Arch Capital Group Ltd. (NASDAQ: ACGL) — Cheap on a Clean Book, Dear on a Cresting Cycle

Independent equity research note. Report date: 2026-06-26. All figures USD unless noted. Primary sources: SEC filings (10-K FY2025, 10-Q Q1-2026, DEF 14A 2026, Form 4 corpus) and public company disclosures.

⚡ Claude’s Take

This block is the author’s own independent opinion and general information, not investment advice. The analysis that follows (Sections 1–15) is presented position-free; only this block takes a view.

Verdict: HOLD / accumulate-on-weakness. Fair-value zone ≈ $95–115 (≈1.4–1.6× book on a 13–15% through-cycle ROE). Accumulate into the low-$80s / sub-$85 (≈1.2–1.25× book). Not-a-short. Medium conviction.

Arch is one of the best-run insurance franchises in the world — a disciplined, counter-cyclical, capital-fungible compounder that has grown book value per share at roughly 16–18% a year for two decades and currently earns a 15–17% operating ROE. You can buy it today at 1.40× a clean book (no AOCI mirage; the 2022 rate-shock losses have fully reversed), a discount to lower-ROE peers Chubb (~1.7×) and Travelers (~2.0×). That is the bull case in one sentence, and it is a real one. The catch — and it is the whole debate — is that the optically arresting ~7× trailing P/E is peak-earnings cheap, not cheap. All three of Arch’s engines are cresting at the same time: reinsurance pricing is rolling over after the 2023–24 hard market, specialty P&C property rates are falling, and the mortgage book is harvesting reserve releases off the most benign credit environment it will ever see. Strip the ~$600M of favorable reserve development, the rate-boosted investment income, and the mortgage-flattered combined ratio, and normalized operating EPS is closer to $8.0–8.8 on a 13–15% through-cycle ROE, not the ~$10.8 the last twelve months printed. At 1.40× book the market is already discounting that normalization — back-solving the justified-P/B formula, today’s multiple implies a durable ROE of only ~11.5–12%, i.e. cost-of-capital-plus with zero re-rating optionality priced in.

So this is quality at a fair price with a thin margin of safety, not a fat one. If Arch’s franchise is what its record says it is, a sustained 13–15% ROE makes 1.40× book modestly too cheap and the stock should compound at roughly book-value growth (~10%/yr) with re-rate upside. If the synchronized soft cycle plus fading releases drag realized ROE to 11–12%, 1.40× is simply fair and you clip mid-single-digits. The asymmetry is gently favorable but not compelling at $94 — hence accumulate-on-weakness rather than chase. Framing (from the factor tape): beta 0.33, LowVolatility +0.86, Value +0.19, ~14% off its October-2024 high, +5.7% over the past year — an out-of-favor, plateaued, low-volatility quality-value name, explicitly not a falling knife and not a crowded momentum trade. Conviction: medium. Bull-flip: operating ROE holds ≥15% for 2–3 quarters with low-teens BVPS growth as the cycle softens → re-rate toward 1.6–1.8× book. Bear-flip: ROE breaks below ~13% accompanied by adverse reserve development and shrinking premium → the franchise is mid-teens-ROE-at-best and 1.40× was the ceiling. Tag: “The cheapest thing about it is the P/E — and the P/E is the one number you should ignore.”

📈 Stock Price Action — Five-Year Event Map

Over five years ACGL roughly tripled — from ~$38 in mid-2021 to an all-time high of $110.75 on 7 October 2024 — riding the post-pandemic hard market in P&C and reinsurance, and has since gone sideways-to-down for ~20 months, trading a $82–103 band. At $94.33 (25 June 2026) it sits ~14.8% below its high, near the middle of a 52-week range of $82.45–$103.39, just under its flattening 200-day EMA (~$93.7). The arc is textbook: a violent re-rating into a hardening cycle, then a plateau as the market began to price the cycle’s turn.

Cycle narrative. (1) Through 2021–22 the stock barely moved despite a hardening market because rising rates pushed unrealized bond losses through AOCI, masking strong underlying earnings — a price lag, not an earnings lag. (2) 2023 was the breakout: the hard market translated into a 79.3% combined ratio and a 21.6% operating ROE, amplified by a one-time ~$1.18B Bermuda deferred-tax benefit (worth ~$3.10/share) as Bermuda introduced a corporate income tax. (3) Into late-2024 the compounding continued and the MCE/Allianz mid-corp acquisition added scale; the stock peaked at $110.75. (4) The ~17% pullback coincided with the market’s recognition that reinsurance pricing was topping (the January-2025 renewal was the cyclical high-water mark) and with the orderly retirement of CEO Marc Grandisson (Nicolas Papadopoulo became CEO, effective 13 October 2024). (5) Since then the tape has been a standoff: still-high reported earnings and aggressive 2025–26 buybacks against a visibly softening pricing environment and a decelerating ROE (21.6% → 18.9% → 17.1% → 15.4% in Q1-2026). The price move is Fact; the attributed causes are Interpretation.

1. Executive Summary

Arch Capital Group is a Bermuda-domiciled global specialty insurer, reinsurer and mortgage insurer that has compounded book value per share at roughly 16–18% annually since its 2001 recapitalization — one of the strongest long-run records in the industry — by relentlessly prioritizing underwriting profit over premium volume and by moving capital opportunistically across three businesses whose cycles do not fully correlate. In FY2025 the company wrote $22.9B of gross premium, earned a consolidated combined ratio of 82.8%, and produced $4,359M of net income to common ($11.60 diluted EPS) and a 17.1% operating ROE, against a year-end common book value of ~$65/share.

The investment question is not whether Arch is a good business — it plainly is — but what you are paying for, and at what point in the cycle. The headline ~7× trailing P/E (7.9th percentile of Arch’s own ten-year history) is the single most misleading number on the page. It rests on earnings that are simultaneously flattered by (i) hard-market underwriting margins now softening, (ii) ~$600M of favorable prior-year reserve development, (iii) a mortgage segment running at a 14.6% combined ratio off peak-benign credit, and (iv) net investment income boosted by a rate cycle that is now easing. Normalize those and through-cycle operating EPS is closer to $8.0–8.8 with a 13–15% ROE. On the metric that actually matters for an insurer — price-to-book against through-cycle ROE — Arch trades at 1.40× a clean book (the 49.8th percentile of its own history, i.e. mid-range, not cheap), which the justified-P/B framework shows is consistent with a durable ROE of only ~11.5–12%.

