After a decade of subpar combined ratios, Zaffino has shrunk AIG to a focused commercial P&C book. The economics still trail Chubb and Travelers, so the stock is only attractive when it trades near tangible book.

Plain English

AIG sells insurance to big companies — covering plane crashes, lawsuits, hurricanes, factory fires. Customers pay premiums up front; AIG promises to pay claims later, sometimes years later. AIG keeps the cash in between and earns interest on it. The trick is charging enough premium to cover claims plus expenses. AIG was bad at this for twenty years, took huge losses, and had to be bailed out in 2008. A new CEO has cleaned it up. The business is now okay — not great. Buy it only when the price is below the value of the company's net worth, because that is your safety cushion if claims come in worse than expected.

Thesis

American International Group is a global commercial property-and-casualty insurer that has spent the last decade dismantling itself: spinning Corebridge (Life and Retirement), reinsuring legacy long-tail liabilities into Fortitude Re, and selling or running off non-core books. What remains is a focused General Insurance franchise (North America Commercial, International Commercial, Global Specialty including Validus Re-related lines, and a smaller Personal Insurance business) plus a residual Corebridge stake being monetized. The business now writes roughly $25B of net premiums at a combined ratio that has finally moved into the low-90s — adequate, but still meaningfully behind Chubb (mid-80s) and Travelers (high-80s). The scorecard tells the story bluntly: composite 60/100, ROIC 10y avg 0.0%, FCF conversion 5y 0.0%, owner earnings TTM negative ($-1.92B), and the deterministic IV model returns negative ($-137 base) because GAAP free cash flow has been overwhelmed by capital-return outflows during the divestiture period. The 10-year share count is down 5.5%, but that buyback was funded by selling assets, not by compounding earnings. The honest way to value AIG is therefore book × normalized ROE, not DCF: at ~$78.77 the stock trades around 1.0x tangible book on roughly $80 of TBVPS, embedding a normalized ROE of ~10%. That is a fair price for a fair business, not a Buffett-style fat pitch. The only way to underwrite an attractive return is to buy materially below book — call it $60-65 — and let buybacks at sub-book prices compound TBVPS while Zaffino narrows the gap to peers. Above ~$95 you are paying for execution that has not yet been earned. Price/IV math: scorecard IV is uninformative (negative); peer-relative book × ROE pegs base-case fair value near $85 with a $60 buy line and $95 trim line.

Moat

AIG is a global commercial and specialty P&C insurer. The Buffett-Munger framework is unusually clear about what an insurance moat looks like, and AIG's relationship to that template is the central question.

1. Cost advantages. Buffett's archetype is GEICO: a structural low-cost producer that uses direct distribution to undercut everyone else, then compounds the gap as scale grows [3]. AIG is the opposite of GEICO. It writes large, complex commercial and specialty risks distributed almost entirely through brokers (Marsh, Aon, WTW), so it carries a meaningful expense load before a single claim is paid. Reinsurance and specialty have some scale economies in catastrophe modeling and balance-sheet capacity, but Berkshire, Munich Re, Swiss Re, Chubb and the Lloyd's market all have those too. There is no durable unit-cost edge here. Verdict: NONE.

2. Intangibles — brand, ratings, regulatory. This is AIG's best moat candidate. Writing $500M of D&O on a multinational, or aviation hull-and-liability, or political-risk on a frontier-market project, requires (a) an A or better financial-strength rating, (b) global licensing across ~70 countries, and (c) a claims organization that pays large losses without litigation games. AIG has all three. Buffett describes the same idea in [4]: 'insurance is the sale of promises… both the ability and willingness of the insurer to pay – even if economic chaos prevails – is all-important. Berkshire's promises have no equal.' AIG's promises are not Berkshire's, but they are credible enough to keep Fortune 500 risk managers on the panel. The erosion risk is that the same is true of Chubb, Zurich, Allianz, and the Lloyd's syndicates — AIG is one of perhaps eight global lead carriers, not a unique node. Verdict on intangibles alone: NARROW.

3. Switching costs. Commercial insurance renews annually. Brokers re-market at every cycle and a 5-10% rate difference reliably moves the account. Long-tail lines (D&O, casualty, workers' comp) have some stickiness because incumbents 'own' the loss history, but this is a soft moat — not the multi-year SaaS contractual lock-in investors sometimes imagine. Verdict: NONE-to-WEAK.

