Industry Primers · Latticework by MOI Global

Know an industry the way its specialists do.

Each primer is a full-length research report on an entire industry, written in the format of an institutional initiation, with a companion slide deck and a two-part podcast. One industry at a time, to expand and deepen your circle of competence.

Free subscribers receive every companion deck and Part One of every podcast.

10
industries covered,
3 sectors
42
hours of narrated audio
across 20 episodes
8
parts in every report,
one architecture
Fridays
a new industry,
delivered to your inbox
Four ways into every industry

Read it, skim it, or listen to it.

Premium

The research report

An institutional initiation on a whole industry: its history, economics and players, the accounts, valuation, AI and the risks. Every figure dated and sourced.

Free

The companion deck

A self-contained slide summary of the report: the spine of the argument, the key exhibits and tables, the investable universe and the checklist.

Free

Podcast, Part One: The Story

The industry’s story, from its founders to today, told for anyone. About 95 min per episode.

Premium

Podcast, Part Two: The Deep Dive

A working method for professionals: the economics, the accounts that mislead, valuation and the questions to ask. About 2 h 36 min.

The library

10 industries, and counting.

No. 10 · November 13, 2026

How managers in New York, Toronto, Stockholm, Zug, London and Paris earn fees on capital locked up for years, why their shareholders out-earned the market and the funds’ own investors, and how specialists appraise the managers and the listed vehicles

No. 9 · November 6, 2026

How insurance brokers in the United States, London, Europe, Australia and India earn a share of premiums without bearing the losses, why the price paid for agencies and for the shares decided the long records, and how specialists appraise the companies

No. 8 · October 30, 2026

How robot makers and their suppliers in Japan, Europe, China, Korea, Taiwan and the United States make money, why the best records belong to companies that keep earning from the machines already at work, and how specialists appraise the companies

No. 7 · October 23, 2026

How royalty companies make money from mines and wells they don’t operate, why the price paid and the share count decide the owners’ records, and how specialists appraise the companies

No. 6 · October 16, 2026

How fertilizer producers make money from gas, rock and ore, why capital allocation through the cycle matters more than a specific asset, and how specialists appraise the companies

No. 5 · October 9, 2026

How homebuilders make money, why the few builders that stopped owning land and kept their debt low made the industry’s great records, and how specialists appraise the companies

No. 4 · October 2, 2026

How aerospace companies make money, why the owners of small certified monopolies in the supply chain compounded far faster than the airframers, and how specialists appraise the companies

No. 3 · September 25, 2026

How regional banks make money, why the gap between the best operators and the industry’s own index is the widest in financial services, and how specialists appraise the companies

No. 2 · September 18, 2026

How property and casualty insurers make money, why most celebrated (re)insurance owner-operators trailed the index over the last decade, and how specialists appraise the companies

No. 1 · September 11, 2026

Research report and companion slide deck on how exchanges and financial market infrastructure businesses make money, what their long-term returns show, and how to appraise their value.

Inside a primer

Eight parts. One architecture. Every industry.

Every report follows the same structure, so once you have read one primer you can find your way around all of them, and compare industries on the same terms.

Every chapter ends with a “What to remember” box; read in sequence, the boxes summarize the report. The appendices give the analyst the working tools: the investable universe, a due diligence checklist, a reading and listening course of letters, filings and recordings of founders and chief executives, and the numbered notes that source every figure.

Start reading
  1. I
    Orientation

    What the industry is, how its businesses make money, and the map of who is paid for what.

  2. II
    Evolution

    The history, from the founders to today, and the case studies of who built value and who lost it.

  3. III
    Structure & Economics

    Margins, moats and the cycle: where the profit in the chain sits, and why.

  4. IV
    The Players

    The owner-operators and their records, judged by the value per share they left their owners.

  5. V
    Speaking the Language

    The accounts and the industry’s statistics, and the adjustments that make them comparable.

  6. VI
    Analysis & Valuation

    How specialists value the companies, with the transactions and the sources of long-run returns.

  7. VII
    The AI Inflection

    What artificial intelligence already pays for, and what is still priced ahead of its revenue.

  8. VIII
    Risks & Debates

    How the companies fail, the warnings that appear first in public documents, and the open debates.

  9. +
    Appendices

    The investable universe, the checklist, the reading and listening course, and the source notes.

From the reports

Original exhibits, drawn from primary sources.

Each report carries dozens of exhibits built from filings, regulators’ statistics and long price histories, every one with its sources and its date.

Property and Casualty Insurance & ReinsuranceFloat grew from $17m to $176bn, and its cost was below zero in 17 of the first 31 years and in each of the last three
AerospaceSeven years of returning more than the cash, then seven years of burning it
Exchanges & Financial Market InfrastructureThe leaderboard since demutualization, and the 80x spread inside it
HomebuildingOn the same margin, NVR earned twice the land owners’ return, because its inventory turned four to five times as fast
FertilizersFour spikes since 1974; in real terms, 1974 is still the record for urea and DAP
US Regional BanksThe count has fallen every year since 1984, and only twice did failures do the work
AerospaceAirbus: from 1% to parity by 2003, and clear ahead only when Boeing stopped delivering
Property and Casualty Insurance & ReinsuranceFourteen of nineteen beat the index; the median edge is a point a year
FertilizersRecords starting before 2008 mostly beat the S&P 500; all ten from 2013 on trailed it
Built to be used

Read it the way your work demands.

No primer needs to be read in order. Each suggests its own paths through the report; these are the usual ones.

The generalist with limited timeParts I and VI, plus the due diligence checklist
The analyst on a specific companyParts III, V and VI, in that order
The student of business historyPart II, and Part One of the podcast
The accountantPart V, with the valuation chapter
The owner-operator hunterPart IV, then the return attribution in Part VI
The risk officerPart VIII, then the economics in Part III
The technologistPart VII
The commuterPart One of the podcast on the way in, Part Two at the desk

Latticework is the Substack of MOI Global, the research-driven membership organization of intelligent investors. For nearly two decades, MOI Global has hosted summits, published research and held the kind of conversations that practicing investors find indispensable.

Read by investors at
  • Berkshire Hathaway
  • Elliott Management
  • Oaktree Capital Management
  • Southeastern Asset Management
and many other independent firms, family offices and endowments
Membership

Start free. Go deep when you are ready.

Free

Sample every industry

  • The companion slide deck of every primer
  • Part One of every podcast: the industry’s story
  • The overview and key findings of each primer
  • Weekly Inspiration and the light editions of our weekly briefings
Subscribe free
Made for the primers

Premium

The full body of work

  • Every full research report, the archive and each new primer
  • Part Two of every podcast: the deep dive for professionals
  • The full Monday Morning Briefing and Special Situations decks
  • In-depth investment ideas and the complete archive
  • Our four flagship online conferences, live, with replays and transcripts
  • The AI Bootcamp for Non-Technical Investors
  • An invitation to the bi-monthly live member call
Become a Premium member

Institutional

For investment firms

  • Everything in Premium
  • Full MOI Global membership, subject to approval
  • The institutional community that has connected fund managers, allocators and firm founders for nearly two decades
  • Group subscriptions for your team
Explore Institutional

Plans and current pricing are on the Latticework subscribe page.

Questions

Frequently asked.

What exactly is an Industry Primer?

A full-length research report on an entire industry, written in the format of an institutional initiation for fundamentally oriented, long-term investors. Each one comes with a companion slide deck and two podcast episodes: Part One tells the industry’s story for anyone, Part Two is the deep dive for professionals.

What can I read for free?

Every free subscriber receives the companion slide deck and Part One of the podcast for each primer, along with the overview and key findings. The full report and the Part Two deep dive are for Premium members.

How often are new primers published?

A new industry arrives on Fridays. Members’ requests inform the order of coverage, so tell us which industry you want next.

Do Premium members get the back catalog?

Yes. Premium includes every primer published so far, as well as each new one as it comes out, together with the rest of the Premium library.

Can my team or firm subscribe together?

Yes. Group subscriptions are available on the subscribe page, and the Institutional tier adds membership in MOI Global (subject to approval) for investment firms.

Is this investment advice?

No. The primers are published for educational purposes. They describe how industries work and how specialists appraise companies; no security mentioned is recommended for purchase or sale.