The bull and bear cases reduce to one contested number: the through-cycle ROE. If it is 13–15% (our base case, and what the franchise’s quality argues), 1.40× book is modestly too cheap and the stock should compound near book-value growth with optional re-rating toward the 1.6–2.0× that Chubb, Travelers and W.R. Berkley command. If synchronized softening across all three engines plus fading reserve releases pull realized ROE toward 11–12%, then 1.40× is simply fair. Capital allocation is a genuine strength — counter-cyclical buybacks (paused at ~$0 in 2023–24 to fund the hard market and the MCE deal, resumed at ~$1.9B in 2025), no value-destructive M&A, comp tied to growth in tangible book value per share — and removes a common way these stories go wrong. The balance sheet is fortress-grade (debt/total-capital ~10%, A+ rated, book free of AOCI distortion). This is a high-quality compounder at a fair-to-slightly-cheap price with the entire cycle turning at once: a HOLD you accumulate on weakness, not a table-pounding buy at $94. No recommendation or price target appears below this section (the labeled Claude’s Take above is the sole exception).

2. Business Overview

Arch Capital Group Ltd. (“Arch”) operates three reportable underwriting segments plus a corporate/“other” segment that houses non-underwriting affiliates and run-off. The business model is conceptually simple and economically powerful: collect premium, hold the resulting “float” until claims are paid, earn investment income on it, and — unlike most insurers — actually generate an underwriting profit on the premium itself. Arch’s defining cultural tenet, stated in essentially every annual letter since founder Constantine Iordanou rebuilt the company in 2001–02, is that it will write business only when the price clears its return hurdle, and will shrink when it does not. Revenue comes from two streams: net premiums earned across the three segments and net investment income on a ~$48B portfolio, supplemented by realized gains and equity in earnings of affiliates (chiefly Somers Re and Coface).

Segment economics, FY2025 (from the FY2025 10-K, $M):

Below the underwriting line, FY2025 added net investment income of $1,625M, net realized gains of ~$464M, and equity in earnings of affiliates of ~$504M, for pretax income of $4,979M and net income available to common of $4,359M.

Insurance segment (~46% of GPW). Primary and excess specialty P&C: casualty (the largest line), professional liability (D&O, E&O, employment practices), property and short-tail (energy, marine, aviation), programs, travel/accident & health, surety, and the construction/national-accounts business expanded by the 2024 Allianz US MidCorp & Entertainment (“MCE”) acquisition. Distributed through wholesale and retail brokers. This segment is the most competitive and the lowest-margin of the three (95.2% combined ratio in FY2025), but it is also the largest premium base and the part of Arch most exposed to the current property-rate softening and casualty social-inflation firming.

Reinsurance segment (~49% of GPW). Treaty and facultative reinsurance across property catastrophe, property non-cat, casualty, marine/aviation, credit & surety, agriculture, and other specialty lines, written globally out of Bermuda, London, and other hubs. This is Arch’s best-in-class engine — an 80.8% combined ratio in FY2025, a fourth consecutive sub-80% underlying quarter — and the business most levered to the reinsurance hard market that peaked at the January-2025 renewals. Management is now deliberately shrinking property-cat as risk-adjusted pricing falls (net premium written down ~6%), a textbook Marathon-style supply-discipline signal.

Mortgage segment (~6% of GPW, ~34% of underwriting profit). Private mortgage insurance (US, via Arch MI, built on the transformative 2016 ~$3.4B acquisition of AIG’s United Guaranty) plus international mortgage reinsurance and GSE credit-risk transfer. At a 14.6% combined ratio — a negative current loss ratio in 2025 because cure rates on prior delinquencies drove reserve releases — this segment is wildly profitable but cyclically peaking: US mortgage credit is about as benign as it gets (delinquency ~2.06%), the in-force book is barely growing, and the releases that flatter the result must fade.

Recurring vs. cyclical. Premium is contractual and renews annually (sticky but re-priced each cycle); investment income is the most recurring stream; underwriting margin is highly cyclical. The crucial structural point — developed in Section 6 — is that the headline 82.8% combined ratio is mortgage-flattered: stripping the mortgage segment, the P&C-only combined ratio is roughly 87.8%, good and in line with Chubb, but not the dominant figure the blended number implies.

Verdict: A diversified, high-return underwriting platform with three genuinely distinct profit engines; the mix disguises how much of the consolidated margin comes from the small, peaking mortgage book.

3. Industry Dynamics

Arch competes in three different industries, each with its own structure and its own position in the capital cycle — and, unusually, all three are softening at once in 2026.

Specialty / E&S P&C insurance. A large, fragmented, broker-intermediated market. Structurally mediocre-to-decent: low switching costs, no customer captivity, capital is mobile, and pricing is cyclical. The 2019–2023 hard market — driven by social inflation, reserve deficiencies at weaker carriers, and reinsurance cost pass-through — has matured into a bifurcated cycle: property and short-tail rates are now falling (double-digit declines in cat-exposed property and E&S property), while casualty and professional lines are still firming as carriers respond to adverse liability trends and “social inflation” (rising jury verdicts, litigation funding). The E&S channel, where Arch is strong, continues to take share from the admitted market — a secular tailwind — but the easy rate-driven margin expansion is over. Structural verdict: mid-cycle, decent not good.

