Chubb CB

Reference table

1. What the business is — and what the vendor payload thinks it is

From the 10-Q: "Chubb Limited is a holding company incorporated in Zurich, Switzerland. Chubb Limited, through its subsidiaries, provides a broad range of insurance and reinsurance products to insureds worldwide. Our results are reported through the following business segments: North America Commercial P&C Insurance, North America Personal P&C Insurance, North America Agricultural Insurance, Overseas General Insurance, Global Reinsurance, and Life Insurance."

Six segments. The vendor's seg_prod block contains one. Its FY2025 entry reads, in full:

> {"Segment Life": 7231000000}

Against $59,783 million of FY2025 revenue. The single disclosed line is 12.1% of the company. There is no data in the payload on North America Commercial P&C, North America Personal P&C, North America Agricultural, Overseas General or Global Reinsurance — which is to say, on 87.9% of Chubb. This is a worse segment-coverage failure than any recorded elsewhere in this programme, including BA's one-of-three and DELL's collapse into "Other Segments."

And seg_geo stops in 2014. The most recent geographic entry is {"UNITED STATES": 10448000000} for fiscal 2014, followed by 2013 and 2012. Twelve years stale, and useless.

Everything that follows therefore comes from the 10-Q. Goodwill by segment at 2026-06-30, which is the only complete segment-level breakdown available in the extracted filing ($M):

The segment-level current-accident-year combined ratios excluding catastrophes, from the 10-Q's own tables, are the most useful disclosure in the file:

North America Personal P&C at 69.9% is the standout — the 10-Q attributes the improvement to "improvement in homeowners and personal excess from lower underlying losses, a lower acquisition ratio resulting from a change in business mix, and a lower administrative ratio." North America Commercial P&C is the one segment moving the wrong way, and the 10-Q names the reason: "a change in the mix of business given the reduced property exposure." That is a deliberate underwriting decision — reducing property risk raises the underlying loss ratio while lowering catastrophe volatility — and it is worth reading as discipline rather than deterioration.

Two structural facts about the company that the payload also does not carry. First, Huatai Group — a majority-owned Chinese subsidiary whose life, property-and-casualty and asset-management businesses sit inside the Life Insurance segment, and which explains most of the $6,022 million of noncontrolling interests on the balance sheet. Second, the acquisition of Liberty Mutual's property-and-casualty insurance business in Thailand, which the 10-Q names as a driver of Asian growth. The vendor's acquisitionsNet line reads −$289M for FY2025 and −$25M for the first half of 2026 — small, consistent with bolt-ons, and the only transaction data available.

2. The underwriting result, which is the whole investment case

The P&C combined ratio, from the 10-Q ($ and percentage points):

Read the bottom row and then the row above it, in that order.

The bottom row — 82.2%, 82.3%, 82.2%, 82.3% across four periods — is the quality signal. It is the combined ratio for business written in the current accident year with catastrophes stripped out: no weather, no reserve releases, no prior-year noise. Chubb has earned approximately 17.8 cents of underwriting profit on every premium dollar, four periods running, with a variance of one tenth of a point. For a company writing across six segments, fifty-plus countries and every major line of business, that consistency is the moat.

The row above it — catastrophe losses of 4.0 points in the quarter against 10.5 points in the first half of 2025 — is the noise. The 10-Q is explicit that the first-half improvement from 90.4% to 83.9% "reflects the impact of the California wildfire catastrophe losses in the prior year, including the unfavorable impact of the ceded reinstatement premiums on the expense ratio." Almost the entire headline improvement is the absence of last year's wildfires, not this year's performance. Anyone reading the headline combined ratio as a 6.5-point operational gain is misreading it.

Prior-period development, which is where insurers hide and where they are found out:

> "Pre-tax net favorable PPD for the six months ended June 30, 2026 was $742 million in our active companies, including net favorable development of $715 million in short-tail lines and net favorable development of $27 million in long-tail lines... Our corporate run-off portfolio had adverse development of $173 million, primarily driven by adverse development for molestation-related claims."

Three observations. First, $742 million favourable against $587 million in the prior-year half is a genuine positive and adds 2.5 points to the reported combined ratio. Second, 96% of it is short-tail — property, auto physical damage and surety — which is the honest kind, because short-tail reserves settle quickly and are hard to manipulate. Long-tail favourable development was only $27 million, with workers' compensation favourable and general casualty adverse. Third, the run-off portfolio's $173 million of adverse development on molestation-related claims is a genuine and growing liability, up from $83 million in the prior-year half. It is disclosed, it is reserved, and it is the kind of long-tail exposure that does not stop.

Gross unpaid losses and loss expenses were $88,018 million at the start of 2026, against $84,004 million a year earlier, with reinsurance recoverable of $18,346 million. Net unpaid losses of $69,672 million is the float — the money Chubb holds and invests until claims are paid — and it grew 5.1% year on year.

3. Financials, capital and the return of capital

The annual series ($M except per share):

Revenue has risen 65.8% in five years and net income has nearly tripled, while the diluted share count has fallen 12.6%. Earnings per share went from $7.82 to $25.91 — a 3.3x increase. FY2022's dip to $12.50 is the mark-to-market and catastrophe year and is the only interruption.

