ACGL — Q2 2026 Financial Report Analysis

Catastrophes hit the insurance book, and buybacks did the work on EPS

Arch Capital earned $1.05 billion for common shareholders in the three months to 30 June 2026, down 14.7% from $1.23 billion a year earlier. The more telling number is underwriting income — the profit made purely from selling insurance, before anything the investment portfolio contributes — which fell 19.7% to $657 million. Two things drove that: catastrophe claims in the insurance segment more than doubled as a share of premium, and the amount of premium Arch is earning shrank.

Diluted earnings per share fell only 7.1%, to $3.00 from $3.23. The gap between the 14.7% drop in profit and the 7.1% drop in per-share profit is entirely share count: Arch bought back 12.4 million of its own shares for $1.2 billion during the quarter, cutting the average diluted share count from 379.9 million to 348.8 million. Management states in the filing that across the first half it "repurchased approximately 94% of our net income in our own shares."

Headline numbers

Two pieces of jargon worth unpacking, because the rest of this analysis leans on them:

• Combined ratio is the share of each premium dollar consumed by claims and expenses. Below 100% means the insurance business itself made money; above 100% means it lost money and any profit has to come from investments. Arch's 83.5% means it kept about 16.5 cents of underwriting profit per premium dollar — still good, but 2.3 cents less than a year ago.
• Net premiums written vs. earned. Written premium is business signed this quarter; earned premium is the slice of past and present contracts recognized as revenue this quarter, spread over the policy term (usually 12 months). Written premium moves first, and earned premium follows it with a lag of about five quarters. Arch's written premium is down 6.9%, which means earned premium — and the underwriting profit calculated on it — keeps falling for several quarters even if nothing else changes.

Where the profit came from, and where it stopped

The insurance segment is the problem. It wrote 48% of the group's net premium and produced $27 million of underwriting income — 4% of the total. A 98.5% combined ratio means the segment kept roughly one and a half cents per premium dollar. Reinsurance and mortgage, on a combined $2.1 billion of premium, produced $630 million.

Catastrophes. Pre-tax current accident year catastrophe losses across insurance and reinsurance, after reinsurance recoveries and reinstatement premiums, were $201 million. In the insurance segment those losses added 7.6 points to the loss ratio versus 2.9 points a year ago, which the filing attributes primarily to the Iran conflict and severe convective storms (hail, tornado and straight-line wind events) in the U.S. That 4.7-point swing on $1.88 billion of earned premium is roughly $88 million of lost margin, and on its own more than explains why insurance underwriting income fell $102 million. Reinsurance moved the other way: 3.0 points of catastrophe load versus 5.5 points a year ago.

Reserve development. When an insurer finds that claims from earlier years are settling for less than it set aside, it releases the surplus into current profit — "favorable prior-year development." All three segments released reserves, and the releases got bigger where the business is going well: reinsurance released $97 million (5.3 points of its loss ratio, versus $81 million and 3.9 points last year), insurance released $27 million (1.4 points versus 0.4), and mortgage released $45 million (15.7 points versus 22.8). The detail matters — reinsurance's release came from property business of the 2024 and 2025 underwriting years, partly offset by $26 million of adverse development in casualty from 2022–2023. Insurance's short-tail property release of $22 million was likewise offset by adverse development in programs business from 2021–2023. Casualty and programs reserves are the line items to watch; they are moving the wrong way in both segments.

Underlying margin, stripped of both. The combined ratio excluding catastrophes and prior-year reserve moves — the cleanest read on this quarter's pricing and claims experience — worsened from 80.9% to 82.5%. It worsened in both underwriting segments: insurance 90.6% to 91.6%, reinsurance 77.2% to 79.9%. So the deterioration is not purely a weather story.

Why premium is shrinking

The top line is falling for three distinguishable reasons, and only one of them is a demand problem.

In insurance, gross premiums written fell 2.9% and net fell 5.1% — but the filing notes that "adjusting for the non-renewal of certain programs related to the MCE Acquisition, net premiums written would have decreased by 1.8%." Arch bought Allianz's U.S. MidCorp and Entertainment businesses in August 2024 and is now walking away from parts of that book. Most of the reported decline is deliberate pruning.

In reinsurance, gross premiums written were actually up 0.2% — flat — while net fell 10.4%. The difference is retrocession: premium Arch pays to other reinsurers to offload risk, which rose from $1,137 million to $1,358 million. The filing cites "non-renewals, share reductions as well as targeted increased retrocessions." Arch is buying more protection at what it evidently judges to be attractive terms, which lowers both expected losses and expected profit. That choice shows up favorably in the expense ratio — 22.9% versus 24.4%, "primarily reflecting the impact of higher profit commissions on retrocessions" — and unfavorably in the shrinking earned premium base.