Reinsurance. A more concentrated, more capital-intensive oligopoly (a handful of Bermuda/European/global players plus alternative capital). Reinsurance went through a violent hardening in 2023 after a decade of soft pricing and back-to-back catastrophe years; January-2023 and January-2024 renewals delivered the best property-cat terms in a generation. By the January-2025 renewals the cycle topped, and 2026 is the down-slope: property-cat risk-adjusted rates are falling (management’s framing: returns compressing from “the 30s toward the high teens”), retentions are stable to loosening, and alternative capital (ILS, catastrophe bonds, sidecars) is flooding back in — the classic Marathon late-cycle signal that high returns attract capital which then mean-reverts margins. Structural verdict: textbook cycle top; the most attractive of the three two years ago, now de-rating fastest.

Private mortgage insurance. A US oligopoly of six players (Arch MI, MGIC, Essent, Radian, National MI, plus runoff books), gated by GSE eligibility requirements (PMIERs capital standards) and underpinned by a structural need: borrowers with <20% down payments require MI for a GSE-conforming loan. This is the most defensible of the three industries — high regulatory barriers, rational pricing via proprietary risk-based engines (Arch’s “RateStar”), and persistent in-force premium. But it is mature and cyclically peaking: housing affordability is stretched, origination volumes are subdued at higher mortgage rates, the in-force book grows slowly, and current earnings reflect the most benign credit losses the cycle offers. A housing/employment downturn would raise delinquencies and reverse the reserve-release tailwind. Structural verdict: genuinely good oligopoly, but ex-growth and at peak credit.

Cross-cutting regulatory/structural factors. Bermuda introduced a 15% corporate income tax (OECD Pillar Two-aligned) effective 2025, ending the territory’s historical zero-rate advantage; the one-time recognition of a deferred-tax asset created the ~$1.18B 2023 benefit, but the ongoing effect is a structurally higher cash tax rate that lowers Arch’s future net margin versus its own history. The reinsurance and mortgage businesses are capital-regulated (BMA, PMIERs); the group is rated A+ (Superior) by A.M. Best.

Overall verdict: Two decent-to-mediocre cyclical industries (specialty P&C, reinsurance) plus one genuinely good but ex-growth oligopoly (mortgage). Arch’s edge is less about being in great industries than about navigating mediocre cyclical ones better than anyone — which raises the question Section 4 must answer: is that a moat, or just management?

4. Competitive Position

The honest answer: Arch is an elite operator in largely commoditized markets, with one genuine but narrow structural moat (mortgage) and a real but reproducible-in-principle process advantage (cycle and capital discipline). It is not a wide-moat franchise in the Chubb/W.R. Berkley brand-and-distribution sense, and it is more key-person-dependent than either.

Applying the Greenwald taxonomy honestly:

• Customer captivity / switching costs: Largely absent in insurance and reinsurance. Brokers move books on price and terms; cedants re-tender annually. There is some habit/relationship stickiness, but nothing that would let Arch hold share at uncompetitive prices.
• Network effects: None.
• Proprietary cost / supply advantage: Partial and real in mortgage. Arch MI’s RateStar risk-based pricing engine and 25+ years of granular loss data give a genuine data/analytics edge in a six-player oligopoly where pricing precision compounds. In P&C/reinsurance the cost advantage is modest — Arch runs a low expense ratio and a lean, underwriter-led structure, but so do the best peers.
• Economies of scale + captivity: Real but narrow in mortgage (PMIERs capital scale + GSE relationships + data) and in the capital-fungibility of the group: because Arch can redeploy capital across three low-correlation engines, it can lean into whichever business is best-priced and starve the others, achieving a blended return that a single-line specialist cannot. That is a structural advantage of the holding-company design, not of any single business.

What proves something real exists. The clinching evidence that Arch’s discipline is more than a slogan is the behavior: the group grew gross premium from ~$18.4B (2023) to ~$22.9B (2025) while holding superior margins during the hard market — then, as pricing softened in 2026, it began deliberately shrinking (reinsurance net premium −6%, ~$250M of MCE business non-renewed) to protect returns rather than chase volume. Few insurers actually do this; most grow into soft markets and pay for it three years later in adverse development. Arch’s record across cycles — through-cycle ROE ranging from ~9.5% in the worst years to ~24% at the peak, ~17.1% in FY2025, and ~16–18% compound book-value-per-share growth over two decades — is the financial fingerprint of a genuine edge.

Direct comparison. Against Chubb, Arch matches the commercial combined ratio but lacks Chubb’s wide brand-and-distribution moat in high-net-worth personal lines and global commercial — Arch is narrower and more cyclical. Against W.R. Berkley, Arch is larger and more diversified (Berkley is a purer specialty operator with arguably an even more decentralized underwriting culture). Against RenaissanceRe / Everest in reinsurance, Arch carries less catastrophe volatility because of its diversification. Against MGIC / Essent in mortgage, Arch MI is a co-leader, advantaged by being part of a diversified group (it can hold MI capital through cycles that would strain a monoline). The single most important caveat: the moat is ~80% demonstrated management quality and ~20% structure, and the architects matter — founder Iordanou, then Grandisson, now Papadopoulo and a recently-consolidated single-President structure under Maamoun Rajeh. The June-2026 reorg modestly raised key-person concentration.

Verdict: A durable advantage exists — proven in the financials — but it is a process-and-design moat (discipline + capital fungibility + a narrow mortgage data/scale edge), not an unassailable franchise. Worth owning; not to be mistaken for a brand monopoly. If the culture degrades, the moat degrades with it.

5. Growth History and Forward Opportunities

History. Arch’s growth has been the right kind — driven by leaning into hard markets and retreating from soft ones, not by acquisitive empire-building or chasing premium. The two transformational moves were the 2016 ~$3.4B acquisition of United Guaranty (creating the mortgage segment at the housing-credit trough — a franchise-defining home run) and the organic build-out of reinsurance and specialty insurance. Gross premium roughly tripled over the past decade; the hard-market surge took GPW from ~$18.4B in 2023 to ~$22.9B in 2025. Crucially, book value per share — the metric Arch manages to — compounded ~16–18% annually, and net premium written grew while combined ratios improved, the signature of quality growth.