The balance sheet at 2026-06-30, filing-verified:

Book value per share of $195.45 is the anchor for this dive and it is computed directly from the filing — $75,372M of Chubb shareholders' equity divided by 385,634,049 shares. At $348.335 the price-to-book ratio is 1.78x.

Capital return, and the pace has doubled:

$2,122 million in six months is 1.6% of the market capitalisation in half a year, an annualised roughly $4.2 billion or 3.2%. Combined with the $3.93 dividend at a 1.13% yield and a 13.7% payout ratio, total shareholder yield is approximately 4.3%. And critically, equity still rose $1,615 million across the half despite the buyback and dividend — the business is generating more capital than it returns.

Cash flow is not a useful lens for an insurer and is not used as one here. The vendor reports FY2025 operating cash flow of $14,537M with capital expenditure of exactly $0, giving "free cash flow" of $14,537M — for a company that plainly spends money on property and technology. capitalExpenditure of zero across all four years is a null-as-zero artefact, and every free-cash-flow-derived metric in this file is discarded. The relevant measures are the combined ratio, book value per share growth, and return on equity.

Return on equity: trailing net income of approximately $11,213M against average Chubb shareholders' equity of roughly $74.5B is approximately 15.0%. The vendor's returnOnEquityTTM field is NULL — it does not compute it — which is itself a small finding.

4. Valuation — priced in or room?

At $348.335 (market cap $134.39B, 385,634,049 shares):

Estimate coverage is asymmetric and the asymmetry matters: 16 analysts on FY2026 and FY2027 EPS but only 7 on revenue, and 7 and 5 respectively on FY2028. For an insurer that is the right way round — earnings are what the street models, revenue is an output of premium volume and investment income — but the FY2028 row rests on 7 analysts and carries no conclusion here.

The most informative single line in the estimate block: FY2026 consensus core EPS of $27.665 is BELOW the trailing $29.09 and below FY2025's GAAP $25.91-plus-realised-gains run rate. The sell-side expects Chubb to earn less in 2026 than in the trailing twelve months. That is normal for a catastrophe-exposed insurer coming off a benign first half — the consensus is loading a catastrophe allowance into the second half — and it means the FY2026 multiple of 12.6x is struck on a deliberately conservative number.

est.ebitdaAvg and est.ebitAvg are rejected on two independent grounds. First, EBITDA is not a meaningful concept for an insurer whose principal expense is claims and whose principal income is premiums and investment return. Second, they carry the fixed-ratio fabrication signature: in every forward year ebitdaAvg is exactly 21.97% of revenueAvg and ebitAvg is exactly 21.36%, to four decimals. All forward valuation uses epsAvg.

Enterprise value is likewise rejected. enterpriseValueTTM of $149.75B treats $88 billion of insurance reserves as absent and $17.6 billion of holding-company debt as leverage. returnOnInvestedCapitalTTM of 3.48% on investedCapitalTTM of $19,931M produces a 3.5% return for a company earning 15% on equity — the framework does not fit and the outputs are discarded.

Peer context. The vendor set — Bank of America, BBVA, Bank of Montreal, Interactive Brokers, ICICI, KKR, Marsh & McLennan, Progressive, Travelers, UBS — contains three genuine insurance comparables (Progressive, Travelers, Marsh & McLennan, the last a broker rather than an underwriter) among seven banks and asset managers. No peer multiples are supplied in the file, so no peer-multiple comparison is drawn.

4a. What today's price assumes (the inversion)

At $348.335 — 11.9x FY2027 consensus core EPS and 1.78x filing-verified book value — the price embeds:

• The current-accident-year combined ratio excluding catastrophes stays near 82%. (Our derivation; the 10-Q supplies the figure.) This is the most fragile assumption in the price, and it is fragile for a reason that has nothing to do with Chubb's execution: 82.2% is what a disciplined underwriter earns near the top of a hard pricing cycle. Every point of deterioration is roughly $450 million of annual pre-tax underwriting income on a P&C net premium base of roughly $45 billion — around 4% of pre-tax income per point.
• Core operating EPS of $27.67 in 2026 and $29.14 in 2027. (Consensus; 16 analysts each.) The first half delivered $14.08 ($6.82 plus $7.26), so the full-year consensus implies $13.59 in the second half — a 3.5% sequential decline that is the catastrophe allowance. The nearest test is the 2026-10-27 print at a $6.32 consensus, against $7.26 delivered in June.
• Book value per share compounds at roughly 6-8% while the buyback runs at roughly 3.2% a year. (Our number, from the H1 movement: equity up 2.2% after $2,122M of repurchase.) At a 15% return on equity and a 13.7% dividend payout, retained earnings alone support high-single-digit book growth; the buyback converts part of it into per-share growth.
• The market keeps paying roughly 1.8x book. (Our number.) At 1.5x book the price is $293; at 2.1x it is $410. On a book value that grows 7% a year, the multiple is worth more than two years of compounding, and it is the widest source of variance in the outcome.
• Catastrophe losses run near the 4-5 point long-run average rather than the 10.5 points of the first half of 2025. (Our number, from the 10-Q's own disclosure.) The 2026-10-27 print covers the peak of the Atlantic hurricane season and is a weather report.