Expand your circle of competence, one industry at a time.

Industry Primer No. 10 · November 13, 2026

Alternative Asset Managers

How managers in New York, Toronto, Stockholm, Zug, London and Paris earn fees on capital locked up for years, why their shareholders out-earned the market and the funds’ own investors, and how specialists appraise the managers and the listed vehicles

A listed alternative asset manager is paid a fee on capital its clients have committed for years and a share of the gains when that capital is sold. Because the fee is charged on committed capital, the management and advisory fees of Blackstone (NYSE: BX) fell only 8.2 percent from 2007 to 2009, while its realized performance fees net of compensation fell 96 percent from 2007 to 2008. The ten largest listed fee-earning managers were worth about USD 531bn at the October 22, 2026 close, and Blackstone alone was 28 percent of them. After 2008 the managers added capital with no end date, through insurers they own and funds sold to individuals: Athene, the insurer that merged with Apollo Global Management (NYSE: APO) in 2022, earned 54 percent of Apollo’s 2025 segment income.

From 2012 to 2025 the managers’ shareholders out-earned both the market and the managers’ own fund investors, but in 2026 every listed manager fell while the market rose. From May 3, 2012 to the end of 2025, seven of eight managers with a full record beat the S&P 500’s total return of 15 percent a year, and The Carlyle Group (NASDAQ: CG), at 13 percent, did not. Blackstone’s shareholders earned 26 percent a year, while the investors in its corporate private-equity funds earned a net internal rate of return of 15 percent since 1987. In 2026, to the October 22 close, every listed manager tracked fell 6 to 35 percent while the S&P 500 returned 13 percent. The report traces the fall to redemptions at the credit funds sold to individuals, two alleged frauds against lenders and the fear that AI would hurt the software companies the funds own.

Listen to Part One freeThe story of the alternative asset managers, from a Toronto company that sold São Paulo’s streetcars and power for cash in 1979 to the founders who built the largest firms by owning a stake beside their clients, told for anyone.

Key findings19 findings across the report. Tap any one to expand.

  1. Part IOrientationThe fee on committed capital makes a manager’s income steadier than its funds’ results, and the managers differ less by strategy than by how long their capital stays.

    On the authors’ USD 10bn buyout fund, with a 1.5 percent fee and 20 percent of the profit above an 8 percent preferred return, the manager collects USD 953m of fees over the fund’s life, USD 750m of them before it sells anything. Its USD 1,610m share of the profit, called carried interest or carry, arrives only in the ninth year and only because the investments doubled, and below about 1.6 times cost the carry is zero while the fees do not change. The share of assets under management (AUM) with no fixed end date runs from 16 percent at TPG (NASDAQ: TPG) to about 70 percent at Blue Owl Capital (NYSE: OWL).

  2. Part IIEvolutionThe industry’s base is still the contract KKR wrote in 1976: a fee on committed capital and a fifth of the profit.

    Jerome Kohlberg, Henry Kravis and George Roberts founded KKR (NYSE: KKR) with USD 120,000, and its funds of 1976 to 1986 earned 26 to 39 percent a year net. The 1987 Fund bought RJR Nabisco for about USD 31bn, mostly with debt, and earned 8.9 percent a year net while its fee was charged on the commitments throughout. Every American platform began with one strategy and added credit, real estate and later insurance, while Europe’s managers grew out of banks and industrial groups and listed as companies: 3i Group (LSE: III) from 1945, EQT (Nasdaq Stockholm: EQT) from the Wallenbergs’ Investor AB in 1994 and Partners Group (SIX: PGHN) in Zug in 1996.

  3. Part IIEvolutionBrookfield’s history is a series of decisions about who would own the fee stream and who would own the capital.

    Its predecessor Brascan, chartered in Toronto in 1899, sold its Brazilian utility for USD 380m, and in 1979 Edper, the Bronfmans’ company run by Jack Cockwell, bought about 30 percent of it in the open market within five weeks, and a federal judge declined to stop it. Its holding companies were folded into one by 1997, controlled by a partnership of its managers. Under Bruce Flatt from 2002 it became a manager of other people’s capital and bought Oaktree and about USD 13bn of insurers. The separations of 2022 and 2025 put the fee stream in Brookfield Asset Management (NYSE, TSX: BAM) and the capital in BN, whose 74 percent of BAM accounted for 61 percent of BN’s share price at the October 22 close.

  4. Part IIEvolutionCapital with no end date changed the managers after 2008, and in 2026 the caps on withdrawals held while the share prices fell.

    At the end of 2025 Athene-related capital was 42 percent of Apollo’s AUM and Global Atlantic 29 percent of KKR’s. The funds sold to individuals allow an exit capped at about 5 percent a quarter, and in the first quarter of 2026 holders of Blue Owl Technology Income Corp asked to redeem 40 percent of its shares. Every fund paid at its stated net asset value (NAV), and the fees continued on the capital that stayed. BREIT, Blackstone’s real estate fund for individuals, reached its repurchase limits in December 2022 and cleared its waiting requests at NAV in fourteen months.

  5. Part IIIStructure & EconomicsThe limited partners pay most of their cost as carry, and in the channel that sells to individuals the distributor takes about two thirds as much as the manager.

    On the worked fund the limited partners (LPs) pay USD 2.6bn in fees and carry, 63 percent of it carry, and keep 11 percent a year of a 15 percent gross return. On BREIT’s Class S terms Blackstone receives USD 2.2 of a USD 9 gross return on USD 100, and the distribution chain USD 1.6. Fees follow fundraising and carry follows the sale of investments: Blackstone’s management and advisory fees rose 28 percent from 2022 to 2025, while its realized performance revenues fell 54 percent from 2022 to 2023. Distributions to LPs have held below 15 percent of NAV for four years by the count of Bain & Company.

  6. Part IIIStructure & EconomicsPay decides the margin on fees, and the contract protects the fee only until the clients decide whether to commit to the next fund.

    In 2025 the margin on fee-related earnings (FRE), the profit on management and related fees, ran from 42 percent at Ares Management (NYSE: ARES) to 69 percent at KKR, each on its own definition. Shareholders kept 61 percent of realized carry at Blackstone and about a quarter at KKR and Apollo. No listed manager discloses how many of its LPs commit again. Lower fees in the largest funds, LPs that invest directly at no fee, the departure of named people and redemption requests above the caps all erode the protection.

  7. Part IVThe PlayersInsurer money and money from individuals explain most of the difference in the managers’ growth.

    In the year to mid-2026 the AUM of Apollo and TPG rose 25 percent, against 4 percent at Carlyle, and 58 percent of the USD 77bn BAM raised in its record second quarter came from its parent’s insurer. Apollo’s insurer earned more than its fees in 2025. Fees grew at every major while carry waited, so in the third quarter of 2026 Blackstone’s FRE rose 6 percent while its distributable earnings (DE), the broader measure that includes realized carry, fell 9 percent.

  8. Part IVThe PlayersThe founders who compounded value added strategies, listed part of the firm and handed over the chief executive’s job while keeping the vote.

    Kravis and Roberts returned 24 percent a year as listed co-chief executives, against 15 percent for the index, and their control through KKR’s Series I preferred share ends on December 31, 2026. Carlyle’s founders, as listed co-chief executives from 2012 to 2017, returned 8.3 percent a year against 15 percent. Stephen Schwarzman holds about 19 percent of Blackstone’s economic shares and controls the vote for its directors, and Ares’s founders hold 80 percent of its votes. Outside New York the founders diluted themselves, and Conni Jonsson owns 3.6 percent of EQT.

  9. Part IVThe PlayersBrookfield is controlled by a partnership of its managers who are paid almost entirely through the share price.

    The partnership owns about 20 percent of BN’s Class A shares, worth USD 18bn at the October 22 close, and elects half its board through Class B shares held in a trust in which Flatt holds 41 percent. Flatt’s cash from 2021 to 2025 was a salary averaging USD 533,292 a year, and nine tenths of his pay for the period came from the share price. Over his tenure BN compounded at 19 percent a year to the end of 2025, against 9.9 percent for the S&P 500, with most of the excess earned before 2012, and on the way its shares fell 72 percent from May 2007 to March 2009.

