AIG — Q2 2026 Financial Report Analysis

Adjusted EPS rose 10% while GAAP EPS fell 10% — almost all of both moves came from the Corebridge exit and the buyback, not from better underwriting

AIG's second quarter of 2026 (three months to June 30, filed on Form 10-Q on August 7, accession 0000005272-26-000076; results released August 6) is a quarter where the two headline earnings numbers point in opposite directions, and neither one is mainly about insurance.

Net income attributable to common shareholders fell to $948 million ($1.78 per diluted share) from $1,144 million ($1.98). Adjusted after-tax income — AIG's own measure, which strips out investment mark-to-market swings, restructuring and divestiture items — rose to $1,069 million ($2.00 per diluted share) from $1,044 million ($1.81). The GAAP decline is the disappearance of a large prior-year investment gain. The adjusted increase is almost entirely the share count.

Two terms that carry most of the weight below. The combined ratio is the share of every premium dollar that goes out as claims plus expenses; below 100% means the insurance business itself made money before counting investment income, so AIG's 89.0% means roughly 11 cents of underwriting profit per premium dollar. Prior-year reserve development is the revision of money set aside in earlier years for claims not yet settled: if the old estimates turn out to have been too high, the release flows into this quarter's profit as "favorable" development, even though the business that produced it was written years ago.

Takeaway: Adjusted after-tax income grew 2.4% ($1,044m to $1,069m) but adjusted EPS grew 10.5%, because the diluted share count fell 7.7%. Roughly eight of the ten points of "growth" management led with are bought back stock, not earned income — and the underlying accident-year combined ratio improved only 0.3 points, with International Commercial's deteriorating 2.3 points.

The Corebridge exit closes, and it explains the GAAP decline

AIG finished disposing of its life and retirement business this quarter. On February 17, 2026 Corebridge Financial repurchased 24.7 million shares from AIG at $30.42 for $750 million; by March 23 AIG's last two board designees had resigned, and at March 31 AIG concluded it no longer had significant influence, moving the stake from equity-method accounting to a marked-to-market equity security. On May 7 it sold the remaining 25.5 million shares at $27.90 for roughly $710 million — a price 8% below where it sold in February, which is why the final tranche produced only a $103 million gain on sale (recorded inside net investment income and excluded from adjusted income).

The comparability effect is large and entirely mechanical. Changes in the fair value of equity securities and the Corebridge stake, plus gains and losses on the share sales, contributed $173 million of pre-tax income this quarter against $464 million a year ago. That $291 million pre-tax swing is the bulk of the $280 million decline in pre-tax income ($1,264m vs $1,544m). Total revenues were flat at $7,085 million only because $344 million of premium growth was almost exactly cancelled by a $339 million drop in net investment income, most of which is the same Corebridge and equity-mark effect.

Corebridge's absence also shows up in the run rate: Other Operations net investment income and other fell to $39 million from $92 million, a decline that includes $27 million of Corebridge dividends that no longer exist, with the rest from lower short-term investment income. Other Operations' adjusted pre-tax loss widened to $142 million from $101 million.

Separately, a $133 million net loss on divestitures for the first half (only $6 million of it in Q2) came from a change in estimate on earn-out consideration tied to the earlier sales of Validus Reinsurance and the global personal travel and assistance business — a reminder that disposals completed in prior years still carry tail exposure through contingent payments.

Growth is real, but a chunk of it is bought rather than written

Net premiums written of $7,516 million grew 9%, with no help from currency — reported and constant-dollar growth were both 9%. The composition matters. Two transactions closed on February 6, 2026: a 35% equity stake in Convex Group, a global specialty insurer, for $2.1 billion, and a 9.9% stake in Onex Corporation, an asset manager, for $642 million. Alongside the equity stake, a whole-account quota share reinsuring 7.5% of Convex's underwriting portfolio took effect January 1, 2026, stepping up to 10.0% in 2027 and 12.5% in 2028.

