-attributed. No claim carries a management speaker_role. No concentration sensitivity test is meaningful across eight claims from seven channels. All quotes are verbatim from the stored claim text.
• Data as-of: all operating figures — segment revenue and margins, net interest income, fee revenue, expenses, EPS, ROTCE, capital ratios, AUC/A, AUM, deposits, book and tangible book value per share and the share count — through 2026-06-30, from the 8-K earnings release filed 2026-07-15 and the 10-Q filed 2026-07-31 · annual data from the vendor payload, with the FY2025 published revenue figure derived as set out in Section 2 · estimates 2026-08-04 · prices 2026-08-04, quote timestamp 1785873599 = 2026-08-04T19:59:59Z ($156.80, +0.38%; 50-DMA $148.73; 200-DMA $126.82; RSI 37.0; MACD +2.34) · knowledge-base claims 2026-08-04. BNY's fiscal year is the calendar year. All figures come from the Synthos vendor data file for BNY or from the SEC filings in the BNY archive; no figure comes from memory, recall or external retrieval.
• Filing archive contents, and one significant gap: 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 source of the balance sheet, capital ratios, book values and share counts); 8-K filed 2026-06-24 (Item 8.01 — the 19% dividend-increase intention and the confirmation that the Stress Capital Buffer has remained at the 2.5% floor since 2020); 8-K filed 2026-07-15 (the complete second-quarter earnings release, the source of every segment and profitability figure in this dive); 8-K filed 2026-07-23 (Series N Noncumulative Perpetual Preferred Stock issuance). All carry preserved [TABLE] statement data. THE GAP: BNY's 10-K does not contain an income statement or MD&A — it incorporates them by reference from the Annual Report (Exhibit 13), which is not in the extracted text. So no full-year FY2025 published revenue or expense figure is available from a filing, and this dive does not estimate around it: the FY2025 revenue figure of ≈$20,068M is explicitly derived from the 10-Q's H1 2025 total of $9,820M plus the vendor earn_cal block's Q3 and Q4 published figures, labelled as derived at every use, and cross-checked against the FY2025 consensus of $20,024.8M (0.2% apart). The 8-K corporate-action scan was run and found NO acquisition, divestiture or merger in the archive window — the brief's flag that BNY is M&A-active is not borne out by any filing here.
• Where the filings contradicted or corrected the vendor (detailed in Section 8): the revenue basis — the vendor reports FY2025 revenue of $40,441M against a published figure of approximately $20,068M, a 101.5% gross-up (FY2024: +114.6%), reconciling through the vendor's own $20,682M interest-expense line, with the same file reporting the identical year on the published basis in both earn_cal (summing to $20,068M) and est.revenueAvg ($20,024.8M) — invalidating priceToSalesRatioTTM, netProfitMarginTTM, evToSalesTTM, ebitdaMarginTTM and revenuePerShareTTM; inc_q changing basis mid-series, reporting the June 2026 quarter net at $5,698M with zero interest expense while every prior quarter is gross at $9.9–10.4B, so that a naive sequential calculation shows a 42% revenue decline in a quarter revenue rose 5%; the $125.3 billion contradiction between bal_a's netDebt of −$97,641M (net cash) and km_ttm's enterpriseValueTTM of $135,325M (implying $27,701M of net debt), with netDebtToEBITDATTM of 2.95x computed against the negative figure; marketCap struck on a stale share count — the vendor's $107,624,227,200 implies 686,379,000 shares, exactly the 10-Q's March 31 figure, against 678,504,000 outstanding at June 30, overstating market capitalisation by 1.16% (we use the filing's June 30 count throughout); bookValuePerShareTTM of $65.61 and tangibleBookValuePerShareTTM of $37.35, neither of which deducts preferred equity, against the 10-Q's $58.82 and $32.81 — a 11.5% and 13.8% overstatement, and the difference between a 4.20x and a 4.78x price-to-tangible-book multiple, which is the multiple the entire valuation argument rests on; est FY2029, a corrupt row with epsAvg, epsHigh, epsLow and netIncomeAvg all exactly ZERO; est.ebitdaAvg/ebitAvg, rejected for a fixed-ratio signature (exactly 25.45% and 18.36% of revenue in every forward year), for netIncomeAvg exceeding ebitAvg in every forward year, and for implying an 18.36% EBIT margin against a reported 39.8% pre-tax operating margin; and seg_geo FY2025, which contains a "Non-US" line of $7,084M that exactly duplicates the sum of its own EMEA, Asia Pacific and Other Countries lines, inflating a naive total by 35%, alongside seg_prod, which covers only 66.4% of published revenue and is missing FY2023 entirely. Where vendor and filing AGREED — recorded: tech.max_dd_from_peak of −3.41% was CHECKED against the ADP-class defect and found CORRECT, because ret_12m of +57.1% implies a price near $99.81 a year ago, within 1% of lo52 of $100.66, so the low preceded the high; earn_cal is on the published net basis and matches the earnings release to the dollar ($5,698,000,000 revenue, $2.46 adjusted EPS for Q2 2026); capitalExpenditure of −$1,553M against D&A of $1,852M shows no COP-class understatement (the check was run); effectiveTaxRateTTM of 20.45% is consistent with the disclosed 21.0%; and bal_a FY2025 total assets of $472,300M matches the 10-Q comparative exactly.