4b. The return bridge (why the multiple moves)

Expected return over the next twelve months decomposes as: core EPS growth (+5.3%, from FY2026E $27.665 to FY2027E $29.138) + multiple drift (roughly HELD, 12.6x to about 12.4x on the respective forward year) + shareholder yield (+4.3%) ≈ +8% to +11%.

Our base assumes the multiple holds rather than expands, and the reason is that 1.78x book is already the upper half of Chubb's historical range. An insurer's multiple is anchored by book value and by return on equity: at a 15% return on equity, a 1.78x price-to-book implies the market is capitalising the excess return over cost of capital at a reasonable but not depressed rate. There is no obvious re-rating catalyst and we are not underwriting one.

Roughly half the expected return is shareholder yield, which is the most robust half. If the multiple held at 12.6x on FY2027E the price would be $367 (+5.4%); at 14x, $408 (+17.1%); at 11x, $321 (−7.9%). The base case does not require the crowd to change its mind about anything — it requires only that Chubb keeps underwriting at 82% and keeps buying back 3% of itself a year.

4c. Variant perception (where we differ, what would surprise)

• We differ from the street by 4.0%, upward — $370 against $355.86 — and it is a small, honest disagreement about the multiple rather than the earnings. With 23 of 43 analysts at buy or better and 18 at hold, the coverage is split. Our number is above consensus because we weight the buyback acceleration (doubled year on year) and the flat current-accident-year combined ratio more highly than the FY2026 EPS estimate that sits below trailing earnings. That is a modest edge and we do not pretend it is more.
• We differ from the vendor data almost completely, and this is the clearest example in the batch of why the filings matter. The payload knows one of six segments, its geography stopped in 2014, it reports zero operating income for a quarter in which the company earned $2.3 billion pre-tax, and it publishes three mutually inconsistent equity-per-share figures. An investor screening this name on the vendor block would not know what business Chubb is in. Every operational figure in this dive comes from the 10-Q.
• We think the current-accident-year combined ratio is the number to watch and that it is under-discussed relative to the headline. The headline combined ratio improved 6.5 points in the first half — from 90.4% to 83.9% — and almost all of it is the absence of the January 2025 California wildfires. The underlying figure moved one tenth of a point. Watchable number: the CAY combined ratio excluding catastrophes in the 2026-10-27 print. Anything above roughly 84% would be the first genuine evidence of the cycle turning, and it would matter far more than a hurricane.
• We think the insider file deserves more weight here than in most names. John Keogh, president and chief operating officer, sold 23,000 shares at approximately $321.50 on 27 May 2026 and gifted 62,352 shares over 27-28 May, taking his holding from approximately 229,000 to 140,970 — a 38% reduction in one week. Executive Vice President Joseph Wayland sold 8,502 shares at $364.54 on 28 July, above today's price. Three officers sold; none bought. This is not a scandal and gifts are frequently charitable, but a 38% reduction by the chief operating officer is the largest single insider position change in this batch and it is reported as such.
• Positive surprise that would force a re-rate: a third quarter with catastrophe losses below 4 points and the current-accident-year combined ratio still near 82%, which would compound book value materially; or an acceleration of the buyback beyond the current $4 billion annualised pace; or the first substantive disclosure of Huatai's contribution as a separate growth engine.
• Negative surprise that would break the thesis: a current-accident-year combined ratio excluding catastrophes above 84%, which would say the pricing cycle has turned; adverse development in the run-off portfolio materially above the $173 million half-year rate on molestation claims; a major catastrophe quarter above 12 points; or any adverse development in China affecting Huatai, which carries $4,853 million of segment goodwill.

Synthos fair values

All three anchors are multiples of the FY2027 consensus core operating EPS distribution (mean $29.138, low $26.199, high $31.631, 16 analysts), each cross-checked against filing-verified book value of $195.45 per share.

• Bear ~$270 — 10.3x the FY2027 consensus LOW of $26.199, and 1.38x current book value. Cross-check: 9.8x FY2026E; 0.9% above the 52-week low of $267.68. The scenario: the property-and-casualty pricing cycle turns, the current-accident-year combined ratio drifts toward 86%, a heavy catastrophe year lands, and the market pays a mid-cycle book multiple. −22.5%.
• Base ~$370 — 12.7x the FY2027 consensus MEAN of $29.138, and 1.89x current book value (roughly 1.77x book as it stands at the end of 2027 on high-single-digit accretion). Cross-check: 13.4x FY2026E; 11.7x FY2028E. Sensitivity, stated openly: 11.5x gives $335 — 3.8% BELOW spot — and 14x gives $408. The whole answer lives in an 11.5-14x band and the market at 11.9x is near the bottom of it. The scenario: underwriting holds near 82%, catastrophes run at the long-run average, the buyback continues at roughly 3.2% a year, and the multiple does not move. +6.2%.
• Bull ~$445 — 14.1x the FY2027 consensus HIGH of $31.631, and 2.28x current book. Cross-check: 16.1x FY2026E; 22.4% above the 52-week high of $363.50. The scenario: a benign catastrophe year, continued sub-83% underwriting, an accelerated buyback, and a re-rating toward the multiple a high-return-on-equity specialty franchise can command. +27.8%.