The 10-Q's segment commentary attributes premium growth in all three segments partly to "the impact of strategic transactions," which is this quota share. That is assumed business written to Convex's underwriting standards, not AIG's own new production, and the step-ups mean two more years of mechanical top-line growth from the same source. Organic growth is genuine underneath it — the filing names Casualty and Financial Lines in North America and Specialty and Property internationally — but headline premium growth and organic premium growth are no longer the same number, and AIG does not break out how much of the 9% is which.

Working the other way, North America Property shrank: management noted growth would have been 11% excluding North America Property lines, and the filing attributes the drag to lower production at Lexington, the excess and surplus lines platform, on Property. CEO Eric Andersen described a market that "has transitioned from an extended phase of broad positive pricing into a more selective environment, where profitability and growth are increasingly dependent on line-specific dynamics" — in plain terms, the broad post-2019 commercial rate increases have stopped, and property rates in particular are now falling enough that AIG is walking away from business.

The three segments moved in three directions

North America Commercial looks like the star and partly isn't. Its combined ratio improved 1.9 points, but the improvement came from 1.8 points of higher favorable prior-year development and 0.6 points of lower catastrophe losses — items that say nothing about how this year's business is priced. On the accident-year view that strips both out, the segment got worse by 0.5 points, with a 0.3-point higher loss ratio the filing attributes to "changes in business mix" and a 0.5-point higher acquisition ratio, also mix. The press release adds the honest version: rate pressure, "particularly in Property."

International Commercial is where the quarter's damage sits. Underwriting income fell a third to $200 million and the combined ratio rose 5.4 points to 91.3%. Catastrophes account for 2.5 points of that, but the accident-year ratio still deteriorated 2.3 points — 1.0 point of higher loss ratio on mix, 1.3 points of higher acquisition costs from "a combination of strong new business growth and changes in business mix," plus 0.6 points of adverse prior-year development driven by prior-year premium adjustments. Growing 11% while the accident-year combined ratio worsens 2.3 points is the pattern worth watching; the segment is buying share in a softening market.

Global Personal produced the cleanest improvement: the combined ratio fell 5.6 points and the accident-year ratio 4.9 points, so this is underlying, not luck. Two-plus points came from the expense side — a 0.9-point lower acquisition ratio reflecting renegotiated commission terms in the US high-net-worth business earning through, and a 1.3-point lower operating expense ratio — with the rest from a 2.7-point lower accident-year loss ratio on mix. Underwriting income went from $25 million to $114 million. Net premiums earned actually fell, on Warranty, so the profit came from margin, not volume.

Catastrophes: a man-made loss in a quarter with no big natural one

Total catastrophe charges were $210 million, or 3.4 points of the loss ratio, against $170 million (2.9 points) a year ago. The composition is unusual: $132 million from "flooding, rainstorms and other," which includes $75 million of net losses from the Middle East conflict; $46 million from windstorms and hail; $30 million from winter storms. There was no single large natural catastrophe. The $75 million war-related loss landed almost entirely in International Commercial, whose cat charge tripled to $89 million from $29 million, and is the main reason that segment's combined ratio deteriorated.

For context on the half: first-half catastrophe losses of $390 million were well below the prior year's $695 million, because the first half of 2025 carried $455 million of wildfire losses. That comparison is most of why AIG's first-half combined ratio improved 4.5 points to 88.1% and first-half underwriting income jumped 68% to $1,460 million — a flattering comp, not a step-change in underwriting.

Reserves: workers' compensation is paying for excess casualty

Favorable prior-year development, net of reinsurance and prior-year premiums, was $145 million on the ratio basis (2.5 points) versus $112 million (2.0 points). The reserve note's gross figure, which includes $26 million of amortization of the deferred gain on the National Indemnity adverse development cover, is $172 million against $128 million.

The line-by-line detail is the more useful read:

Workers' compensation releases of $177 million, from excess-of-loss sensitive and primary business, are doing the heavy lifting, alongside $79 million from property programs. Against that, US Excess Casualty strengthened by $74 million on "unfavorable experience in first-layer umbrella and high excess casualty" — adverse again, as it was in the year-ago quarter ($106 million) and across both first halves ($68 million adverse in 2026, $99 million in 2025). Net excess casualty reserves stand at $3.2 billion against $6.0 billion gross.