4. Network effects. None in the underwriting business. There is a mild data flywheel — more policies underwritten produces a richer loss database which improves pricing — but every large global carrier has the same flywheel and reinsurance brokers commoditize the data through industry exposure databases. Verdict: NONE.

5. Pricing power. P&C is structurally cyclical. Buffett describes the iron rule in [6] and [1]: 'a sound insurance operation needs to… be willing to walk away if the appropriate premium can't be obtained. Many insurers pass the first three tests and flunk the fourth.' AIG's history is an almost textbook case of flunking the fourth test — chasing volume in soft markets, taking large reserve charges in 2010, 2016 and again on long-tail casualty in 2022-2024. The Zaffino-era underwriting reform is real (combined ratio improved from ~110 in 2017 to low-90s in 2024-2025), but pricing power in commercial P&C is mostly the cycle, not the franchise.

Stress test. Give a competitor $10B and 5 years. In commercial P&C this absolutely buys you in: see Berkshire Hathaway Specialty Insurance going from zero to ~$10B of premium in roughly a decade, or Fairfax/Allied World/Sompo expanding leadership lines materially through hiring teams. The barrier to entry is regulatory and ratings — meaningful, but not insurmountable for any well-capitalized entrant willing to be patient.

Synthesis. AIG has a NARROW intangibles moat (global licenses, A rating, claims-paying reputation, broker relationships in specialty lines like aviation, energy and political risk) and essentially nothing else. It is not GEICO. It is not Berkshire Reinsurance. It is a credible global lead carrier whose economics will track the cycle and management discipline more than any structural advantage.

Moat verdict: NARROW.

Management

Peter Zaffino became CEO in March 2021 after running General Insurance under Brian Duperreault from 2017. The Duperreault-Zaffino regime inherited a mess: AIG had taken roughly $15B of cumulative reserve charges between 2010 and 2017, the General Insurance combined ratio was ~110 in 2017, and the franchise had been hollowed out post-crisis. The simplification record is genuine.

Reinvest. Reinvestment in the underwriting business has been disciplined and unglamorous: the AIG 200 program took ~$1B+ out of the cost base, the firm exited unprofitable books (much of consumer lines internationally, parts of personal lines, mass-affluent), and built a global specialty platform with the Validus Re acquisition (later partially divested). The AIG Re/Validus moves around 2018-2024 are the kind of decisions Buffett warns about in [1]: 'For decades, General Re was the Tiffany of reinsurers… The General Re of 1998 was not operated as the General Re of 1968.' Zaffino's analog is the Validus reinsurance unit which AIG bought in 2018 and substantially divested in 2024 to RenaissanceRe — a textbook 'walk away when the cycle turns' move.

Acquire. Validus (2018, $5.6B) was bought near a cycle bottom and the reinsurance piece sold near a cycle peak. That is good capital allocation. Smaller bolt-ons (Glatfelter brokerage, etc.) are minor and have been managed pragmatically.

Debt. AIG has been a net deleverager. Senior unsecured debt is down materially from ~$28B post-crisis. The interest-coverage figure of 0.0 in the scorecard reflects GAAP earnings volatility from the Corebridge deconsolidation accounting, not actual debt-service stress — financial leverage at the holding company is roughly in line with A-rated peers.

Buybacks. This is the dominant capital action and the place Buffett's framework is most demanding. Share count is down 5.5% over 10 years (scorer: -0.0548). That is a modest reduction given the size of asset sales. The buyback program — running $1.5-2B per quarter in 2024-2025 funded by Corebridge sell-downs and Validus Re proceeds — is happening at prices around 0.95-1.1x tangible book. That is roughly fair, not the sub-IV bargain Buffett demands. Munger's standard is 'buybacks below intrinsic value'; AIG's are at-or-near IV. So buybacks are accretive but not a value-creation engine. Grade for buyback discipline: B-.

Dividends. ~$0.40/quarter, payout ratio ~25-30% of normalized earnings. Sensible.