Base is 6.2% above spot; asymmetry roughly 1.24:1 to the upside (22.5% down, 27.8% up), plus a 4.3% total shareholder yield. That is a modest payoff ratio on a genuinely high-quality business, and with zero independent conviction behind it, it is a Watch with a named entry rather than a position.

5. Knowledge base — zero claims, stated as a finding

Raw hits on a case-sensitive entity and text sweep: ZERO. Entity matches: 0. Name-level claims: 0. Discarded from the broader sweep: 5.

The primary sweep ran case-sensitive entity terms CB, Chubb, Chubb Limited and Huatai, together with a case-sensitive text sweep on Chubb and Greenberg, across all 51,928 distilled claims. It returned exactly zero hits.

That result is worth stating precisely, because the batch brief specifically flagged "CB" as a short-ticker homograph risk requiring a case-sensitive check. The homograph risk did not materialise — not because the check filtered noise, but because no entity in the Synthos knowledge base is tagged CB or Chubb at all. There is nothing to disambiguate.

A broader, case-insensitive text sweep on chubb, property/casualty, P&C insur and reinsur returned five hits, and all five are discarded:

> 2024-10-20 · bullish 65 · thesis · no entities · channel: business_breakdowns — "Hardening coastal insurance (10-20% of cost) favors scaled owners who can spread risk and reach reinsurers; underinsured mom-and-pops become forced sellers." A real-estate claim about insurance costs, not an insurer claim.

> 2025-01-23 · neutral 75 · principle · no entities · channel: no_priors — "Regulated admitted insurers barred from pricing in reinsurance costs and forward-looking risk models will exit high-risk markets — market failure until rules change." The most substantively relevant of the five — it describes the regulatory dynamic behind Chubb's own "reduced property exposure" — but it names no company.

> 2026-04-10 · neutral 80 · principle · no entities · channel: odd_lots — "The 12 International Group P&I clubs insure ~90% of oceangoing tonnage and jointly buy the world's largest reinsurance policy, giving purchasing power for cheap high-limit cover (~$3B) despite being competitors." Marine mutual insurance, a market Chubb does not operate in this way.

> 2026-03-14 · neutral 70 · principle · no entities · channel: we_study_billionaires — "Long-track-record investors must evolve within their wheelhouse (Davis into reinsurance/international; Buffett from cigar-butts to quality) to keep succeeding." An investing-philosophy claim.

> 2026-06-01 · bullish 68 · thesis · no entities · role: independent · channel: empire — "Re, an on-chain reinsurance business, uses stablecoin markets as its capital function, paying real 12-14% yield from a genuinely productive business — 5-7x leverage on treasuries. In portfolio." A crypto claim about a different company whose name is the word "Re".

Conclusion, stated as a finding. Breadth 0, claim count 0, net conviction none. The Synthos knowledge base contains no claim about Chubb Limited, and — more broadly — no entity-tagged claim about any large property-and-casualty insurer anywhere in the five hits this sweep surfaced. That is a coverage gap in the knowledge base, not a judgement about the company, and it should be recorded as such: the ingest pipeline's source list does not include voices that discuss insurance underwriting. For this dive the consequence is direct: there is no independent conviction to support a position, no concentration test to run, and a name with a 1.24:1 payoff ratio and an empty lane defaults to Watch.

6. Data integrity — the worst vendor file in this batch

Eight findings. Chubb's payload is the least usable in this batch: five of six segments missing, geography twelve years stale, a quarter reported with zero operating income, and three mutually inconsistent equity-per-share figures.

1. seg_prod contains ONE of SIX reportable segments — REJECTED in full. The FY2025 entry is a single line, "Segment Life": $7,231M, against reported revenue of $59,783M. The 10-Q states the six segments explicitly: "North America Commercial P&C Insurance, North America Personal P&C Insurance, North America Agricultural Insurance, Overseas General Insurance, Global Reinsurance, and Life Insurance." 87.9% of the company is absent from the payload. All segment analysis in this dive comes from the 10-Q's goodwill and combined-ratio tables.

2. seg_geo stops at fiscal 2014 — twelve years stale, REJECTED. The most recent entry reads {"UNITED STATES": 10448000000} for 2014, followed by 2013 and 2012. There is no geographic data for any year since. No geographic figure appears in this dive.

3. The first quarter of 2026 is reported with operatingIncome and incomeBeforeTax of ZERO — a null-as-zero defect. The same row reports netIncome of $2,320M and incomeTaxExpense of $646M, which together imply pre-tax income of at least $2,966M. A quarter cannot have zero pre-tax income and $646 million of tax. No trailing operating-margin figure in this dive uses that quarter, and operatingProfitMarginTTM of 18.19% is computed by the vendor across a series containing this zero and is not used.