The insurance segment's expense ratio went the other way, rising to 35.5% from 33.6%, which the filing attributes to transitional costs from the MCE acquisition and to spreading fixed costs over a smaller earned premium base. Note also that the prior-year comparison is flattered: MCE purchase accounting lowered the 2025 second-quarter expense ratio by about 0.6 points because deferred acquisition costs on the acquired book were not recognized at closing.

Mortgage was the only segment growing net premium, up 7.5% on flat gross premium (+0.3%) because Arch terminated certain Bellemeade Re and quota share agreements on U.S. primary business — again, retaining more of what it already writes rather than writing more.

Investments, tax and the GAAP-versus-operating gap

Net investment income rose 3.0% to $417 million. The composition is worth noting: the pre-tax investment income yield fell to 3.91% from 4.25% a year ago (and 3.99% last quarter), so the increase came from a larger portfolio, not better rates — the filing credits "growth in average invested assets, due in part to strong operating cash flows." On a per-share basis, though, investment income rose 12%, from $1.07 to $1.20, because there are fewer shares. The same buyback arithmetic that cushioned EPS is flattering this line.

Net income fell 14.7% while after-tax operating income — which strips out realized investment gains and losses, currency moves, equity-method income and transaction costs — fell only 8.8%, to $893 million. The divergence is almost entirely last year's comparison: Q2 2025 included $229 million of net realized gains, against a $17 million realized loss this quarter. Pulling the other way, two items helped reported GAAP profit and are excluded from operating profit: equity-method investment income of $196 million (up from $162 million) and a $10 million foreign-exchange gain against an $88 million loss a year ago, a $98 million swing. The effective tax rate also fell to 13.4% from 14.7%.

Arch issued $2.0 billion of senior notes in June ($600 million at 5.250% due 2036, $1.4 billion at 5.950% due 2056), used part of the proceeds to tender for older 2043 and 2046 notes — booking a $16 million pre-tax gain — and intends to retire its 4.011% notes maturing in 2026. Interest expense was already $44 million versus $38 million; it will be structurally higher from here, since the new coupons are roughly 150–200 basis points above the debt being retired.

Takeaway: Strip out the buyback and the picture is a company whose underwriting engine is getting smaller and slightly less profitable at the same time — earned premium down 8.1%, underlying combined ratio up 1.6 points, and an insurance segment at 98.5% that contributed just 4% of group underwriting income. Arch is converting that underwriting capital into shares rather than into premium, and at $1.2 billion bought above book value in the quarter, that trade dilutes book value per share even as it props up earnings per share.

Capital, book value and what to watch

Book value per share — total common equity divided by shares outstanding, the metric Arch itself calls "the key driver of Arch Capital's share price over time" — reached $68.04, up 2.8% in the quarter and 15.0% year over year from $59.17. The filing is explicit that the quarterly gain was "partially offset by $1.2 billion of shares purchased at an average price higher than the book value per share." Buying back stock above book value reduces book value per share; Arch did it anyway, at scale, which is a statement that management considers the shares cheap relative to earnings power rather than relative to accounting equity. Year to date Arch has repurchased 20.7 million shares for $1.9 billion, against $359.7 million in the same period of 2025, and had $2.2 billion of authorization left at quarter end after a $3.0 billion increase approved on 19 April 2026.

Arch, like most property-casualty insurers, does not publish earnings guidance, so there is no management forecast to evaluate. What the filing does disclose about forward risk: the company targets limiting its net probable maximum loss from a 1-in-250-year catastrophe in any single geographic zone to roughly 25% of tangible shareholders' equity — a ceiling, not a forecast, and the relevant one heading into the second half, which contains peak Atlantic hurricane season.

Our read on trajectory:

• Underwriting income should keep compressing into 2027 even on normal catastrophe experience. Earned premium follows written premium with roughly a five-quarter lag, and written premium is down 6.9%. With the underlying combined ratio also up 1.6 points, both terms of the underwriting-profit equation are moving against Arch at once.
• The insurance segment is the swing factor. At 98.5% it is one bad quarter from an underwriting loss. The pruning of MCE programs should help the margin as the weaker business runs off, but the transitional expense drag and the adverse development in programs and healthcare lines are unresolved.
• The earnings mix is getting more dependent on items that are finite or excluded from operating profit. Mortgage reserve releases fell to $45 million from $64 million and will shrink as the strong 2024–2025 accident years are exhausted; equity-method income and currency gains helped reported profit but are not underwriting.
• The offsets are real, though. A falling share count is compounding into every per-share line, investment income is growing on asset growth despite a lower yield, and increased retrocession buying should make catastrophe quarters less punishing than the 7.6-point insurance load just seen.

Sources: Arch Capital Group Ltd. Form 10-Q for the quarterly period ended 30 June 2026 (filed 4 August 2026, SEC accession 0000947484-26-000124) and the second-quarter 2026 earnings release furnished as Exhibit 99.1 to Form 8-K (filed 28 July 2026, SEC accession 0000947484-26-000118). All figures are as reported by the company; percentage changes are computed from those reported figures.