CB — Chubb Limited

Deep-value analysis: Chubb Limited (NYSE: CB) — REFRESH

Generated 2026-09-02 · Price: … (Yahoo, 09/02/2026) · Market cap: …B · Sector: Financial Services — Insurance P&C/Specialty · HQ: Zurich, Switzerland · Reference thesis: the 2026-07-28 report (36 days)

Primary sources: SEC EDGAR — 10-K FY2023/FY2024/FY2025 (filed 02/23/2024, 02/27/2025, 02/27/2026), 10-Q Q2 2025 (07/28/2025), 10-Q Q1 2026 (04/28/2026), 8-K plus Ex-99.1 (press release) and Ex-99.2 (Financial Supplement) for Q2 2026 results (07/21/2026). Secondary: data pack 09/02/2026, research brief 09/02/2026, yfinance, GBL v21 tracker.

Executive summary

This analysis’s mode. The mechanical triage (the quarter-over-quarter table, 09/01/2026) requested REFRESH for a single reason: of the July thesis’s seven falsifiers, six were automatically checked and all six hold, and one — “organic premium growth supports profitable scaling, despite price softening” — remained unverifiable, because the tracker was missing a second consecutive quarter of revenue growth. Nothing new was filed with the SEC after the thesis date; the price moved … from the triage moment and … from the reference report (… → …). The price move is not the reason for the revisit and I don’t build the thesis around it.

What I inherit unchanged (verified as unchanged, not re-derived): the six-segment structure and the moat sources (10-K FY2025, Item 1); the Cigna Asia 2022 acquisition history and the gradual build-up in Huatai to 87.2%; the Dividend Aristocrat status with 33 years of increases; the AM Best A++ rating; the conclusion “clean accounting, no restatements, no material weakness, no goodwill impairment”; and the pre-mortem scenario structure. All remain valid — I summarize them, not rewrite them.

What I re-derived, and what changed. Three things, all material:

1. The broken falsifier repaired itself, but not cleanly. Consolidated net earned premiums grew … in Q2 2026 (13,889 vs 13,125 mn) and … in Q1 2026 (13,457 vs 12,000 mn) — two consecutive positive quarters, so the falsifier formally holds. But net written premiums, which are the leading indicator, grew only … consolidated and … at constant currency, and on P&C … and … at constant currency (Financial Supplement Q2 2026, p.1). Property and short-tail lines fell … (… at constant currency). Real organic growth, stripped of FX, is practically zero in P&C. The falsifier didn’t break, but it’s stretched thin.

2. The old report’s owner-earnings base was methodologically wrong. The July report used TTM GAAP EPS (…) as a proxy. For a …-portfolio insurer, GAAP EPS contains volatile mark-to-market (−… mn in Q1 2026, +… mn in Q3 2025) and, more importantly, an abnormally light catastrophe load over the last four quarters: … mn pre-tax TTM versus a norm of 5.43% of P&C earned premiums (the 2023-2025 average catastrophe ratio), i.e. … mn. The correct base is normalized TTM core operating income: … mn, i.e. …/share — 4.9% below the old figure, and nearly identical to analyst consensus for 2026 (… stockanalysis.com, 25 analysts). The external cross-check confirms the normalization.

3. The old claim “buybacks funded from profit, not debt” was too strong. In FY2025 Chubb repurchased … (versus … in 2024) while financial debt rose from … mn to … mn (+… mn). Net debt issuance covered roughly 55% of the increase in buybacks. Not alarming — the … of profit comfortably covers the … returned to shareholders, and leverage remains 18.4% of adjusted market cap — but the correct wording is “partly funded with new debt, at leverage that remains moderate,” not “exclusively from profit.”

Updated thesis. The business remains exceptional and has improved operationally: P&C combined ratio 83.8% in Q2 2026 (from 85.6%), record adjusted net investment income … (…), operating ROTE 21.2%, tangible book value per share … year-over-year. Earnings quality is HIGH (Sloan accruals …, TTM CFO/NI 1.34x, favorable reserve development doesn’t explain the margin improvement). The problem remains price, but it’s less severe than in July: at … the stock is 7.8% above the triangulation’s median model, versus 1.8% in July, but the Monte Carlo simulation (20,000 scenarios, entirely missing from the old report) gives a median of … MOS and an … undervaluation probability. The divergence between the no-growth models (… to …) and the modest-perpetual-growth ones (… to …) is the whole story: at … the market is implicitly pricing roughly 1% perpetual growth at a 9% cost of capital, for a company growing tangible book value per share in double digits.

Estimated value: range …–…/share; median of the five triangulated models: … Verdict: MONITOR (deep GBL score 19/30, 63%), with higher conviction than in July. The buy threshold remains below … (roughly 10.8x normalized earnings, 1.53x book value), the level at which four of five models simultaneously flip into positive discount.