  10. Part IVThe PlayersIn a listed vehicle the shareholder owns the capital and pays the fee, and who holds the vote explains the widest discounts better than the size of the fee.

    At the October 22 close every vehicle in the report traded below its own measure of value, from about 26 percent at 3i and Greenlight Re (NASDAQ: GLRE) to 64 percent at Tetragon Financial Group (Euronext Amsterdam, LSE: TFG), whose listed shares carry no vote. Pershing Square Holdings (LSE, Euronext Amsterdam: PSH) paid about USD 697m of fees in 2025 and stood 35 percent below NAV on October 20, 2026, while its manager, Pershing Square Inc. (NYSE: PS), was worth about USD 18bn. Greenlight Re’s book value per share grew 1.4 percent a year from its 2007 listing, and Third Point Investors, now Malibu Life Holdings (LSE: MLHL), traded 61 percent below its book.

  11. Part IVThe PlayersThe managers have spent their own capital mostly on insurers and on other managers, while Blackstone paid its capital out.

    KKR, Apollo and Brookfield put about USD 30bn into insurers from 2021, and the insurers’ earnings depend on a spread that at Athene narrowed from 1.9 percent in 2023 to 1.1 percent in the second quarter of 2026. The managers bought by other managers cost a median 6.5 percent of the AUM announced, mostly paid in the buyers’ stock. Blackstone paid out USD 42bn from 2015 to 2025 and bought nothing large after 2008. Pershing Square sold 10 percent of its business in May 2024 at a USD 11bn valuation and was worth about USD 18bn listed at the October 22 close.

  12. Part VSpeaking the LanguageEach manager defines its own earnings measures, so a reader must name the measure before comparing any two firms.

    Fee-earning AUM, FRE and DE are each defined by the firm that reports them, and the internal rate of return rises with any delay in calling capital. KKR’s second-quarter 2026 FRE rose 37 percent as reported and 19 to 26 percent like for like, after KKR moved about USD 160m of performance fees into FRE and did not restate the prior year. Blackstone’s 2025 DE of USD 7.1bn was 2.4 times its GAAP net income attributable to the company. Brookfield publishes three sets of accounts, and BN’s DE already contains about three quarters of BAM’s, so the two are never added.

  13. Part VIAnalysis & ValuationWith the carry and the balance sheet taken out, the market pays for the managers’ fee streams at very different multiples.

    At the October 22 close the listed managers traded between 10 times (Carlyle) and 27 times (BAM) their own earnings measures, with a median of 16 times. With the net accrued carry and the balance sheet taken out at their stated values, the market paid 24 times after-tax FRE at Blackstone and at Ares, about 19 times at KKR and Apollo once their insurers are valued at an assumed 10 times earnings, and 8.7 times at Carlyle. At a required return of 9 percent, Blackstone’s multiple assumes a fee stream growing about 4.7 percent a year indefinitely, against about 10 percent a year from 2012 to 2025.

  14. Part VIAnalysis & ValuationRebuilt from market prices and its own book values, Brookfield Corporation is worth 24 percent more than its share price and 29 percent less than Brookfield’s own estimate.

    The report puts BN at USD 47 a share, against a close of USD 38 on October 22 and Brookfield’s estimate of USD 67 at June 30, 2026. Two management judgments make most of the gap: the insurer at 15 times earnings adds USD 7.3 a share, and future target carry at 10 times adds USD 12. The fact that would settle the argument is realized carry, net USD 520m over the twelve months to June 2026 against Brookfield’s target of about USD 2.7bn a year.

  15. Part VIAnalysis & ValuationThe shareholders earned more than the funds’ investors because they owned the fee on a growing base, and the entry price decided each buyer’s result.

    Of Blackstone’s 23 log points a year from 2012 to 2025, 14 came from growth in DE per share, 4.9 from dividends and 4.5 from a higher multiple. A buyer of Blackstone at its USD 31 offer of June 2007 lost about 86 percent by February 2009 and still earned 13 percent a year to the October 2026 close, against 11 percent for the index. The failures were the firms whose capital could leave: Fortress was sold in 2017 at 56 percent below its offer price before dividends, and the holders of Och-Ziff lost about 96 percent.

  16. Part VIIThe AI InflectionAI reached the managers first as an asset and then as a risk, and a shareholder owns the fee on the AI assets rather than the assets.

    Blackstone said its data-center platform had grown from USD 130bn of total value in January 2026 to USD 185bn by mid-year. Funds managed by Blue Owl own 80 percent of the Hyperion data-center campus in Louisiana, about USD 27bn of development cost, which Meta Platforms (NASDAQ: META) leases from 2029 on a four-year initial term. To the low of March 12, 2026 six large US managers lost 31 to 42 percent of their price, against 20 percent for the software index, and by October software had recovered while the managers had not. USD 10bn of new data-center capital adds under 1 percent to Blackstone’s market value at the market’s own multiple.

  17. Part VIIIRisks & DebatesThe managers rarely fail as businesses, because their fees are contracted, but they fail their shareholders through what they added.

    The risks are an insurer whose spread can narrow, funds that promise individuals quarterly liquidity, and clients who can leave after a scandal. The alleged frauds of 2025 at Tricolor, an auto lender, and First Brands, an auto-parts maker, fell on the banks and other lenders that financed them. The managers’ own test is the share of loans on which interest is no longer recorded at their business development companies: 2.4 percent at cost at Ares Capital (NASDAQ: ARCC) in June 2026 and 7.1 percent at FS KKR Capital (NYSE: FSK). At Athene, the fall from the 2025 spread to the second quarter’s is worth about USD 1.5bn a year, 44 percent of 2025’s spread earnings.

  18. Part VIIIRisks & DebatesThe ten open debates come down to three tests: whether private credit’s losses stay small, whether individual investors’ capital stays put, and whether the founders’ firms keep their clients after the founders.

    Each debate has a settling fact that a filing will report, from the non-accrual loans and the results of the tender offers to KKR’s governance after December 31, 2026 and the discount of Pershing Square Holdings by mid-2027. Partners Group’s new co-chief executives, Roberto Cagnati and Juri Jenkner, are to start on January 1, 2027, subject to the approval of FINMA, the Swiss regulator.

  19. AppendicesThe appendices give the analyst the working tools: the investable universe of 56 listed names, from the US majors and the European managers to the listed vehicles and the largest business development companies, with sixteen private partnerships beside them, a checklist of 29 questions in the order of Chapter 22, a reading course of letters, filings and investor days with a listening course of 45 recordings of founders and chief executives, and the numbered notes that source the report.
Read the full report and hear the deep dive Premium members get the complete Alternative Asset Managers primer, every primer in the archive and each new one as it is published.
Industry Primer No. 9 · November 6, 2026

Insurance Brokers

How insurance brokers in the United States, London, Europe, Australia and India earn a share of premiums without bearing the losses, why the price paid for agencies and for the shares decided the long records, and how specialists appraise the companies

A broker earns a share of the premium its client pays an insurer and never pays the claim, so it needs little capital. On a US middle-market commercial account the commission is roughly 10 to 15 percent of premium. People are the main cost: compensation and benefits took 50 percent of 2025 revenue at Brown & Brown (NYSE: BRO) and 58 percent at WTW (NASDAQ: WTW), while capital spending ran from 1.1 to 2.4 percent of revenue at the five largest listed brokers. Those five, Marsh & McLennan (NYSE: MRSH; MMC until January 2026), Arthur J. Gallagher & Co. (NYSE: AJG), Aon (NYSE: AON), WTW and Brown & Brown, were worth USD 272bn at the October 21, 2026 close, 84 percent of a listed industry worth about USD 325bn. Their shares trade in New York, and Marsh, Aon and WTW earn about half their brokerage revenue outside the United States.

From 2005 to 2025 two of the five largest brokers beat the S&P 500, and much of the difference came from the price at the start. With dividends reinvested, Gallagher returned 15 times the money, 15 percent a year, and Aon 13 percent a year, against 11 percent for the S&P 500 total-return index, while Marsh compounded at 12 percent, about the market’s rate. Brown & Brown returned minus 3.6 percent a year over the five years from the end of 2005, when its shares were priced for the hard market of 2001 to 2003, and 15 percent a year from the end of 2010. The consumer brokers did worst: SelectQuote (NYSE: SLQT), a Medicare broker, returned 0.05 times the money from its first close in May 2020 to the end of 2025.