This is the standard US casualty problem — jury awards and settlement values on old liability claims rising faster than the reserves set for them — and AIG is currently absorbing it out of workers' compensation redundancy. That works as long as the comp releases keep coming. They are finite; the excess casualty deterioration has so far not stopped.

Investment income is quietly weaker than it looks

General Insurance net investment income was exactly flat at $871 million. That flat line contains something new: AIG's 35% share of Convex's net income, less amortization of purchase-accounting basis differences, now runs through General Insurance net investment income, and was not in the prior-year figure at all. Flat with a new earnings stream added means the pre-existing portfolio's income went down. The consolidated attribution in the 10-Q is consistent: excluding the Corebridge and equity marks, net investment income fell on lower income from alternative investments and mortgage loans, partly offset by higher income from available-for-sale bonds as maturing money is reinvested at higher yields. On an adjusted basis total net investment income fell 5% to $908 million, all of it in Other Operations.

Core operating return on equity — AIG's preferred profitability measure, which excludes the Corebridge stake and unrealized bond marks — fell to 11.1% from 11.7% even as adjusted EPS rose. The arithmetic is straightforward: core operating book value per share rose 17% to $74.43 from $63.71 as Corebridge proceeds converted into cash and other assets inside the core, while adjusted after-tax income grew 2%. AIG has raised the capital base faster than it has raised the earnings on it. Whether the Convex and Onex stakes eventually close that gap is the central question for the next two years.

Capital return is the earnings growth story

AIG returned $904 million in the quarter: $641 million of buybacks (about 8 million shares) and $263 million of dividends. For the first half, 15 million shares for $1.2 billion plus $504 million of dividends. The quarterly dividend rose 11% to $0.50 from $0.45, and the board declared another $0.50 on August 6, payable September 30. A further 2 million shares went for roughly $195 million in July, with $2.6 billion left under authorization as of July 31.

Shares outstanding fell to 524.7 million from 559.8 million, down 6.3% in a year. At the Q2 pace the remaining authorization is roughly four quarters of repurchase. Leverage is unstressed — total debt to total capital of 18.1%, or 17.6% on AIG's adjusted-capital basis — so the buyback is not being funded by the balance sheet. But it does mean the reported EPS growth rate is running well ahead of the income growth rate, and that gap closes once the pace of repurchase normalizes or the authorization requires refreshing.

Outlook

Management gave no numeric quarterly guidance, reiterating only confidence in "our 2025 Investor Day financial objectives" and describing the market as selective and line-specific rather than broadly hardening.

What the filing supports as a forward read:

• Premium growth is well underwritten in North America and less so internationally. The Convex quota share steps from 7.5% to 10.0% of Convex's book in 2027, so reported net premiums written will keep growing partly for structural reasons. Judge the franchise on accident-year combined ratios, not the top line; on that measure International Commercial is the segment to watch, having grown 11% while its accident-year ratio deteriorated 2.3 points.
• Third-quarter catastrophe exposure is untested. Q2's $210 million contained no major natural event and $75 million of man-made loss. Atlantic hurricane season falls in Q3; AIG's Q3 combined ratio has far more variance than the 89.0% just reported.
• Excess casualty is the reserve risk. Adverse development of $74 million in the quarter and $68 million year to date, against workers' compensation releases of $177 million and $185 million, means the net favorable number depends on the comp redundancy outlasting the casualty deterioration. If comp releases slow before excess casualty stabilizes, the reported combined ratio moves up by more than a point without anything changing in the current book.
• Core operating ROE has further to travel. At 11.1% and falling, with a 17% larger core equity base, AIG needs the Convex earnings contribution and the redeployed Corebridge proceeds to produce returns above the current portfolio yield, or the buyback has to keep shrinking the denominator. The first is a 2027–2028 question; the second has about four quarters of authorization behind it.

The underwriting franchise is performing acceptably — an 88.1% accident-year combined ratio with a 30.8% expense ratio is a good result in a softening market, and Global Personal's improvement is the clearest genuine margin gain in the quarter. But the reported earnings growth is being produced by the share count, and the two largest swing factors in the P&L this quarter, the Corebridge mark and the Middle East loss, were neither of them underwriting decisions.