Composition. Predominantly organic, cycle-driven growth (rate × exposure × disciplined new business), supplemented by opportunistic bolt-ons (Barbican at Lloyd’s; the 2024 MCE/Allianz mid-corp book). The MCE deal added ~$1.5–2B of annualized US middle-market commercial and entertainment premium and broadened distribution, but Arch is already pruning the acquired book (~$250M non-renewed in 2026) where it doesn’t clear the return hurdle — discipline applied even to freshly-bought premium.

Forward opportunities.

• E&S share gains: the secular migration of risk from the admitted to the excess & surplus market continues to favor Arch’s specialty platform.
• Casualty firming: as property softens, casualty and professional lines are still hardening; Arch can rotate capital toward the firming lines (the capital-fungibility advantage in action).
• International mortgage / GSE credit-risk transfer: a longer-runway extension of the mortgage franchise beyond the mature US PMI market.
• Capital return as a “growth” lever: with the soft cycle reducing attractive underwriting deployment, Arch is redirecting capital to buybacks (~$1.9B in 2025, +$783M in Q1-2026) — book-value-per-share accretion substitutes for premium growth.

The honest counter-point: much of the reported growth of the last three years was price (rate), not exposure, and that tailwind is reversing. With all three engines softening, 2026–27 premium growth will be flat-to-down by design — management would rather shrink profitably than grow unprofitably. Growth from here is therefore book-value compounding (retained underwriting profit + investment income + buybacks), not top-line expansion.

Verdict: A high-quality, organic, cycle-disciplined growth record — but the top-line growth engine is now in its down-phase, and forward “growth” is really compounding at a decelerating ROE. High-quality, but no longer fast.

6. Financial Quality

This is the heart of the thesis. Arch’s financials are genuinely strong, but the headline numbers are at a cyclical peak and flattered in several ways that a 7× P/E does not advertise.

Five-year trend (reconciled to 10-K/10-Q; $M except per-share):

(1) GAAP vs. operating earnings. The ~7× trailing P/E sits on GAAP TTM EPS of ~$13, which includes ~$464M of net realized investment gains and ~$504M of equity-method affiliate income that Arch itself excludes from “operating” earnings. On Arch’s operating definition, TTM operating EPS is ~$10.8 (operating P/E ~8.7×). The GAAP/operating gap flips sign by year: GAAP understated earnings in 2022 (unrealized bond losses), overstated them in 2023 by the one-time ~$1.18B Bermuda deferred-tax-asset benefit (~$3.10/share), and runs above operating in 2024–25 on market gains. The signal in the operating series is unambiguous: operating ROAE is decelerating — 21.6% (2023) → 18.9% (2024) → 17.1% (2025) → 15.4% (Q1-2026). The cycle is rolling.

(2) Reserve-release dependence. FY2025 underwriting income benefited from roughly $600M of net favorable prior-year reserve development (~12% of pretax income, ~$1.36/share of operating earnings), led by the mortgage segment ($235M release; its current accident-year loss ratio was effectively negative, producing the 14.6% combined ratio). Q1-2026 carried ~$200M of favorable development (~5 points of combined ratio). Reserve releases are a legitimate but finite and pro-cyclical earnings source — they flatter the result in good years and reverse in bad ones. The quality question is how much current-accident-year margin remains once releases normalize; on a current-year basis Arch’s combined ratio is several points higher than reported.

(3) The mortgage-flattered combined ratio. Mortgage is ~6% of premium but ~34% of underwriting profit at an ~85% margin. The consolidated 82.8% combined ratio is therefore not representative of the P&C franchise: stripping mortgage, the P&C-only combined ratio is ~87.8% — still good and broadly in line with Chubb, but ordinary rather than exceptional. Investors anchoring on “82.8% — best in class” are double-counting the cyclically-peaking mortgage book.

(4) Book value — clean, and the number to trust. Unlike the life insurers (MET, AIG) where a 2022 rate shock crushed GAAP book through AOCI and created a “richest-ever P/B” mirage, Arch’s book is clean. AOCI swung from −$1.65B in 2022 to roughly zero by 2025 because the fixed-income portfolio is short (~3.34-year duration); the rate-shock losses pulled to par as bonds matured. There is no AOCI distortion in Arch’s 1.40× P/B — it is 1.40× real, recovered book. Book value per common share compounded ~16–18% annually over the long run; even at a decelerating ROE, retained earnings keep book growing at low-double-digits.

(5) Investment portfolio & income. A ~$48B portfolio, high-quality and short-duration, generated NII that tripled from ~$546M (2021) to $1,625M (2025) as reinvestment yields rose — a powerful but now-cresting tailwind, since further Fed easing lowers reinvestment yields at the margin. Equity in affiliates (Somers Re, Coface) adds ~$500M but is lower-quality, more volatile income.

(6) Balance sheet & cash conversion. Fortress-grade: debt/total-capital ~10.1%, total capital ~$26.9B, A+ (Superior) ratings, and operating cash flow of ~$6.2B in FY2025 (~1.40× net income, >1.0× every year) — clean cash conversion, no divergence of earnings from cash. The MCE acquisition used standard purchase accounting ($276M goodwill, no one-time gain), and there have been no goodwill impairments.

Normalized view. Putting it together: four tailwinds are peaking simultaneously (softening reinsurance/property pricing, fading reserve releases, peak-benign mortgage credit, cresting investment yields). A reasonable through-cycle operating EPS is ~$8.0–8.8 on a 13–15% ROE — meaningfully below the ~$10.8 TTM print. That makes the trailing multiple “peak-earnings cheap,” and shifts the analysis onto price-to-book vs. normalized ROE.

Verdict: Economics are genuinely high-quality and the balance sheet is pristine, but reported earnings are at a cyclical high, flattered by releases, mortgage mix, and rate-boosted NII. Do the economics improve with scale? Yes, modestly (expense leverage, capital fungibility) — but the swing factor is the cycle, not scale, and the cycle is turning down.

7. Capital Allocation

Capital allocation is Arch’s strongest non-underwriting attribute and a key reason the franchise compounds — it is run like an investor’s company, not a premium-maximizer’s.