Listen to Part One freeThe story of the insurance broker, from London’s marine market of the 1570s to the founders in the United States, London, Australia, Germany and India who built the largest firms by giving their salespeople a stake, told for anyone.

Key findings16 findings across the report. Tap any one to expand.

  1. Part IOrientationA broker’s revenue follows insurance prices and insured values while the insurer bears the loss, and the industry’s layers move on different cycles.

    Marsh spent about 1.1 percent of its 2025 revenue on capital assets while its operations brought in cash equal to 20 percent. A broker also earns interest on premiums it has collected and not yet passed to the insurer: on the report’s USD 500,000 commercial account about USD 2,260 a year at 2026’s short rate and about USD 50 at 2021’s. Retail broking, wholesale brokers who place hard risks, managing general agents (MGAs) who underwrite on an insurer’s behalf, reinsurance broking and consulting each have their own cycle, and in the third quarter of 2026 Marsh Risk grew 5 percent on its underlying basis while Marsh’s Reinsurance segment fell 6 percent. Only the United States has a listed industry of size.

  2. Part IIEvolutionTwo legal rules made the broker’s business: only members could sign policies at Lloyd’s, and a New York court held in 1904 that an agent’s renewals belong to the agent.

    The first made the Lloyd’s broker the world’s wholesaler of hard-to-place insurance. The second made an American agency something that could be sold, so in 1961 J. Hyatt Brown paid USD 75,000 for his father’s agency, about 1.2 times its revenue, and bought its renewals. American groups later bought the London houses, paying a premium of 58 percent for Sedgwick in 1998 and 34 percent for Jardine Lloyd Thompson in 2018, because a London house was worth more inside an American distribution network than on its own.

  3. Part IIEvolutionConsolidation ran less on economies of scale than on a difference in price between private agencies and listed brokers.

    In 2010 a well-run private agency changed hands at about 6 times EBITDA guaranteed plus an earnout, by the figures of Reagan Consulting, an adviser to agencies, while the five largest listed brokers traded at about 10 times on the authors’ GAAP-based proxy. Private equity followed that difference, and the count of announced US and Canadian agency deals by OPTIS Partners rose from 184 in 2009 to a peak of 1,111 in 2021. After the Justice Department stopped the merger of Aon and WTW in 2021, the largest brokers bought or agreed to buy five large private brokers built by acquisition, NFP, McGriff, AssuredPartners, Accession and USI Insurance Services, for about USD 61bn from 2024 to 2026.

  4. Part IIEvolutionA broker’s value is destroyed by conflicted revenue and by revenue booked before the cash arrives.

    In October 2004 New York’s Attorney General alleged that insurers had agreed to pay Marsh more than USD 1bn of contingent commissions, payments for the volume or profitability of the business it placed with them, to steer clients their way. The contested payments equaled more than half of Marsh’s 2003 net income, and its shares fell 48 percent in four sessions, about USD 12bn of value, nearly fourteen times its USD 850m of restitution. The Medicare brokers booked a policy’s lifetime commissions at the sale and cut the estimates when customers left sooner: eHealth (NASDAQ: EHTH) carried USD 1,123m of commissions receivable at the end of 2025 against USD 554m of revenue for the year, and GoHealth went through Chapter 11 in 2026.

  5. Part IIIStructure & EconomicsA broker is paid before anyone knows whether the policy makes money, so its margins hold while insurers’ results swing.

    Kinsale Capital (NYSE: KNSL), an insurer that writes only through brokers, paid them an average commission of 15 percent of gross written premium in 2025. On a hard-to-place risk that passes through a retail broker, a wholesaler and an MGA, the intermediaries keep about 27 cents of the premium dollar on the authors’ illustration. In 2023 the US property and casualty insurers paid out 102 cents in claims and expenses for every dollar of premium, while Marsh’s GAAP operating margin was 23 percent. In 2009, with commercial premiums falling about 5 to 6 percent in every quarter, Marsh’s retail broking fell only 1 percent on its underlying basis.

  6. Part IIIStructure & EconomicsPeople cost decides a broker’s margin, and the producers who hold the client relationships are both its main asset and its weak point.

    One point of the compensation ratio is worth about USD 737m of pre-tax income a year to the five largest brokers together. Interest on the premiums a broker holds costs almost nothing to earn, and Marsh’s went from USD 15m in 2021 to USD 497m in 2024 and USD 403m in 2025. Producers can leave: four Florida leaders of Marsh resigned on July 21, 2025 to build the US business of Howden, a private broker founded in London in 1994, and Marsh’s complaint alleged about 90 resignations in a day. The court let the injunction’s bar on soliciting Marsh’s clients lapse after September 19, 2026.

  7. Part IVThe PlayersThe five largest listed brokers are five different businesses, so no organic growth rate or margin compares directly with another.

    WTW is 54 percent consulting and Brown & Brown almost none, and Brown & Brown earns 86 percent of its brokerage revenue in the United States against about half at Marsh, Aon and WTW. Gallagher’s revenue grew 21 percent in 2025 on 6 percent organic growth. Aon, with the highest adjusted operating margin of the five at 32 percent, agreed on August 31, 2026 to acquire USI for USD 17bn in cash, which takes its net debt from 2.4 to about 4.2 times adjusted EBITDA on the authors’ pro forma. Outside New York the listed brokers are networks and marketplaces, such as Steadfast Group (ASX: SDF) in Australia and PB Fintech (NSE: POLICYBZR) in India.

  8. Part IVThe PlayersThe longest records were built by managers who bought small agencies at a disciplined price and paid their people in stock, and every record depends on the multiple at which it began.

    Under J. Patrick Gallagher Jr., chief executive since 1995, Gallagher returned 15 percent a year against 11 percent for the S&P 500, and Greg Case’s Aon returned 14 percent a year from April 2005 against 11 percent. Robert Kelly’s Steadfast returned 15 percent a year in Australian dollars from its 2013 listing against 8.4 percent for its benchmark. The higher valuations of 2013 to 2022 were worth about 5 points a year of Dan Glaser’s 19 percent a year at Marsh, while John Doyle, who took over in January 2023 with the multiple near its peak, has returned 5.6 percent a year against 22 percent for the index.

  9. Part IVThe PlayersThe acquisition programs work because a small agency costs about 9 to 11 times EBITDA and is worth more inside a major, and private capital now competes for the same agencies.

    From 2010 to 2025 the five largest spent USD 78bn of cash on acquisitions, and Gallagher spent USD 31bn of it while its share count rose from 103 million to 257 million. An earnout adds about 12 to 14 percent of an agency’s price, and the US tax deduction for purchased goodwill returns about 1.1 times its EBITDA. Buyers backed by private equity have done 69 to 76 percent of the deals OPTIS counts every year since 2019. Without the savings it expected, Aon paid about 23 times USI’s adjusted EBITDA against its own 13 times, by the authors’ computation.

  10. Part VSpeaking the LanguageOrganic growth, adjusted earnings and fiduciary balances are each defined by the company that reports them, and each misleads a reader who takes it at face value.

    In the second quarter of 2026 Brown & Brown’s total revenue rose 30 percent with Accession, while its organic revenue fell 0.7 percent without contingent commissions and rose 0.7 percent with them. The listed brokers use seven definitions of organic growth, so a gap of less than a point between two companies is within what the definitions alone produce. Adjusted earnings add back the amortization of acquired customer lists, and Gallagher’s 2025 adjusted EPS of USD 11 was 86 percent above its GAAP EPS of USD 5.7. At the end of 2025, 38 percent of Gallagher’s balance sheet belonged to its clients and insurers.

  11. Part VIAnalysis & ValuationAt the October 21, 2026 close the market paid little for the brokers’ acquisition programs.