4. Three mutually inconsistent equity-per-share figures — resolved from the filing. The payload reports bookValuePerShareTTM of $205.712, shareholdersEquityPerShareTTM of $191.786, and tangibleBookValuePerShareTTM of $138.313. The first is total equity including $6,022 million of noncontrolling interests divided by shares — noncontrolling interests do not belong to Chubb shareholders and must not appear in a per-share book value. The 10-Q settles it: Chubb shareholders' equity of $75,372M at 2026-06-30 divided by 385,634,049 shares outstanding gives BOOK VALUE PER SHARE OF $195.45, and that is the figure used throughout this dive. Note that priceToBookRatioTTM of 1.816 is computed off the middle figure, not the one labelled book value — internally inconsistent, and superseded here by the filing-derived 1.78x.

5. tangibleBookValuePerShareTTM deducts goodwill only, and the intangibles line is unstable — flagged, not used. The vendor's $138.313 implies equity less goodwill only ($73,757M − $20,207M ≈ $53,550M). The 10-Q's own definition, from the compensation note, is "Chubb shareholders' equity less goodwill AND intangible assets attributable to Chubb, net of tax" — a stricter test that would give roughly $89-95 per share. Compounding the problem, the vendor's intangibleAssets line jumped from $9,600M (FY2024) to $19,224M (FY2025) — a $9,624M increase against total acquisition spending of $289M that year. That is a classification change, not an economic one, and we cannot resolve which convention it now uses. We therefore anchor on book value per share ($195.45, filing-verified) and state that price-to-tangible-book sits somewhere between roughly 2.5x and 3.9x depending on the definition, without claiming a single figure.

6. est.ebitdaAvg and est.ebitAvg carry the fixed-ratio fabrication signature and are conceptually void for an insurer — REJECTED. In every forward year ebitdaAvg is exactly 21.97% of revenueAvg and ebitAvg is exactly 21.36%. Separately, EBITDA has no meaning for a business whose largest expense is claims. All forward valuation uses epsAvg.

7. Enterprise value, invested capital and free cash flow are structurally inapplicable — REJECTED as a class. enterpriseValueTTM of $149.75B treats $88,018M of insurance reserves as absent. returnOnInvestedCapitalTTM of 3.48% on investedCapitalTTM of $19,931M produces a 3.5% return for a company earning approximately 15% on shareholders' equity. And capitalExpenditure reads exactly $0 in all four cash-flow years, making "free cash flow" identical to operating cash flow at $14,537M — a null-as-zero artefact that invalidates every free-cash-flow multiple in the file. returnOnEquityTTM is NULL and is computed here directly.

8. earn_cal revenue actuals do not reconcile to any consistent basis. The Q3 2025 entry reports revenue "actual" of $16,139M against an estimate of $12,982M — a 24% beat that is a basis mismatch, not a result; the same field's Q2 2026 entry reports $14,705M against the inc_q figure of $15,816M. The EPS actuals are internally consistent as core operating earnings per share ($7.49, $7.52, $6.82, $7.26) and are used; the revenue actuals are not used anywhere.

Correct and worth recording: the share count is an exact match (10-Q 385,634,049 outstanding at 2026-06-30 against 385,796,000 implied by market capitalisation ÷ price, a 0.04% difference); debtToEquityRatioTTM of 0.240x and interestCoverageRatioTTM of 14.1x are both consistent with the FY2025 balance sheet; dividendPerShareTTM of $3.93 and dividendPayoutRatioTTM of 13.7%; and quote.yearHigh/yearLow ($365.91/$265.30) sit within 0.9% of tech.hi52/lo52 ($363.50/$267.68), with tech used throughout.

Non-equity tripwire — checked and passed. CB is common stock, CHF 0.50 par value, NYSE-listed, per the 10-Q balance sheet (400,120,847 shares issued, 14,486,798 in treasury, 385,634,049 outstanding). Price of $348.335 is not par-like; beta is 0.406; the dividend is variable and has been raised over time; volume was 1.11M shares (~$385M of turnover); the 52-week band of $267.68 to $363.50 is a 36% range. Note the Swiss domicile and the CHF-denominated par value — this is a foreign private issuer's US listing filing on domestic forms, and it files 10-K and 10-Q rather than 20-F. This is common equity.

7. Technicals

• Price $348.335. −4.2% from the 52-week high of $363.50; +30.1% above the low of $267.68. Position within the annual range: 84th percentile.
• Above both moving averages, and the spacing is tight: +2.9% above a 50-day average of $338.52; +9.1% above a 200-day average of $319.22. Both rising.
• RSI 58.1 — the most neutral reading in this batch, against BMY's 87.1, LMT's 81.0 and APP's 39.0. MACD +3.27.
• Maximum drawdown from peak over the trailing year: −4.2% — the shallowest in this batch by a wide margin, and consistent with a 0.406 beta.
• Relative performance: 3-month +7.1% against SPY +7.6% — essentially in line; 6-month +11.9% against SPY +11.1%; 12-month +30.3% against SPY +24.3% and QQQ +30.8%. A 6-point twelve-month outperformance from a low-beta insurer is a good result, and the three-month figure shows the outperformance has stalled.
• Sentiment: 1 strong buy, 22 buy, 18 hold, 2 sell across 43 analysts; consensus target $355.86 (+2.2%), median $361, high $387 (+11.1%), low $301 (−13.6%). The street's implied upside of 2.2% is the lowest in this batch.