Communication quality. Disclosure under Zaffino has improved markedly versus the Hancock era. Quarterly decks separate accident-year ex-cat combined ratios from prior-year development, which is the right way to read a P&C insurer. Reserve-charge surprises have shrunk in size and frequency. Investor-day targets (mid-90s combined ratio, mid-teens core ROE) have been broadly hit. There is still some 'adjusted' earnings massaging — adjusted-after-tax-income excluding Corebridge mark-to-market — but it is defensible given the deconsolidation noise.

The honest read. Zaffino is a competent operator who has done what was needed: shrink the perimeter, raise underwriting standards, return capital. He is not Ajit Jain [6] and AIG is not building a structural cost or capacity advantage. The 'four insurance disciplines' from [6] — understand exposures, assess loss probability, price for profit, walk away — are now being practiced more often than not, but the prior twenty years of letters give little room to assume permanence.

Capital allocator: B.

Industry

Apply Porter's Five Forces to global commercial property-and-casualty insurance.

1. Rivalry — HIGH. The commercial P&C market is fragmented globally and competitive at the top: Chubb, Travelers, AIG, Zurich, Allianz/AGCS, Liberty Mutual, Tokio Marine, Hartford, plus the Lloyd's market and Bermuda specialty carriers (Arch, Everest, RenaissanceRe). Berkshire Hathaway Specialty Insurance has gone from zero to ~$10B in roughly a decade, and alternative capital (cat bonds, sidecars, ILS funds) commoditizes catastrophe capacity that used to be a source of profit. Pricing is famously cyclical: Buffett has warned for decades that in soft markets 'the other guy is doing it, so we must as well' [6] is the dominant industry psychology. The 2025 letter [3] explicitly notes 'additional capital entered the market, resulting in lower pricing or decelerating rate increases in several important lines… we expect these primary insurance businesses to face continued headwinds in 2026.'

2. Threat of new entrants — MEDIUM. Capital is genuinely a barrier (a global lead carrier needs $40B+ of equity, an A or better A.M. Best rating, and licenses in dozens of jurisdictions), and so is the claims-paying track record described in [4]. But the entrant doesn't need to recreate AIG from scratch — they can hire a team of underwriters, lift a book, and rent ratings via fronting/MGA structures. The last decade has seen exactly this with Bermuda specialty start-ups and InsurTech-backed MGAs. Net: meaningful but not impassable.

3. Buyer power — MEDIUM-HIGH. Three brokers — Marsh, Aon, WTW — intermediate the vast majority of large commercial accounts. They aggressively re-market at renewal, run quote-by-quote competitions, and have driven structural pressure on insurer expense ratios. Large corporate buyers also self-insure through captives for working-layer risk and only buy excess capacity from the market, capping primary premium growth.

4. Supplier power — LOW-MEDIUM. The 'suppliers' are reinsurers and capital. Reinsurance pricing is itself cyclical and currently softening per [3]: 'significant price declines in property reinsurance.' Talent is the binding supplier — senior underwriters with track records can command large packages and move books between carriers. Capital itself is plentiful at the moment.

5. Substitutes — LOW-MEDIUM. Captives, parametric covers, and cat bonds are real but partial substitutes. They erode profit pools at the edges (catastrophe reinsurance most acutely) but do not replace the underwriting of complex casualty, D&O, professional liability, aviation, energy or political risk where AIG has competence.

Value-pool location and trajectory. The commercial P&C value pool is now ~5-6 years past the start of a hard market that began in 2019. Casualty pricing is still firming in 2025-2026 because of social inflation and adverse loss-cost trends. Property and reinsurance pricing have rolled over. The pool is therefore migrating from property toward casualty — but casualty is exactly where AIG has had its worst reserve experience historically, and where the long tail means today's earned premium will not be tested for 5-10 years.

Combined assessment. Insurance generates two profit streams: underwriting profit (small and cyclical for non-GEICO carriers) and float income (large and steady when interest rates are 4-5% as they are now). The float-income tailwind is the single biggest thing supporting AIG earnings in 2025-2026 and will fade if the Fed cuts.

Industry Verdict: Average. Commercial P&C is structurally a fair-to-okay business — better than airlines, worse than software — and the cycle is mid-to-late stage.

Inversion

Now I am the short seller. The bull case is 'turnaround under Zaffino, buybacks below IV, normalized 13% ROE.' Here is why each leg breaks.