Share repurchases — the Marathon tell. Arch’s buyback cadence is genuinely counter-cyclical, the behavior the Capital Returns framework prizes (repurchases by year, $M): 2021: 1,234 · 2022: 586 · 2023: ~0 · 2024: 24 · 2025: 1,889 · Q1-2026: +783. The pattern maps perfectly to opportunity cost: Arch bought back stock when it was cheap (2021), then paused to ~$0 in 2023–24 to redeploy every available dollar into the once-in-a-decade reinsurance hard market and to fund the MCE acquisition, then resumed aggressively in 2025–26 as the softening market reduced attractive underwriting deployment and capital piled up. That is exactly the discipline most insurers lack. The one demerit: 2025–26 repurchases ran at ~$94 average vs. ~$65 book — i.e. ~1.4× book, ordinary capital-return rather than the value-accretive sub-book buybacks Arch executed in 2018–20. Buying at 1.4× book is fine but not the arbitrage of the past.

No common dividend. Arch has never paid a common dividend (only ~$40M/year of preferred dividends) — a deliberate, tax-efficient, reinvestment-first policy that has served holders far better than a payout would have, given the ~16–18% book compounding. (Note: ROIC.ai’s data feed mislabels a 2024 “$1.9B dividend” — this is a buyback misclassification, not a payout.)

M&A scorecard — above-average to excellent. United Guaranty (~$3.4B, 2016, from AIG at the housing-credit trough) created the entire mortgage engine — now ~6% of premium and ~34% of underwriting profit — and stands as one of the best insurance acquisitions of the decade. MCE/Allianz (closed 1 Aug 2024): $450M cash for ~$174M of net assets ($276M goodwill), with ~$852M of net cash received at close (reserve-heavy book) — a modest bolt-on already being pruned where it doesn’t clear the hurdle. Watford → Somers Re (a capital-light reinsurance sidecar) monetized third-party capital; Arch trimmed its stake in January 2026. No impairments, no overpayment, no integration disasters.

Insider activity — neutral-to-mildly-positive. The Form 4 corpus is dominated by routine grants, option exercises, and tax withholding (codes A/F/M/G) — no discretionary dumping. The one notable conviction signal: director Daniel Houston (former CEO of Principal Financial) bought 5,300 shares at $94.085 (~$499K) on 30 April 2026 — a single, high-credibility open-market purchase, not a cluster. Non-executive Chairman John Pasquesi (a founding investor via Otter Capital) holds several million aligned shares. Directors and officers as a group own ~3.3% (~11.6M shares) — meaningful skin in the game.

Compensation — well-aligned. The DEF 14A ties long-term incentives to absolute three-year growth in tangible book value per share (target ~11%/yr) plus a relative-TSR modifier, and states comp focus is “primarily on growth in book and tangible book value per share.” This is the correct governor — it rewards per-share value creation and return on capital, not premium growth or size. The CEO’s pay is ~77% performance-based. Minor ding: 2025 say-on-pay support fell to ~84% on one-time 2024 CEO-transition grants.

Verdict: Intelligent, disciplined, return-focused capital allocation — counter-cyclical buybacks, no dividend dogma, accretive M&A, a book-value-per-share comp metric, and aligned insiders. This is the part of the thesis that is not in doubt. The only quibble is that current buybacks at 1.4× book are value-neutral rather than value-additive.

8. Changes and Headwinds — Last Two Years

Leadership transition (completed, orderly). Long-time CEO Marc Grandisson retired outright — ceasing as CEO and leaving the board — effective 15 October 2024; Nicolas Papadopoulo, a 23-year Arch underwriting insider, became CEO effective 13 October 2024 (he did not become executive chairman; the roles separated cleanly). In a further step, a 3 June 2026 reorganization consolidated segment leadership under a single President, Maamoun Rajeh, who now runs Insurance, Reinsurance and Mortgage, following the departure of François Gansberg. Net: an orderly internal succession that preserves the culture, but a modest increase in key-person concentration.

The cycle turn (the dominant headwind). Reinsurance pricing peaked at the January-2025 renewals and is now in its down-phase; property and short-tail P&C rates are falling; alternative capital is returning. Management’s response is the correct one — deliberately shrinking (reinsurance net premium −6%, MCE pruning) — but the consequence is decelerating ROE (18.9% → 17.1% → 15.4%) and flat-to-down premium. This is a cyclical, not structural, headwind, but it is the single biggest driver of the stock’s plateau.

MCE/Allianz integration. Closed Aug-2024; integrating a US mid-corp commercial book into the Insurance segment, with disciplined pruning underway. No adverse surprises disclosed to date, but a reserve-heavy acquired book always carries development risk.

Bermuda corporate income tax. The new 15% Bermuda CIT (effective 2025) structurally raises Arch’s cash tax rate versus its zero-tax history; the 2023 one-time DTA benefit is in the past, leaving a permanently higher go-forward tax drag on net margin.

Capital-return acceleration. Buyback authorizations were re-upped by +$2.0B (Sep-2025) and +$3.0B (Apr-2026), signaling management’s view that returning capital beats deploying it into a softening market — both a positive (discipline) and a tacit admission that underwriting opportunity is shrinking.

Reserve and social-inflation watch. Casualty/liability reserves across the industry face social-inflation pressure (rising verdicts, litigation funding); Arch’s heavy favorable development to date is a strength, but the casualty book is where any future adverse surprise would emerge.

Verdict: The changes are net neutral-to-mildly-strengthening on governance and capital allocation (orderly succession, accelerated buybacks, intact discipline), but the operating environment has clearly turned — the headwind is cyclical earnings normalization across all three engines, not a franchise or management problem.

9. Risk Analysis (Risk Matrix)

Net risk read: No existential or balance-sheet risk; the dominant, clustered risks are all cyclical earnings risks (pricing, releases, mortgage credit, NII) that compound in a downturn — which is precisely why the market caps the multiple at 1.40× book despite a 17% trailing ROE.