Today's move and what it does to the entry

CB closed 2026-08-04 at $348.335, up 0.03% or $0.105 from $348.23. It opened at $344.44, traded $342.22 to $348.94, and closed near the high on 1.11M shares. On a day when most of this batch moved sharply, Chubb was flat — which is precisely what a 0.406-beta insurer does, and is a reminder of what this security is for.

The honest read: there is nothing in this chart to act on, in either direction. No gap, no extreme, no divergence. The entry question is entirely a valuation and conviction question, and the answer to both is "not quite."

What would change it:

• A price near $300 — roughly 10.3x FY2027E and 1.53x book value, close to the street's low target — which would put our base $370 at a 23% premium. That is 14% below today's close and inside the range this stock traversed in the last twelve months.
• Or the 2026-10-27 print with catastrophe losses below 4 points and the current-accident-year combined ratio still at or below 82%, which would confirm that the underwriting discipline survives a hurricane quarter. At that point paying 1.8x book is comfortable.

8. Insiders — three officers sold, and the chief operating officer cut his holding 38% in a week

The reading, and this is the one genuinely notable insider file in the batch.

John Keogh, the president and chief operating officer, reduced his position from approximately 229,146 shares to 140,970 in two days — 23,000 shares sold on the open market for approximately $7.4 million, and 62,352 shares given away across 27 and 28 May. That is a 38.5% reduction in the second-most-senior executive's personal holding, executed in a single week.

Two qualifications and one that cuts the other way. Gifts are most often charitable or family transfers and carry no view on the stock — but 62,352 shares is $20.1 million at the transaction-period price, and it is far larger than the sale beside it. The open-market sale of 23,000 shares is itself modest against a residual holding worth approximately $49.1 million at today's close. What does not soften: it is a sale, not a tax withholding, and there is no F-InKind transaction anywhere in this file.

Juan Luis Ortega exercised options at $139.01 and sold the resulting 3,886 shares at $322.08 — a net-settled exercise-and-sell, mechanical, and the residual holding of 28,859 shares is a 12% reduction. Joseph Wayland sold 8,502 shares at $364.54 on 28 July, one week after the second-quarter print and 4.7% ABOVE today's closing price, retaining 33,749.

What the file contains is therefore: zero purchases, zero tax withholdings, three officers selling on the open market, and the chief operating officer reducing his stake by more than a third. That is a mild negative and it is reported as one. No transaction by Evan Greenberg, the chief executive, appears anywhere in the file — neither purchase nor sale.

9. Verdict, kill-criteria and flip conditions

Watch.

What is genuinely excellent: a current-accident-year combined ratio excluding catastrophes of 82.2%, held within one tenth of a point across four consecutive reported periods — the single most impressive operating statistic in this batch; a headline combined ratio of 83.8% in the June quarter; $742 million of net favourable prior-period development in the half, 96% of it in short-tail lines; North America Personal P&C at 69.9%, improved 2.3 points; $75,372 million of shareholders' equity growing 2.2% in a half-year despite $2,122 million of buyback; a share count down 1.4% in six months with the repurchase pace doubled year on year; 0.24x debt-to-equity and 14.1x interest coverage; a return on equity near 15%; and a 0.406 beta with a −4.2% maximum twelve-month drawdown.

What holds it at Watch: a base fair value of $370, only 6.2% above spot, with a payoff ratio of 1.24:1; a street consensus target of $355.86 implying just 2.2% upside, the lowest in this batch, with 18 of 43 analysts at hold; a FY2026 consensus EPS of $27.67 that sits BELOW the trailing $29.09; a price-to-book of 1.78x that is in the upper half of the range this franchise has historically commanded; an insider file in which three officers sold, none bought, and the president and chief operating officer cut his holding 38% in a week; a third-quarter print that is a hurricane report; and — decisively for the tier — a knowledge-base lane of exactly zero claims in 51,928.

The distinction that matters. Chubb is a better business than most names in this batch and it is not expensive. What it does not have is a reason to act today: 6% of upside, no independent conviction, and a seasonally hostile quarter directly ahead. In the Synthos frame, that is a Watch with a specific number attached — roughly $300, or 1.53x book — and a specific event that would substitute for it, namely a clean hurricane quarter. Owning this at 1.53x book with an 82% combined ratio would be a genuinely good position; owning it at 1.78x for 6% is not.