    The five largest traded at a median 14 times trailing adjusted EBITDA, from WTW’s 13 to Gallagher’s 17, and their median P/E on a uniform proxy had run from 14 at the end of 2009 to 29 at the end of 2021. A 5.5 percent free-cash-flow yield with 5 percent organic growth prices a return of about 11 percent a year before any acquisition. On the authors’ worked example, a dollar of agency EBITDA bought at 9 times costs 9.4 times net and is worth about 14 times inside a major after 5 percent attrition, a gain of 4.2 times, which turns into a loss at the 2009 trough multiple.

  12. Part VIAnalysis & ValuationThe brokers’ returns came either from buying revenue or from earning more on it, and the starting multiple dominated both.

    From 2010 to 2025 Gallagher returned 18 percent a year and Brown & Brown 15 percent, compounding revenue at 13 to 14 percent a year, roughly half of it acquired. Aon returned 16 percent on revenue growth of 4.8 percent a year and earned the rest from margin and buybacks, while the S&P 500 returned 14 percent. In the one soft market of the record, from the end of 2005 to the end of 2010, the five largest earned a median 1.9 percent a year against 2.3 percent for the index, as their earnings held and their multiples fell to the market’s.

  13. Part VIIThe AI InflectionAI has improved the brokers’ operations without yet reaching their revenue or their pay costs, and the exposure lies in small policies sold through forms.

    On February 9, 2026, the morning Insurify announced a car insurance comparison app inside ChatGPT, the five largest brokers lost about USD 25bn of market value while the S&P 500 rose. By October 21 Gallagher stood 7.0 percent above its February 6 close and Goosehead Insurance (NASDAQ: GSHD), which sells home and auto insurance through franchised agencies, 32 percent below it. Aon cut its invoicing cycle from 22 to 11 days, yet Marsh’s compensation ratio went only from 59 percent in 2019 to 58 percent in 2025. A commercial broker’s largest risk is an insurer paying less commission for less work, and no insurer had announced such a cut by October 22, 2026.

  14. Part VIIIRisks & DebatesA broker carries no underwriting loss, but it carries the placements it arranges and, when it is private, heavy debt.

    Of twelve large brokers that left the record between 1980 and 2026, nine were sold or merged, two were taken private and one, GoHealth, went bankrupt. Marsh recorded USD 569m of charges related to Greensill in the first nine months of 2026, 37 percent of its third-quarter adjusted operating income, and Aon USD 197m of settlement expenses after the collateral fraud at Vesttoo in 2023. S&P Global Ratings put the leverage of Acrisure at 9.6 times EBITDA at the end of 2025 against about six times on its lenders’ definitions, and on floating-rate debt each 100 basis points of interest costs a private broker 6 percent of EBITDA at 6 times leverage.

  15. Part VIIIRisks & DebatesThe open debates of October 2026 will be settled by numbers that the companies and the regulators publish by October 2027.

    The slowdown in growth is cyclical if Aon, Marsh Risk and WTW’s Risk & Broking hold 4 percent organic growth through two more quarters of falling property prices. The savings Aon reports against its plan will settle the USI purchase, and a priced offering will settle the confidential filing Hub International made in June 2026. Whether brokers’ pay becomes a regulatory risk again will be settled by Australia’s revised brokers’ code and by the final regulation of IRDAI, India’s insurance regulator.

  16. AppendicesThe appendices give the analyst the working tools: the investable universe of 33 listed companies worth about USD 325bn at the October 21, 2026 close, with the largest private brokers beside them, a checklist of 22 questions in the order of Chapter 21, a reading course of letters and filings with a listening course of 42 recordings of founders and chief executives, and the numbered notes that source the report.
Read the full report and hear the deep dive Premium members get the complete Insurance Brokers primer, every primer in the archive and each new one as it is published.
Industry Primer No. 8 · October 30, 2026

Robotics

How robot makers and their suppliers in Japan, Europe, China, Korea, Taiwan and the United States make money, why the best records belong to companies that keep earning from the machines already at work, and how specialists appraise the companies

Most of the industry’s value and most of its demand sit outside the United States. The 41 listed companies at the core of the industry were worth USD 1,134bn at the close of October 16, 2026, and about two-thirds of that value is listed outside the United States. China installed 59 percent of the 603,307 industrial robots installed in 2025, by the count of the International Federation of Robotics (IFR), while Japan still makes about 38 percent of the world’s robots. Component makers and arm makers each hold about 30 percent of the value of the listed core, and surgical robots about a quarter.

The best long-term records belong to companies that keep earning from the machines already at work. Intuitive Surgical (NASDAQ: ISRG), which takes about 84 percent of its revenue from instruments, service and leased systems, returned 23 percent a year from its 2000 listing. Fanuc (TSE: 6954), the largest maker of industrial arms, takes 17 percent of its sales from service and returned 5.2 percent a year in yen from January 2000. The report traces that difference through the history, the economics, the accounts and the valuation of the industry.

Listen to Part One freeThe story of the robot industry in twenty sections, from an inventor’s patent of 1954 and the first robot sold in 1961 to the builders of Japan, Europe, China and the United States and a humanoid maker’s listing in 2026, told for anyone.

Key findings15 findings across the report. Tap any one to expand.

  1. Part IOrientationA robotics company earns most reliably from what follows the sale, because the arm is at most about half of what a factory pays and the spare parts, service and consumables recur for as long as the machine runs.

    A factory buys a robot cell, the arm with its gripper, sensors, software and setup engineering, and the IFR puts the turnover of whole robot systems at about 2.7 times the USD 19bn of robots installed in 2025. Service on the machines already in use brought Fanuc 17 percent of its sales in the year to March 2026, and in surgery the instruments used in each procedure recur the same way. The industry is five businesses (industrial robots, components, warehouse systems, surgical robots, and service and humanoid robots), and each sells a long-lived machine or its parts and earns again from the installed base.

  2. Part IIEvolutionA robot industry grows where its customers install robots in large numbers.

    Unimation sold the first Unimate to General Motors (NYSE: GM) in 1961, but by 1982 American plants had installed about 4,000 robots in twenty years while Japanese plants installed 16,000 in ten, so American makers never had the buyers to build a supply chain. Japan built a chain of arms, servo motors, controllers and reducers that made it the world’s robot maker, and the downturn of 2008 to 2010, when Fanuc’s operating profit fell 71 percent, showed that the chain is a cyclical exporter. Europe’s car plants made KUKA and ABB (SIX: ABBN) two of the world’s largest arm makers, and both were later sold: KUKA to Midea Group (SZSE: 000333; HKEX: 0300) in 2016, and ABB’s robot business to SoftBank Group (TSE: 9984) under an agreement of October 2025.

  3. Part IIEvolutionChina has become the industry’s largest buyer and is building its own robot makers, but foreign makers still hold most of its car plants.

    China became the largest market in 2013 and raised its robot density, the number of industrial robots per 10,000 manufacturing employees, from 68 in 2016 to 470 in 2023, past Germany and Japan. Chinese makers supplied 57 percent of China’s installations in 2024 and 55 percent in 2025, as foreign makers grew faster in a record year for China’s car plants. In 2024 Chinese suppliers provided 90 percent of the robots installed in China’s metal and machinery plants and only 31 percent of those in its car plants.

  4. Part IIEvolutionThe case studies show that value lasts where a company earns from its installed base and disappears where it has none.

    Intuitive built a business paid on every procedure, and its shares rose 210 times from its listing. In 2014, after an FDA warning letter and a wave of product-liability suits, Intuitive’s revenue fell 5.9 percent and its service revenue still grew. iRobot sold the Roomba with no service business on the machines already in homes, and it ended in January 2026 owned by its contract manufacturer, with nothing for its shareholders.

  5. Part IIIStructure & EconomicsThe profit in the chain goes to the parts a customer cannot easily replace.

    Component makers serving all of manufacturing earn the most: in the latest fiscal year Keyence (TSE: 6861) kept 51 percent of its sales as operating profit, Airtac International (TWSE: 1590) 30 percent and SMC (TSE: 6273) 23 percent. Makers tied to robot orders mostly earned less than Fanuc’s 21 percent, Harmonic Drive Systems (TSE: 6324) only 4.3 percent. Intuitive kept 29 percent of its USD 10bn of 2025 revenue because it is paid about USD 1,900 of instruments on every procedure, and about 84 percent of its revenue recurs. An arm maker is protected by the cost of switching, since a plant would have to retrain its engineers on a new controller and replace its spares and service, and that protection held in China’s car plants.