10. Valuation Discussion (Embedded Expectations)

The wrong lens and the right lens. The trailing GAAP P/E of ~7.2× (7.9th percentile of Arch’s own ten-year range) is the wrong anchor: it is flattered by ~$600M of reserve releases, the residual of the 2023 Bermuda DTA, and peak-cycle underwriting margins. The right lens for an insurer is price-to-book against through-cycle ROE (the justified-P/B framework), cross-checked by a normalized P/E. On book, Arch trades at 1.40×, the 49.8th percentile of its own 2015–2025 range (~1.1×–1.6×) — mid-range, not cheap.

Peer comparison (approximate, current):

The striking fact: Arch trades at a P/B discount to Chubb, Travelers and W.R. Berkley despite a comparable-or-higher current ROE. Bulls read this as a mispricing; bears read it as the market correctly discounting the durability of Arch’s ROE (more reinsurance-cyclical, more reserve-flattered) versus Chubb’s steadier mid-teens.

Embedded expectations. Back-solving the justified-P/B identity, P/B = (ROE − g) / (COE − g): at a cost of equity of ~9.5% and long-run growth ~4–5%, a 1.40× multiple implies a sustainable through-cycle ROE of only ~11.5–12% — below the current 17% and below our 13–15% normalized estimate. For reference, a durable 13% ROE justifies ~1.6× book; 14% justifies ~1.8–2.0×. So the market is pricing ROE to decay to roughly cost-of-capital-plus, with no credit for durability above ~12% and no re-rating optionality. That is the crux: the entire valuation debate collapses into one contested number — the normalized ROE.

Scenario analysis (value zones, not targets; explicit assumptions):

• Bear (~$78–88): All three engines soften faster than expected, reserve releases fade, mortgage credit normalizes; normalized operating EPS ~$8.0, ROE → 11–12%, book compounds ~7–8%/yr; multiple de-rates to ~1.15–1.25× book. The market’s implied case.
• Base (~$95–110): Discipline holds normalized ROE at ~13–14%, operating EPS ~$8.8–9.5, book compounds ~10–12%/yr; multiple stays ~1.40× book. Total return ≈ book-value growth, no re-rate. Most likely.
• Bull (~$120–140): Mortgage stays benign and casualty firming sustains a 15–17% ROE longer than feared, and the market re-rates Arch toward the 1.7–2.0× book its quality and peer set arguably warrant. Requires both durability and multiple expansion.

Embedded-expectations conclusion: At 1.40× a clean book on a franchise that has earned 15–24% ROEs across the cycle and compounded book at 16–18%, the market is arguably pricing peak-normalization slightly too pessimistically — conditional on the through-cycle ROE genuinely being 13–15% rather than 11–12%. The reward for being right is modest re-rating plus ~10% book compounding; the penalty for being wrong (ROE → 11–12%) is a fair-value stock that goes nowhere. Favorable asymmetry, thin margin of safety. No price target, no recommendation.

11. Variant Perception

Consensus. Sell-side and the market broadly regard Arch as a best-in-class, well-managed specialty (re)insurer whose earnings are at/near a cyclical peak and whose stock is therefore fairly priced — “great company, fine but not compelling price.” The ~7× P/E is widely understood to be peak-earnings, and the stock’s 20-month plateau reflects a market waiting for the cycle to play out. Consensus price targets cluster modestly above the current price, implying low-single-digit upside — i.e., a consensus “hold-ish.”

The strongest bull case. Arch is a category-killer allocator whose 16–18% book-value-per-share CAGR is the only number that matters over time, and you can buy that compounding machine at 1.40× a clean book — a discount to lower-ROE peers — with three diversified engines, a genuine mortgage moat, fortress capital, and management that shrinks into soft markets. The reserve-release and mortgage-credit “flatters” are themselves evidence of conservative reserving and disciplined underwriting that recur. If the through-cycle ROE is 14–15% (its long-run average), the stock is meaningfully cheap and should re-rate toward 1.7–2.0× book as the market gives Arch the multiple its quality warrants. Low beta (0.33) and out-of-favor positioning mean you’re paid to wait.

The strongest bear case. The “cheap” is an illusion of peak earnings. Operating ROE is already decelerating hard (21.6% → 17.1% → 15.4%), the reinsurance engine that drove the upcycle is now in its down-phase with capital flooding back in, the consolidated combined ratio is mortgage-flattered (P&C-only ~87.8%, ordinary), and ~$600M/yr of reserve releases that can’t recur are propping up the result. Normalize all of it and ROE settles at 11–13%, book growth slows to high-single-digits, and 1.40× book is exactly fair — there is no mispricing, just a high-quality company correctly priced for a softening cycle. The peer “discount” to Chubb/Travelers is deserved because Arch’s ROE is less durable and more cyclical. You make ~book-growth and get no re-rate.

The 3–5 assumptions that matter most:

1. Through-cycle ROE: 13–15% (bull) vs. 11–12% (bear). Everything hinges here.
2. Reserve-release durability — are the releases conservative-reserving recurring, or a finite cushion about to fade (or reverse in casualty)?
3. Mortgage credit — does the benign environment persist, or does a housing/employment downturn reverse releases into losses?
4. Reinsurance soft-cycle depth — a shallow, orderly softening (margins stay attractive) vs. a 2017-style race to the bottom.
5. Multiple re-rating — will the market ever award Arch a CB/TRV-style 1.7–2.0× book, or is 1.1–1.6× its structural range?

Falsification tests. Falsifies the bear: two-to-three quarters of operating ROE holding ≥15% with low-teens book-value-per-share growth as pricing softens — proof the franchise sustains elite returns through the down-cycle → re-rate. Falsifies the bull: operating ROE breaking below ~13% accompanied by adverse prior-year development and shrinking premium — proof Arch is a mid-teens-ROE-at-best, fully-priced name → de-rate toward 1.2× book. Both falsifiers key off the same two observables: realized operating ROE and the sign/size of reserve development.