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

• The current-accident-year combined ratio excluding catastrophes above 84% in any quarter. That is the pricing cycle turning, and it is worth roughly 4% of pre-tax income per point.
• Adverse prior-period development in the active companies, or run-off adverse development materially above the $173 million half-year rate on molestation-related claims.
• A catastrophe quarter above 12 combined-ratio points outside a genuinely exceptional event — the first half of 2025 ran 10.5 points on the California wildfires.
• A suspension or material slowing of the buyback, which supplies roughly 3.2% of the annual return.
• Any adverse development affecting Huatai Group, which carries $4,853 million of Life-segment goodwill and underlies most of the $6,022 million of noncontrolling interests.
• A change to the Swiss tax regime materially affecting the holding-company structure.

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

• A price near $300 — roughly 10.3x FY2027E and 1.53x filing-verified book value — with the underwriting intact. This is the single most likely path to a Buy and it is 14% below today's close.
• The 2026-10-27 print with catastrophe losses below 4 points and the current-accident-year combined ratio at or below 82%, which would confirm the discipline survives the hurricane season and remove the principal near-term uncertainty.
• An acceleration of the buyback beyond roughly $5 billion annualised, which would take total shareholder yield above 5% and materially improve the return bridge.
• The first substantive separate disclosure of Huatai's economics, which would convert the one genuinely differentiated asset in this company from an assertion into a number.
• Any knowledge-base coverage of the insurance sector appearing at all. At present the lane is empty and the sector is uncovered, and that alone is worth roughly a tier.

Where CB fits in the Synthos Framework Portfolio. The financials sleeve, at 0% today with a 2-3% target on a fill near $300 or a confirming October print. On batch overlap: COF is also in this batch and also in the financials sleeve, but the exposures are opposite — Chubb is a low-beta underwriter with a 4.2% maximum drawdown, Capital One is a levered bet on the consumer credit cycle — and holding both is a genuine barbell rather than a concentration. Logged as a tracked Synthos call (Watch) as of 2026-08-04 at $348.335.

Single biggest risk: a soft market. The 82.2% current-accident-year combined ratio excluding catastrophes is the whole investment case, and it is what a disciplined underwriter earns near the top of a hard pricing cycle. Nothing in the filings tells you what Chubb earns three years into a soft one, and the 10-Q already shows the first small sign: North America Commercial P&C's underlying combined ratio deteriorated 0.7 points to 81.8%, which the company attributes to "a change in the mix of business given the reduced property exposure." Each point of combined-ratio deterioration is roughly $450 million of annual pre-tax underwriting income — about 4% of pre-tax profit. A four-point drift toward 86% would take approximately 16% out of earnings, and the multiple would compress with it. That is the $270 bear case, and it requires no catastrophe, no reserve failure and no management error — only for competitors to start pricing more aggressively, which is what they always eventually do.