  6. Part IIIStructure & EconomicsThe cycle is deep and shows first in the orders.

    Orders lead sales by one to two quarters, and Fanuc warns that long delivery times make customers order early, so orders run ahead of demand near a peak. The upturn of 2026 came from chip plants and data centers, and Yaskawa Electric (TSE: 6506) booked record orders of JPY 185bn for June to August 2026, up 38 percent. Fanuc’s company margin has run from 41 percent in the year to March 2015 to 17 percent in the year to March 2020, and Harmonic Drive’s fell from 14 percent to about zero in the year to March 2024.

  7. Part IVThe PlayersJapan’s founders built the industry’s strongest operating records, and their shareholder records turned on what they did with the cash.

    Fanuc returned 6.5 percent a year in yen before it doubled its payout and adopted its first buyback policy on April 27, 2015, against 0.2 percent for the Japanese market, and 3.2 percent a year since against 11 percent for a TOPIX fund. It again holds JPY 993bn of cash and securities, 18 percent of its market value. At SMC, Palliser Capital asked in 2026 for JPY 600bn of buybacks over two years, and the board resolved JPY 50bn. The best Japanese records belong to companies whose revenue reaches beyond the robot: Daifuku (TSE: 6383), a maker of warehouse systems, returned 15 percent a year in yen from January 2001 and Keyence 12 percent.

  8. Part IVThe PlayersOutside Japan, most founders who invented a new kind of robot sold early, and many of the largest makers now belong to owners who need no public return from robots.

    Universal Robots, the Danish maker of cobots that work beside people without a safety cage, went to Teradyne (NASDAQ: TER) for USD 285m plus up to USD 65m, and Kiva to Amazon (NASDAQ: AMZN) for USD 775m as announced, while Intuitive’s founders kept their company. KUKA reports inside Midea with no published margin, and SoftBank agreed to pay USD 5.4bn for ABB Robotics, about 3 percent of its own market value. The report judges every sale, purchase and cash pile by the value per share it left the owners who stayed.

  9. Part VSpeaking the LanguageThe industry’s numbers mislead until they are adjusted, because its statistics cannot be added together and its accounts differ by country.

    Installations are counted by the IFR, robots made by China’s statistics bureau, orders by the Japan Robot Association (JARA) and each maker, and humanoids on each maker’s definition. Japanese accounts understate what the owner holds, because the cash sits in three balance-sheet lines, and counting all three takes Fanuc’s enterprise value from 35 to 29 times its average operating profit over ten years. Chinese accounts overstate what the business earns, because grants and tax relief sit inside operating profit: other income was 80 percent of the 2025 pre-tax profit of Estun Automation (SZSE: 002747). In surgery and warehouses the contract decides when revenue appears, so the installed base, the backlog and the cash are the steadier record.

  10. Part VIAnalysis & ValuationRobotics companies are valued best on average operating profit over a full cycle with their cash counted, because the trailing P/E is highest when profit is lowest.

    Fanuc’s P/E ran from 13 at March 2009 to 53 at March 2021, and on the cycle measure with cash counted it trades at 29 times against a median of 37 times for the featured companies. Whole robot businesses changed hands at 1.5 to 2.4 times revenue against a listed median of 5 times. Each purchase is judged by what it left each share: after Estun bought Cloos of Germany, its revenue grew 3.3 times from 2018 to 2025 while its diluted earnings per share fell from CNY 0.12 to CNY 0.05.

  11. Part VIAnalysis & ValuationOver the long run, growth in revenue per share and the revenue from installed machines separated the winners, and net cash did not.

    Growth in revenue per share produced 91 percent of Intuitive’s return from the end of 2004, while 55 percent of Fanuc’s return from March 2008 came from a higher P/E as its net margin fell. In the two full downturns of 2008 to 2010 and 2018 to 2020, the arm and reducer makers lost from 55 percent to all of their operating profit and 50 to 80 percent of their share price, while Keyence and Intuitive lost share price through lower valuations and largely kept their profits. The listed companies’ records look better than the industry’s, because most of its failures were private or were sold.

  12. Part VIIThe AI InflectionAI already pays in conventional robots, while humanoids are valued far ahead of their revenue.

    The IFR grades machine vision as “already established and industry-grade” and general-purpose physical AI as “not yet industry-grade”, and AI pays today in vision, inspection and Amazon’s warehouse fleet routing. Private and listed humanoid makers are valued at about USD 120bn against about USD 235m of 2025 humanoid revenue, and Unitree Robotics (SSE STAR: 688836) sold 5,215 humanoids in 2025, mostly to research and education buyers. The only paid, multi-year humanoid deployment a customer has announced is that of GXO Logistics (NYSE: GXO) with Agility Robotics, of June 27, 2024. AI as a programming tool for conventional arms is already shipping and favors the makers with the most machines in use, so the report values a company with the humanoids earning nothing.

  13. Part VIIIRisks & DebatesRobotics companies fail in a few known ways, and each warning appears first in a public document.

    They fail when their machines leave nothing to service afterward, when they depend on one customer or one country, or when they stretch their balance sheets at a cycle peak. The warnings sit in the order book, the receivables, the customer note and the service share. Fanuc’s robot orders ran at 0.84 and 0.74 times its robot sales in the middle quarters of 2023 while its sales still held.

  14. Part VIIIRisks & DebatesThe open debates of October 2026 will be settled by dated numbers that anyone can read.

    The Japanese orders of October 2026 to March 2027 will show whether the highest orders in JARA’s series, booked in April to June 2026, were a new cycle or orders placed early. The IFR’s figure for China’s domestic share in 2026, due in September 2027, will show whether China’s makers take the car plants. Intuitive’s procedure growth, guided at 13.5 to 15.5 percent for 2026, will measure its contest with the surgical systems of Medtronic (NYSE: MDT) and Johnson & Johnson (NYSE: JNJ), which reached the US market in December 2025 and July 2026.

  15. AppendicesThe appendices give the analyst the working tools: the investable universe of 86 listed and 22 unlisted companies, a checklist of 35 questions, a reading course of letters and filings with a listening course of 45 recordings of founders and chief executives, and the numbered notes that source the report.
Read the full report and hear the deep dive Premium members get the complete Robotics primer, every primer in the archive and each new one as it is published.
Industry Primer No. 7 · October 23, 2026

Natural Resources Royalty & Streaming

How royalty companies make money from mines and wells they don’t operate, why the price paid and the share count decide the owners’ records, and how specialists appraise the companies

Listen to Part One freeThe story of royalty and streaming, from a railroad’s Texas land grant of 1888 and a Nevada gold royalty of 1986 to the first silver stream and the oil of the Permian

Key findings

The overview and key findings of this primer are open to every Latticework subscriber, free or paid.

Read the findings on Latticework
Read the full report and hear the deep dive Premium members get the complete Natural Resources Royalty & Streaming primer, every primer in the archive and each new one as it is published.
Industry Primer No. 6 · October 16, 2026

Fertilizers

How fertilizer producers make money from gas, rock and ore, why capital allocation through the cycle matters more than a specific asset, and how specialists appraise the companies

Fertilizers are three commodity businesses whose costs are set by geology and by the price of natural gas, and most of the listed value sits outside the United States. Nitrogen is made in chemical plants from gas, phosphate from rock dissolved in sulfuric acid, and potash is a salt dug from deep mines. The 23 featured companies on exchanges open to most investors were worth $248bn at the close of September 25, 2026, and four state-controlled Gulf producers made up 41 percent of it. Saudi Arabian Mining Company (Tadawul: 1211), known as Ma’aden, is about a quarter of the total on its own, while the largest company that makes only fertilizer is Nutrien (NYSE, TSX: NTR). China, India and Brazil used half the world’s fertilizer in 2024.

Bought without selection, the fertilizer majors split into one winner and the rest, and capital allocation through the cycle made the difference. From their listings to September 25, 2026, the four long-listed majors returned between 3.6 percent a year for The Mosaic Company (NYSE: MOS) and 21 percent for CF Industries Holdings (NYSE: CF) with dividends reinvested, against about 11 percent for the S&P 500 total return. Without CF the other three averaged about half the market’s return. Under Stephen Wilson, from its 2005 listing to the end of 2013, CF compounded 15 times against 1.8 times for the index. His successor, Tony Will, cut the share count by 45 percent, but he took over after the stock had risen about fifteen-fold in eight years, and his shareholders earned 7.5 percent a year against 14 percent.