Factor/positioning read. The tape supports “out-of-favor quality-value, not a falling knife”: beta 0.33, LowVolatility +0.86, Value +0.19, Quality +0.06, only mild Momentum (+0.27), Growth −0.15; rs_12m +3.7, ~14% off the high, +5.7% over the past year, max drawdown a shallow −14%. Factor-similar names are Loews, Chubb, Everest and Kinsale plus the insurance ETFs — a low-volatility quality-value insurer that has plateaued, the kind of name that re-rates on evidence rather than momentum, and that rarely offers a violent entry. Where consensus may be offsides: the market’s implied ~11.5–12% ROE looks too pessimistic for a franchise of this quality — if the cycle softens in an orderly way.

12. Fact vs. Interpretation Table

13. Open Questions

1. What is the true normalized through-cycle ROE? The whole valuation hinges on 11–12% vs. 13–15%, and only multiple quarters of soft-cycle data will resolve it.
2. How much reserve cushion remains, and where? The magnitude and line-of-business mix of remaining redundancy (especially casualty vs. mortgage) determines how long releases can support earnings — and whether casualty turns adverse.
3. How deep and disorderly will the reinsurance soft market be? An orderly fade (margins stay attractive) vs. a 2014–2017-style multi-year grind are very different outcomes for the best engine.
4. Mortgage credit at the next downturn: how much of the segment’s profit reverses if unemployment rises and home-price appreciation stalls?
5. BVPS definition reconciliation: common book value per share (~$66) vs. some data feeds showing ~$73 — confirm preferred/temporary-equity treatment for precise P/B (immaterial to the thesis, ~1.40× either way once defined consistently).
6. Key-person risk post-reorg: does the single-President structure under Rajeh and the new CEO preserve the underwriting culture that is the moat?
7. Capital-return ceiling: at what point does buying back stock at ~1.4× book become a less attractive use of capital than holding dry powder for the next hard market?

14. What Must Be True

For the bull case (Arch is modestly cheap; re-rate toward 1.6–2.0× book):

• Through-cycle operating ROE must hold 13–15%, not decay to 11–12%.
• Reserve development must stay favorable (or at worst neutral) — no casualty social-inflation surprise.
• Mortgage credit must stay benign through the cycle; book-value-per-share must keep compounding low-double-digits.
• The market must eventually award Arch a CB/TRV-style multiple (1.7–2.0× book).
• Falsification test: if operating ROE prints below ~13% for two-to-three consecutive quarters with adverse prior-year development and shrinking premium, the bull thesis is dead — Arch is a fully-priced mid-teens-ROE name.

For the bear case (1.40× book is fair-to-full; dead money or de-rate):

• Synchronized softening across reinsurance, specialty P&C and mortgage must compress realized ROE to 11–12%.
• Reserve releases must fade (or reverse), exposing an ordinary P&C-only combined ratio (~88%).
• NII run-rate must roll over as the Fed eases.
• Falsification test: if operating ROE holds ≥15% for two-to-three quarters while pricing softens and book-value-per-share grows low-teens, the bear thesis is dead — Arch is sustaining elite through-cycle returns and is too cheap at 1.40× book.

Both falsifiers key off the same two observables: realized operating ROE and the sign/size of reserve development. Watch those two numbers each quarter; they settle the debate.

The analysis above (Sections 1–14) is presented without an investment recommendation or price target. The sole, clearly-labeled exception is the opinion block at the top, which is the author’s own independent view.

15. Source Appendix

See Appendix B — Source Appendix below for the full evidence trail with URLs and access dates, and Appendix A — Diligence Questionnaire for the standard diligence question set.

APPENDIX A — Standard Diligence Questionnaire

Arch Capital Group Ltd. (NASDAQ: ACGL) — supplemental to the research memo (not counted toward the memo length standard). Report date 2026-06-26. Answers grounded in the research log; Fact / Interpretation / Assumption labeled where it matters.

General

What thoughtful questions have other investors asked about this company? The dominant questions: (1) Is the ~7× P/E “cheap” or is it peak-earnings? (2) What is the normalized through-cycle ROE once the hard market fades and reserve releases normalize? (3) How much of the consolidated combined ratio is mortgage-flattered? (4) Will the market ever award Arch a Chubb/Travelers-style book multiple, or is its 1.1–1.6× range structural? (5) Does the post-Grandisson leadership preserve the underwriting culture? (6) How exposed is the mortgage book to a housing downturn?

Cyclicality & Earnings Nature

Are earnings at a cyclical high or low? Fact/Interpretation: A cyclical high. Operating ROE has decelerated 21.6%(2023) → 18.9%(2024) → 17.1%(2025) → 15.4%(Q1-2026); reinsurance pricing peaked at Jan-2025 renewals; mortgage credit is peak-benign; ~$600M/yr favorable reserve development and rate-boosted NII both flatter the result.

Driven by external environment or internal actions? Both, but the recent earnings level is mostly external (hard market + benign credit + high rates). The relative outperformance (disciplined underwriting, capital fungibility, no soft-market chasing) is internal and durable.

How stable are revenues? Premium is contractual and renews annually — relatively stable in volume but re-priced each cycle. By design, premium will be flat-to-down in 2026–27 as Arch shrinks into the soft market. Net investment income is the most recurring stream.

Outlook for products/services? How big is the market? Large, mature, cyclical: global specialty P&C, reinsurance (cycle top), and US private mortgage insurance (~$1.5T+ insurance-in-force industry, mature). E&S share-gain is the main secular tailwind. Growth from here is book-value compounding, not top-line expansion.

Business Quality & Competitive Moat

Is the industry getting more or less competitive? More — reinsurance and property pricing are softening with alternative capital returning (Marathon cycle top); casualty is still firming. Mortgage is a stable six-player oligopoly.

How profitable is the business (ROIC/ROE)? Very — operating ROE 15–17% (peak 21.6%), ~16–18% long-run book-value-per-share CAGR. Through-cycle normalized ROE estimated 13–15% (Interpretation).

How profitable is the industry — competitors, barriers? P&C/reinsurance: decent but cyclical, low barriers, capital-mobile. Mortgage: genuinely good — high regulatory barriers (PMIERs, GSE eligibility), six players, rational risk-based pricing.

Can the business be easily understood? Moderately. The three-segment structure, float economics, and combined-ratio math are learnable, but reserve adequacy and cycle timing require judgment.