• Basis note: analyst estimates are on adjusted diluted EPS, and BNY's adjustments are among the smallest in this programme — June-quarter GAAP diluted EPS of $2.45 against adjusted $2.46, a 0.4% difference, and adjusted noninterest expense of $3,431M against reported $3,439M. No conclusion here turns on the adjustment. Trailing adjusted EPS of $8.64 is the sum of the four quarterly earn_cal actuals ($1.91 + $2.02 + $2.25 + $2.46) and is labelled as such. "Total revenue" throughout means total fee and other revenue plus NET interest income, as BNY publishes it, never the vendor's gross-up. Return on tangible common equity, tangible book value per common share and adjusted figures are the company's own Non-GAAP measures, defined in the "Explanation of GAAP and Non-GAAP financial measures" section of the earnings release, and are used because they are the measures on which a custodian bank is actually valued.
• Estimate coverage: 9 analysts on FY2026 EPS, 10 on FY2027 — the anchor for all three fair values — with an FY2027 range of $9.728 to $10.477, a 7.7% spread; only 3 on FY2028; FY2029 excluded as corrupt. Revenue coverage: 9, 9, 5 and 2 analysts respectively.
• Peer note: the vendor peer set — AMG, AMP, ARCC, BEN, BLK, BX, IVZ, JHG, MAIN, NTRS, PFG, SEIC, STT and TROW — contains two genuine custody-bank peers (State Street, Northern Trust) among fourteen names, alongside nine asset managers and two business development companies. No peer-multiple comparison is drawn, because the payload supplies only peer market capitalisations and no peer multiples.
• Fair-value caveat: the $128 / $172 / $205 anchors are multiples of the FY2027 consensus EPS distribution — 13.2x the low of $9.728, 16.9x the mean of $10.187, and 19.6x the high of $10.477 — each cross-checked against the filing's tangible book value of $32.81 per share at 2026-06-30 (3.90x, 5.24x and 6.25x respectively). Stated arithmetic, not a discounted cash flow. The base is sensitivity-disclosed: 15x gives $153, below spot; 18x gives $183. The base case assumes the price-to-tangible-book multiple stays roughly flat while tangible book compounds at 10-11%, which implies mild forward-P/E compression — a deliberately more conservative construction than holding the earnings multiple flat.
• Bank-metric note: cash-flow-based measures are not used anywhere in this dive — the vendor's operating cash flow swings from $15,068M (FY2022) to $5,912M (FY2023) to $687M (FY2024, with free cash flow of MINUS $782M) to $6,730M (FY2025), because for a custodian bank operating cash flow includes changes in trading assets, receivables and payables and bears no relation to earning power. freeCashFlowYieldTTM (2.51%), priceToFreeCashFlowRatioTTM (39.9x), evToFreeCashFlowTTM (50.2x) and incomeQualityTTM (0.564) are all rejected. The measures used are return on tangible common equity (31.3%), pre-tax operating margin (39.8%), operating leverage (+419 bps), tangible book value per share ($32.81, +11.0% year on year), CET1 (11.0%), Tier 1 leverage (5.9%), LCR (111%) and NSFR (130%) — all filing-sourced.
• Half-weight management framing: the 19% dividend increase is an announced intention "subject to approval by the Company's Board of Directors" per the 8-K of 2026-06-24, and the "efficiency savings" and "platforms operating model" language is company framing. Both are half-weighted under house rules and identified as such at every use.
• Timing: second-quarter 2026 results were released 2026-07-15, twenty days before this dive, and beat the consensus adjusted EPS estimate by 10.3% ($2.46 against $2.23) — the fifth consecutive beat, averaging 10.1%. The next print is 2026-10-15, 72 days away (consensus adjusted EPS $2.25 and revenue $5,514M, both below the June quarter's actuals). The most recent filings are the 10-Q of 2026-07-31 and insider Forms 4 of 2026-07-17 and 2026-07-31. 2026-08-04 carried no company-specific filing.
• 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.