Listen to Part One freeThe story of fertilizer in fourteen sections, from the guano islands of Peru and the first ammonia plant of 1913 to the potash mines of Saskatchewan and the closing of the Strait of Hormuz, told for anyone.

Key findings18 findings across the report. Tap any one to expand.

  1. Part IOrientationA producer earns the gap between the world price of its nutrient and its own cost, so its place on the cost curve decides its margin.

    A tonne of urea from Louisiana and one from Qatar fetch the same price at the same port. At the World Bank’s August 2026 gas prices a Louisiana plant made a tonne of ammonia for about $161 in all and a Dutch plant for about $757, and the price settles near the cost of the most expensive plants the world still needs, usually European ones. Canada, Russia and Belarus have mined about 63 percent of the world’s potash for twenty-five years, and Morocco and Western Sahara hold 68 percent of the world’s phosphate rock reserves.

  2. Part IIEvolutionNitrogen became a manufacturing business priced on the cost of making it, while phosphate and potash stayed with the owners of the deposits.

    Until 1913 every nutrient came from a natural deposit that a state owned or taxed, from Peru’s guano and Chile’s nitrate to Germany’s potash. The first synthetic ammonia plant, at Oppau in Germany, began production on September 9, 1913. On one study’s count, Chilean nitrate’s share of world nitrogen consumption fell from 55 percent that year to 32 percent a decade later. From 1945 to 2003 states and cooperatives built the mass industry, and the American gas shock of 2000 to 2003 bankrupted Farmland Industries and Mississippi Chemical.

  3. Part IIEvolutionEach price spike paid for capacity that arrived in the next trough.

    PotashCorp spent $9.8bn of its $14bn of capital for 2008 to 2016 in the five years 2009 to 2013, and the new tonnes arrived as prices broke in 2013 and reached their low in 2016. CF decided its one expansion in 2012 with urea at $399 a tonne and started the plants in 2016, when urea averaged $194. With the Strait of Hormuz effectively closed to commercial traffic since March 2, 2026, urea rose 82 percent from February to April, to $857 a tonne, and it had fallen 54 percent by August while the strait stayed shut.

  4. Part IIEvolutionThe price paid for capacity, set against its cost new and the point in the cycle, decided the owners’ outcome.

    In 2010 PotashCorp’s board rejected a cash offer of $130 a share, about $40bn, from BHP Group (ASX, LSE, NYSE: BHP). After PotashCorp merged with Agrium into Nutrien in 2018, a holder’s shares were worth 67 cents for each dollar BHP had offered at the September 25, 2026 close, before dividends. CF bought Terra Industries in the 2010 trough for $4.6bn, sold about 11 percent of its nitrogen business to the farm cooperative CHS in 2016 for $2.8bn, at an implied value twice CF’s own, and cut its share count to 42 percent of its count after the Terra purchase.

  5. Part IIIStructure & EconomicsThe producer’s share of the profit swings with the scarcity of its input, while the retailer’s share holds.

    From 2018 to 2025 the adjusted EBITDA of Nutrien’s retail segment varied with a coefficient of variation of 0.21, against 0.56 for its potash segment. The farmer buys once a season with the grain price in mind, and a tonne of urea cost 4.0 tonnes of maize in April 2026, the highest monthly reading since 2000, and 1.7 tonnes by August. Nitrogen is applied every year while phosphate and potash can be skipped for a season, which is why the demand cuts of 2026 fell on phosphate.

  6. Part IIIStructure & EconomicsA nitrogen plant on cheap gas keeps the difference between European and American gas, but only as far as the world still needs the European tonne.

    A tonne of ammonia takes 31 to 36 million Btu of gas, so the August 2026 gap of $18 per million British thermal units (MMBtu) was worth about $667 a tonne at CF’s gas use. CF earned gross margins of 46 to 51 percent in 2011 to 2013 but 35 to 39 percent on similar gaps in 2023 to 2025, because after the American plants of 2016 and 2017 the world needed Europe’s ammonia less. New plants, at about $2,600 to $2,900 a tonne of capacity, need at least about $450 to $505 a tonne of ammonia to earn their cost of capital on the report’s estimate, against the $521 CF realized in 2025.

  7. Part IIIStructure & EconomicsPotash earns a rent on mines paid for decades ago, phosphate’s margin belongs to whoever owns the rock, and the ore body is the industry’s one lasting advantage.

    In 2025 Nutrien kept about $161 of a net price of $252 a tonne of potash after its cash cost and Saskatchewan’s taxes, while BHP’s new Jansen mine will cost about $2,024 of capital for each tonne of annual capacity. On August 2026 prices a producer of diammonium phosphate (DAP) that buys its rock, sulfur and ammonia lost money, while one that mines its own rock kept about $173 a tonne more. In the 2016 low PotashCorp and Mosaic, which own mines, stayed profitable while CF lost money.

  8. Part IVThe PlayersThe featured producers sort by their nutrient’s cost curve and their place on it, and their accounts show the sort.

    CF earned operating income plus depreciation of 45 percent of its FY2025 sales on American gas, while Grupa Azoty (WSE: ATT), Poland’s state-controlled producer on European gas, earned a trailing EBITDA margin of 6.9 percent and carries net debt nearly six times its market value. In July 2026 Yara International (OSE: YAR) agreed to acquire the Gulf Coast Ammonia plant at Texas City, still in commissioning, for about $1,000 a tonne of capacity.

  9. Part IVThe PlayersCapital allocation decided how much capacity stood behind each share, and the start date decided whether the record beat the index.

    Measured against the S&P 500 total return over the same window, the three tenures that spanned the boom of 2003 to 2008 beat the index, among them Bill Doyle’s PotashCorp at 10 times against 1.9 from December 2000. Every chief executive’s window in the set that began in 2014 or later and lasted more than a year trailed it. Nassef Sawiris sold nearly all the plants of OCI N.V. (Euronext Amsterdam: OCI) and paid out about $7bn, yet a holder since the 2013 listing has 0.83 times the money in euros.

  10. Part IVThe PlayersThe state is a producer, a buyer and a gatekeeper, and its decisions move more tonnes than any company’s.

    Eleven of the 23 featured companies have a state or a state company as controlling or largest holder, and they carry 58 percent of the featured value. Producers inside the Strait of Hormuz shipped 27 percent of traded urea in 2024. The antitrust matter is two federal investigations, reported in March and announced in May 2026, and the farmer suits against CF, Nutrien, Mosaic, Koch and Yara entities and others, consolidated as MDL No. 3187 in Kansas. All are allegations the companies deny, no charge has been filed and no court has ruled on the merits.

  11. Part VSpeaking the LanguageMost arguments about a fertilizer company are arguments about the basis of a tonne or a cost, so every measure is read with its definition.

    Two analysts can quote the cost of the same potash mine from the same release and differ by 64 percent, because Jansen costs $2,024 per annual tonne of potassium chloride and $3,318 per tonne of potassium oxide (K2O), the unit of world statistics. Nutrien’s potash cash cost is $58, $65 or $91 a tonne, depending on what the definition leaves out, and CF reports its capacity in short tons while Mosaic prices its sulfur in long tons.

  12. Part VSpeaking the LanguageThe income statement hides three numbers a fertilizer analyst reads first: the reserve behind the depletion charge, the cost of closing mines and waste stacks, and the claims between EBITDA and the shareholder.

    Mosaic’s asset retirement obligations, its discounted cost of closing mines and stacks, rose from 8.8 percent of its equity in 2015 to 22 percent in 2025, and adding them, $2.6bn, to its enterprise value lifts its mid-cycle multiple of EBITDA from 5.7 to 6.9 times. International accounting rules let Nutrien reverse $780m of phosphate impairments in the 2022 spike, a gain a company reporting under US rules could not have booked.

  13. Part VIAnalysis & ValuationSpecialists value a producer on mid-cycle earnings and against the cost of a new tonne, because a single year’s multiple misleads at the top and the bottom of the cycle.