The single event that kills this. A major casualty reserve charge in 2026-2028 — somewhere between $2B and $5B — covering accident years 2017-2022. AIG's casualty book has been priced through years of social inflation: nuclear jury verdicts have grown ~40% from 2019 to 2024, third-party litigation funding has industrialized plaintiffs' bar economics, and the 'incurred but not reported' (IBNR) factor on long-tail casualty is the biggest single estimate on the balance sheet. Travelers, Chubb, and Hartford have all taken accident-year casualty charges in 2023-2025; AIG's claim that its book is 'derisked' relative to peers is exactly the line every prior AIG management took before the next charge. One $3B reserve charge takes a year of normalized earnings to zero, and the multiple compresses from ~1.0x book to 0.7-0.8x book overnight. Stock drops to $55-60.

Why the moat is narrower than bulls think. The bull narrative is 'global specialty franchise impossible to replicate.' Two facts contradict this. First, AIG just sold the most attractive specialty piece — Validus Reinsurance — to RenaissanceRe in 2024 because it was sub-scale and competitively disadvantaged. If the franchise were truly moaty, you don't sell it at the cycle peak. Second, Berkshire Hathaway Specialty Insurance reaching ~$10B of premium in a decade [3] proves that a well-capitalized entrant with a brand and discipline can credibly compete. The 'global licenses, A rating, claims-paying reputation' moat is shared with Chubb, Zurich, Allianz, AXA XL, and Tokio Marine — that is not a moat, that is the price of admission to the top division.

Why management is worse than it appears. The Zaffino restructuring narrative is real, but compare results to the proper benchmark. Chubb has compounded book value per share at ~10% with combined ratios in the mid-80s through the same cycle. AIG has bought back stock funded by selling Corebridge and Validus Re — that is liquidating, not compounding. Total shareholder return since the Duperreault arrival in 2017 has trailed the S&P 500 P&C index. The 5.5% share-count reduction over 10 years (scorer: -0.0548) is paltry against what Chubb and Travelers achieved organically. The buybacks are happening at ~1.0x book — accretive but not value-creating. The IV-low/IV-base/IV-high values in the scorer (negative because GAAP free cash is consumed by the divestiture-and-buyback financial engineering) are a tell: the model cannot find a positive earnings stream.

What bulls are extrapolating that won't hold. Three bull extrapolations are vulnerable. (1) 'Combined ratio drops to mid-80s.' This requires both expense ratio improvement and accident-year loss ratio improvement to continue simultaneously; the former is mostly done, the latter depends on rate-on-rate exceeding loss trend, which is now reversing in property and reinsurance per [3]. (2) 'Float income at 4.5%+ forever.' If the Fed cuts to 3% over 2026-2028, AIG's net investment income falls $500-800M, all of which drops straight through to pretax. (3) 'Buybacks compound TBVPS at 8%.' At 1.0x book, buybacks are TBVPS-neutral — they don't compound it, they just reduce share count.

Valuation trap (multiple compression / regime change). AIG trades at ~1.0x tangible book today. The historical AIG multiple range during periods of weak underwriting was 0.6-0.8x book. If the cycle turns and a reserve charge hits, the multiple compresses by 25-40% from a starting earnings base that is itself peak-cyclical. That is the canonical value trap: you bought at 1.0x book on $80 TBVPS expecting compounding, you end up at 0.75x book on $75 TBVPS after a charge — a 30% drawdown — and you have to wait 3-5 years to re-earn it. The 10-year P/E average of 5.65 in the scorecard is a hint: the market has historically refused to give AIG a normal earnings multiple, because the market does not believe the earnings.

If I am right, the stock could be worth $50 within 2-3 years.

Lollapalooza Bias Check

Several biases are active in me right now.

Anchoring. The current price is $78.77 and the scorecard's intrinsic-value range is unhelpful (negative across low/base/high because the deterministic model is producing nonsense on a financial during a divestiture period). I am therefore anchored to tangible book per share — roughly $80 — as my reference point. That anchor is not necessarily 'value'; it is just the most stable number I have. I should be honest: I am defaulting to book because DCF doesn't work here, not because book is the right answer.

Authority and social proof. The Buffett canon is unusually rich on insurance because Berkshire is itself one of the great insurance operators. It is tempting to apply Buffett's praise of GEICO and Ajit's reinsurance unit [3][6] as a kind of moat checklist and conclude AIG falls short. That is mostly correct, but I should be careful not to use 'AIG is not GEICO' as if that were the same as 'AIG is uninvestable.' Many fair businesses at fair prices have made fine returns; AIG just isn't a fat pitch.