Undermined by foreign low-cost labor? No — capital-and-underwriting business, not labor-cost-exposed.

Do brands matter? Nature of competition? Switching costs? Brand matters modestly (ratings and relationships more than consumer brand); competition is on price/terms/capacity; switching costs are low in P&C/reinsurance (annual re-tender) and somewhat higher in mortgage (in-force persistency).

Financial Condition & Balance Sheet

Assets not fully recognized on the balance sheet? The franchise/underwriting-culture value and the conservatively-reserved redundancy (a source of favorable development) are not balance-sheet items. Interpretation.

Off-balance-sheet liabilities? None material disclosed; reinsurance recoverables and reserve adequacy are the key estimation risks (on-balance-sheet but judgment-laden).

How conservative is the accounting? Conservative — persistent favorable reserve development, short asset duration, clean cash conversion (OCF ~1.40× NI), high-quality investment portfolio. Watch: casualty reserve adequacy is the perennial industry risk.

How CapEx-hungry? Capital-light operationally (no heavy physical capex); the “capex” is underwriting capital deployed into risk — which Arch allocates counter-cyclically.

Capital Allocation & Management

How much FCF, and how is it used? Insurer “FCF” ≈ operating cash flow plus released capital. OCF ~$6.2B FY2025. Used for: underwriting deployment when priced well, then share buybacks (~$1.9B 2025, +$783M Q1-26) when not; no common dividend ever.

Significant acquisitions recently? MCE/Allianz mid-corp book (closed Aug-2024, $450M, $276M goodwill, being pruned). Historically: United Guaranty ($3.4B, 2016 — franchise-defining home run), Barbican, Watford/Somers.

Buying back shares? Yes — counter-cyclically (paused ~$0 in 2023–24 to fund the hard market/MCE, resumed aggressively 2025–26 at ~1.4× book). Authorizations re-upped +$2.0B (Sep-2025), +$3.0B (Apr-2026).

Issuing shares to insiders? Routine equity comp only; no excessive issuance. Net share count is being reduced via buybacks.

Compensation policy? Fact: LTI tied to 3-year growth in tangible book value per share + relative-TSR modifier (~11%/yr target); ~77% of CEO pay performance-based; explicitly not premium/size-driven. A genuine return-on-capital governor. Insiders own ~3.3%.

Motivations of management? Per-share value creation and underwriting profit, aligned by the TBVPS comp metric and meaningful insider ownership (Chair Pasquesi a founding investor).

Valuation & Market Data

ADR, MLP, or K-1 issuer? None — Bermuda-domiciled common stock filing 10-K/10-Q with the SEC; not an ADR, not a K-1. (US holders should note potential PFIC considerations for non-US insurers, though Arch’s active-insurance status generally addresses this; consult a tax advisor — Assumption.)

Dividend policy? No common dividend (deliberate reinvestment-first policy); preferred dividends only (~$40M/yr).

How profitable is the business? Highly — see ROE/combined-ratio above.

Net income diverging from cash from operations? No — OCF ~1.40× net income, >1.0× every year; clean conversion.

Risks & Downside

What would cause the stock to decline? Operating ROE breaking below ~13%; adverse casualty reserve development; a mortgage-credit downturn reversing releases into losses; a major catastrophe year; faster-than-expected reinsurance/property softening; NII roll-over as rates fall; multiple de-rating below 1.4× book.

Risk of catastrophic loss? Low — diversified, A±rated, ~10% debt/capital, short-duration high-quality assets. A single-year cat or credit event compresses earnings but does not threaten solvency.

Chance of total loss? Very low — fortress balance sheet, diversified risk, conservative reserving.

Recent News & Events

Has the business environment changed recently? Yes — the pricing cycle has turned across all three engines (reinsurance topped Jan-2025; property rates falling; casualty firming; mortgage at peak credit). Management is responding by shrinking and returning capital.

Significant acquisitions? MCE/Allianz (Aug-2024), now being pruned; Somers stake trimmed (Jan-2026).

Change in accounting policies? Bermuda 15% corporate income tax effective 2025 (one-time ~$1.18B DTA benefit recognized 2023; structurally higher go-forward tax rate).

Recent changes — markets, facilities, management? CEO transition (Grandisson retired Oct-2024; Papadopoulo CEO); 3-June-2026 single-President reorg under Maamoun Rajeh (Gansberg departed); buyback authorizations re-upped twice; one director open-market purchase (D. Houston, Apr-2026).

APPENDIX B — Source Appendix

Arch Capital Group Ltd. (NASDAQ: ACGL) — Report date 2026-06-26. Public primary sources prioritized; all material facts traceable to filings.

Primary — SEC Filings (CIK 0000947484)

Primary — Quantitative Data Helpers

Peer Cross-Read (public company filings)

Key Figures Quick-Reference (all reconciled to filings)

• FY2025: GPW $22,878M; NPW $16,476M; NPE $17,065M; underwriting income $2,933M; consolidated combined ratio 82.8%; NII $1,625M; NI to common $4,359M; diluted EPS $11.60; operating ROAE 17.1%.
• Segments FY2025 combined ratios: Insurance 95.2% / Reinsurance 80.8% / Mortgage 14.6%. Mortgage ~6% of premium, ~34% of underwriting profit; P&C-only combined ratio ~87.8%.
• Operating EPS TTM ~$10.8; operating P/E ~8.7×. Normalized through-cycle EPS ~$8.0–8.8; ROE 13–15% (author estimate).
• Book value per common share ~$66 (Q1-2026); P/B ~1.40×; AOCI fully recovered (no distortion).
• Debt/total-capital ~10.1%; total capital ~$26.9B; OCF ~$6.2B (~1.40× NI); A+ (Superior) rated.
• Buybacks ($M): 2021: 1,234 · 2022: 586 · 2023: ~0 · 2024: 24 · 2025: 1,889 · Q1-26: 783. No common dividend ever.
• Price $94.33 (2026-06-25); market cap ~$33.2B; ~14.8% off ATH $110.75.