Provenance & disclosures

• Traceability: ZERO knowledge-base claims name Chubb Limited out of 51,928 distilled claims (breadth 0, claim count 0, net conviction none). A CASE-SENSITIVE entity sweep on CB, Chubb, Chubb Limited and Huatai, plus a case-sensitive text sweep on "Chubb" and "Greenberg", returned EXACTLY ZERO hits. The short-ticker homograph risk the batch brief flagged for CB did not materialise, because no entity in the knowledge base is tagged CB or Chubb at all — there is nothing to disambiguate. A broader case-insensitive text sweep on chubb, property/casualty, P&C insur and reinsur returned five hits, all discarded as sector noise and none entity-tagged to any insurer: business_breakdowns 2024-10-20 (a real-estate claim about coastal insurance costs), no_priors 2025-01-23 (conviction 75, the most substantively relevant — regulated admitted insurers exiting high-risk markets — but naming no company), odd_lots 2026-04-10 (conviction 80, marine P&I mutuals), we_study_billionaires 2026-03-14 (an investing-philosophy claim), and empire 2026-06-01 (a crypto claim about an on-chain business named "Re"). This is a coverage gap in the knowledge base rather than a judgement about the company: the ingest source list contains no voice that discusses insurance underwriting. No concentration sensitivity test is possible on an empty lane. The consequence for this dive is direct and is stated in the verdict: a name with a 1.24:1 payoff ratio and zero independent conviction defaults to Watch.
• Data as-of: fundamentals — combined ratios, segment goodwill, prior-period development, equity, share count and buyback through 2026-06-30, all filing-verified from the 10-Q filed 2026-07-28; annual figures for FY2025 and earlier from the vendor payload cross-checked against the 10-K filed 2026-02-27 · estimates 2026-08-04 · prices 2026-08-04, quote timestamp 1785873598 = 2026-08-04T19:59:58Z ($348.335, +0.03%; 50-DMA $338.52; 200-DMA $319.22; RSI 58.1; MACD +3.27) · knowledge-base claims 2026-08-04. Chubb's fiscal year is the calendar year. The company is incorporated in Zurich, Switzerland, and files on US domestic forms (10-K, 10-Q). All figures come from the Synthos vendor data file for CB or from the SEC filings in the CB archive; no figure comes from memory, recall or external retrieval.
• Filing archive contents: 10-K filed 2026-02-27 (fiscal 2025); 10-Q filed 2026-04-28 (March 2026 quarter); 10-Q filed 2026-07-28 (June 2026 quarter — the source of essentially every operational figure in this dive, because the vendor payload cannot supply them); 8-K filed 2026-05-22; 8-K filed 2026-06-10; 8-K filed 2026-07-21. All carry preserved [TABLE] statement data.
• Where the filings supplied what the vendor could not (detailed in Section 6) — this is the most comprehensive substitution in the batch: seg_prod contains ONE of SIX reportable segments (a single "Segment Life" line of $7,231M against $59,783M of revenue), so all segment analysis comes from the 10-Q's goodwill and combined-ratio tables; seg_geo stops at fiscal 2014, so no geographic figure appears in this dive at all; the Q1 2026 income statement reports operatingIncome and incomeBeforeTax of ZERO against $2,320M of net income and $646M of tax, so no trailing operating-margin figure uses it; three mutually inconsistent equity-per-share figures ($205.712 including $6,022M of noncontrolling interests, $191.786 excluding them, $138.313 tangible) are superseded by the filing's $75,372M of Chubb shareholders' equity ÷ 385,634,049 shares = $195.45 book value per share, giving 1.78x price-to-book; tangibleBookValuePerShareTTM deducts goodwill only against the 10-Q's stricter own definition, and the intangibleAssets line jumped $9,624M year on year against $289M of acquisition spending — an unresolvable classification change, so no single price-to-tangible-book figure is claimed; est.ebitdaAvg/ebitAvg rejected for a fixed-ratio signature (exactly 21.97% and 21.36% of revenue in every forward year) and for being conceptually void for an insurer; enterprise value, invested capital and free cash flow rejected as a class, with returnOnInvestedCapitalTTM of 3.48% against an actual return on equity near 15% and capitalExpenditure of exactly $0 in all four years; returnOnEquityTTM is NULL and is computed here directly; and earn_cal revenue actuals reconcile to no consistent basis (Q3 2025 "actual" of $16,139M against a $12,982M estimate) and are not used, though the EPS actuals are internally consistent as core operating EPS and are. Where vendor and filing AGREED: the share count is an exact match (10-Q 385,634,049 at 2026-06-30 against 385,796,000 implied); debtToEquityRatioTTM of 0.240x; interestCoverageRatioTTM of 14.1x; the dividend of $3.93 at a 13.7% payout; and quote versus tech 52-week ranges within 0.9%.
• Basis note: the earn_cal "actual" EPS figures are core operating earnings per share, Chubb's own non-GAAP measure, which excludes realised and unrealised investment gains and losses. Trailing core operating EPS is $29.09 ($7.49 + $7.52 + $6.82 + $7.26) against trailing GAAP EPS of $28.53 — an unusually small gap for an insurer, because the investment portfolio has been quiet. Analyst estimates are on the same core operating basis and all forward multiples in this dive use it. The trailing multiple is shown on both (12.0x core, 12.2x GAAP) and no conclusion turns on the choice.
• Estimate coverage: 16 analysts on both FY2026 and FY2027 core operating EPS — FY2027 is the anchor for all three fair values, with a range of $26.199 to $31.631 (20.7%); only 7 analysts on FY2026 and FY2027 revenue, and 7 and 5 on FY2028 EPS and revenue. The FY2028 row carries no conclusion. Note that FY2026 consensus EPS of $27.665 sits below trailing core EPS of $29.09 — the sell-side is loading a second-half catastrophe allowance, and the FY2026 multiple is therefore struck on a deliberately conservative number.
• Peer note: the vendor peer set — Bank of America, BBVA, Bank of Montreal, Interactive Brokers, ICICI, KKR, Marsh & McLennan, Progressive, Travelers, UBS — contains seven banks and asset managers among three insurance-adjacent names (Progressive and Travelers are genuine underwriting comparables; Marsh & McLennan is a broker). No peer multiples are supplied in the file, so no peer-multiple comparison is drawn.
• Fair-value caveat: the $270 / $370 / $445 anchors are multiples of the FY2027 consensus core operating EPS distribution — 10.3x the low of $26.199, 12.7x the mean of $29.138, and 14.1x the high of $31.631 — each cross-checked against filing-verified book value of $195.45 per share (1.38x, 1.89x and 2.28x respectively). Stated arithmetic, not a discounted cash flow. The base is sensitivity-disclosed: 11.5x gives $335, which is BELOW spot; 14x gives $408. The base case assumes NO multiple expansion; roughly half the expected return is the 4.3% total shareholder yield, which is the most robust component.
• Timing: second-quarter 2026 results were released 2026-07-21, fourteen days before this dive, beating the consensus core operating EPS estimate by 7.2% ($7.26 against $6.77). The next print is 2026-10-27, 84 days away — the peak of the Atlantic hurricane season and therefore the most catastrophe-exposed quarter of Chubb's year, which is an explicit input to the neutral short-horizon stance. The most recent insider transaction is dated 2026-07-28 and is a sale. 2026-08-04 carried no company-specific filing and the stock was flat, which is what a 0.406-beta insurer does on a broad up-day.
• 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.