Confirmation bias and recency. AIG has a 20-year reputation for blowing up — 2008 AIGFP, 2010 reserve charges, 2016 reserve charges, 2017 hurricane losses. I am pre-disposed to seek out evidence that the new regime is just the old regime in fresh paint. The combined-ratio improvement under Zaffino is real and I should give it appropriate weight, not dismiss it as cyclical. Conversely, recency bias also cuts the other way: the 2024-2025 results are unusually clean and may be making me too willing to believe in normalization.

Commitment / consistency. Once I tag a stock 'Too Hard' or 'Avoid,' it becomes psychologically expensive to revisit. The honest answer here is 'Hold/fair price' — a less satisfying conclusion than 'Buy aggressively' or 'Avoid.' I should resist the temptation to force a strong call where the evidence supports a moderate one.

Deprival super-reaction tendency. If AIG fell 25% on a reserve charge tomorrow, the very same analysis would call it a Buy at 0.75x book. The price-to-conclusion sensitivity is high here — that is itself a signal that the underlying business is not a compounder, only the price is the variable.

Incentive-caused bias (in management, not me). The biggest second-order Lollapalooza is in AIG itself: P&C underwriters across the industry are paid on accident-year combined ratios that are estimated, with the true combined ratio not knowable for 5-10 years on long-tail lines. This is a structural incentive to under-reserve. Munger would say: assume the bias is at work, demand a margin of safety, and don't pay above book.

10-Year Outlook

In 2036, will AIG be recognizably the same business? Probably yes. The simplification trajectory — sell Corebridge, divest Validus Re, run off Fortitude Re, focus on commercial and specialty P&C — is essentially complete. AIG in 2036 is most likely a $25-30B premium global commercial-and-specialty insurer with combined ratios oscillating between high-80s in hard markets and high-90s in soft markets, mid-teens core ROE in good years and high-single-digits in bad years. That is a fundamentally similar shape to AIG today, and similar to Chubb, Zurich, and Travelers a decade out.

Customer base larger? Modestly. Global commercial-insurance premium grows roughly with global nominal GDP plus a real growth premium for cyber, climate, and complex casualty exposures — call it 5-6% nominal. AIG's organic premium growth is likely to track or slightly trail this depending on cycle discipline.

Profit per customer higher? Uncertain. The dominant variable is the loss-cost trend on long-tail casualty (social inflation, jury verdicts, litigation funding) versus rate adequacy. There is no clear edge for AIG here over Chubb or Travelers; if anything, peers are ahead.

Moat wider? No. Specialty franchises like aviation and energy may quietly erode as captives, MGAs, and Lloyd's syndicates absorb working-layer risk. The intangibles moat (global licenses, A rating, claims reputation) is shared with several peers and not widening.

Single biggest threat. A 1-in-50-year catastrophe sequence (multiple major hurricanes plus a Cat 5 California earthquake) combined with adverse casualty reserve development. This is the tail-risk that wiped out AIG in 2008-2010 in slow motion via AIGFP and reserve charges; the modern AIG is better-capitalized but the underlying exposure to long-tail uncertainty is structural to the industry.

Does the business pass the 10-year test as a compounder? The franchise will exist; whether it compounds intrinsic value at 8-10% per year depends on cycle management and reserve discipline that the prior twenty-year track record does not support. As a fair-priced cyclical at book value, it is investable. As a 'forever' compounder, it is not.

CONFIDENCE: medium

Position Guidance

• Recommendation: Hold
• Conviction: medium
• Target buy price: $62 (roughly 0.78x tangible book — the level at which a reserve-charge scenario is priced in and buybacks become genuinely value-accretive)
• Target trim price: $98 (roughly 1.22x tangible book — above this you are paying for sustained mid-teens ROE that AIG has not earned over a full cycle)
• Position sizing: 0-2% of portfolio for existing holders; new buyers wait for $62 entry. Treat as a cyclical financial, not a compounder. Size to survive a 30-40% drawdown without forced selling. Pair-trade against Chubb if expressing a view on AIG's relative execution rather than absolute insurance exposure.