    At the end of March 2022 the majors traded at about 6 to 7 times the 2022 earnings they went on to report, and at the end of 2016 PotashCorp traded at 48 times its 2016 earnings. At the September 25, 2026 close the nine Western fertilizer producers traded at a median of 6.3 times mid-cycle EBITDA, and CF’s management uses a mid-cycle figure worth about $3.2bn more than CF’s eleven-year average. Against the cost of new capacity, Nutrien’s potash was valued at about 37 percent of Jansen’s cost per tonne and CF’s ammonia at 78 to 85 percent of the cost of Blue Point, its new Louisiana plant.

  14. Part VIAnalysis & ValuationA fertilizer stock bought without selection has trailed the index, and the entry price relative to the cycle predicted the record best.

    Over the twenty years from September 29, 2006, five of seventeen listed producers beat the index, and the median, Qinghai Salt Lake Industry (SZSE: 000792), returned 4.9 times, a little over half of it. From 2019 to 2025 a falling multiple cost every major 5.5 to 13 points of return a year. After the entry price came the cost position, restraint at the peak and leverage, while retail integration, size and a nutrient’s scarcity predicted nothing. The report treats the evidence as suggestive, because it rests on seventeen companies over one cycle.

  15. Part VIIThe AI InflectionThe AI that a fertilizer investor can count sits in Nutrien’s potash mines, and the larger risk to the producers runs through the price of American gas.

    Nutrien mined 53 percent of its potash ore with automation in the first half of 2026, against 22 percent in 2023, and the value lies in capital saved, since it prices its next tonnes from existing mines at $200 to $300 of capital against about $2,024 at Jansen. Data centers now compete for American gas, and a $2 rise in the Henry Hub price would take about a quarter of 2025’s gap between European and American gas, although the Energy Information Administration’s September 2026 outlook has Henry Hub falling to $3.3 in 2027.

  16. Part VIIIRisks & DebatesFertilizer producers fail when a high-cost position or a single asset meets a price collapse with debt, and shareholders more often lose through dividend cuts.

    The American cooperatives failed on gas in 2002 and 2003, and Sirius Minerals, after $1.1bn had been invested in its mine, was taken over by Anglo American (LSE: AAL) in 2020 for about £405m once its bond financing failed. No producer listed in North America or Europe in 2005 has failed since. Mosaic cut its quarterly dividend by 91 percent in seven months in 2017, and its undiscounted cost of closing its phosphogypsum stacks, about $3.0bn at the end of 2024, equals about 41 percent of its market value.

  17. Part VIIIRisks & DebatesThe open debates of 2026 will be settled by dated numbers, from potash capacity to the antitrust docket.

    On the US Geological Survey’s estimates, world potash capacity grows 17 percent by 2029, so output must grow about 4 percent a year to hold 2025’s operating rate. Urea fell 54 percent from its April peak with the strait still shut, so the debate over the Gulf producers turns on whether buyers will pay more for supply from outside the Gulf. In the antitrust debate, over allegations the companies deny, the plaintiffs allege that record prices drew no increase in supply, and CF answers that it ran at 97 percent of its gross ammonia capacity in 2025. The Justice Department, the Federal Trade Commission and the Kansas court will settle it.

  18. AppendicesThe appendices give the analyst the working tools: the investable universe of 69 listed companies, with the Moscow listings and the unlisted producers beside them; a checklist of 30 questions in ten groups, from the producer’s cost curve to the proxy statement; a reading course that opens with the producers’ own letters and calls; and the numbered notes that source the report.
Read the full report and hear the deep dive Premium members get the complete Fertilizers primer, every primer in the archive and each new one as it is published.
Industry Primer No. 5 · October 9, 2026

Homebuilding

How homebuilders make money, why the few builders that stopped owning land and kept their debt low made the industry’s great records, and how specialists appraise the companies

Most new American houses are built for sale by a builder that controls the lot, and half of them are permitted in seven Sunbelt states. Americans started 939,700 single-family houses in 2025, and 72 percent of them were built for sale in that way. The 34 listed companies the report features were worth about $217bn at the close of September 18, 2026, and 39 percent of that value sits outside the United States, most of it in Japan and Britain. D.R. Horton (NYSE: DHI), at $39bn, is the most valuable listed homebuilder among them. Demand is set by the buyer's monthly payment: at the mid-September 2026 mortgage rate of about 7 percent, a $400,000 loan costs $2,648 a month, and one point of rate changes the payment by $262.

Bought blind, the industry has lagged the market, and its great records belong to the few builders that stopped holding land with borrowed money. Kenneth French's construction portfolio, which mixes homebuilders with contractors, compounded at 8.7 percent a year over the century to July 2026, against 10 percent for the US market. Of the 27 builders and land developers listed in the United States in 2005, six failed in or through Chapter 11. Over the twenty years to September 2026, NVR (NYSE: NVR), which buys finished lots under option, was the only one of nineteen listed builders to beat the S&P 500 total return, with 13 percent a year against 11 percent. Over the same years Hovnanian Enterprises (NYSE: HOV) lost 8.9 percent a year, and its book equity was negative at every fiscal year end from 2010 to 2020.

Listen to Part One freeThe story of American homebuilding in fourteen sections, from the balloon loans of the 1920s and Levittown’s potato fields to the builders that now hold their land by option, told for anyone.

Key findings18 findings across the report. Tap any one to expand.

  1. Part IOrientationA builder's return on equity is its margin times its inventory turns times its leverage, and the terms after the margin decide most of the gap.

    In FY2025 NVR and D.R. Horton earned almost the same gross margin, 21 and 22 percent, yet NVR earned 33 percent on its average equity and D.R. Horton 15 percent. NVR's inventory turned 5.5 times in the year against 1.4 times at D.R. Horton, and at D.R. Horton's turns NVR would earn about 8 percent. The option also caps the loss: in the five years from 2006 NVR wrote off about $654m of deposits, 57 percent of its 2006 equity, and still earned a profit every year.

  2. Part IIEvolutionThe industry was built on the federally insured long mortgage, and every bust from 1973 to 1991 began when credit tightened.

    The Federal Housing Administration, created in 1934, insured long, fully amortizing loans and turned the price of a house into a monthly payment, so that a Levittown house of 1950 cost $7,990, or $56 a month for a veteran. A household with the same budget could borrow 44 percent as much at the rate of October 1981 as at that of April 1971. U.S. Home and NVR, the largest builders to fail in 1991 and 1992, failed on the debt behind their land and acquisitions while their buyers kept closing on houses.

  3. Part IIEvolutionThe bust of 2006 to 2011 broke the builders that owned land bought with debt.

    After 1992 the listed builders used the bond and stock markets to buy one another and to add land. Monthly single-family starts then fell 81 percent from January 2006 to March 2009, and the listed builders wrote off between 57 percent (NVR, deposits only) and 140 percent (Standard Pacific) of their 2006 equity. In November 2007 Lennar (NYSE: LEN, LEN.B) sold about 11,000 homesites carried at about $1.3bn for $525m, about 40 cents on the dollar. The six builders that failed held land concentrated in one or two states and bought with debt.

  4. Part IIEvolutionThe survivors moved their land into options, and outside owners are now buying the listed builders.

    Lennar controlled 19 percent of its homesites by option in 2010 and 98 percent at the end of fiscal 2025, after spinning off its land into Millrose Properties (NYSE: MRP). A mortgage rate of about 2.7 percent in January 2021 gave the builders the best margins in their history, 29 percent on D.R. Horton's home sales in fiscal 2022. Outside owners bought four listed builders in the thirteen months to July 24, 2026, when Berkshire Hathaway (NYSE: BRK.A, BRK.B) bought Taylor Morrison for about $6.8bn of equity.

  5. Part IIEvolutionThe four case studies show that a builder's results follow the financing of its land and of its buyers.

    NVR came out of its 1992 bankruptcy buying finished lots under option, and it has spent 109 percent of its net income since 2010 on buybacks. TOUSA's joint venture paid $826m for Transeastern Properties in August 2005, and TOUSA filed for Chapter 11 in January 2008. Britain's Help to Buy loans to buyers of new homes lifted the underlying operating margin of Persimmon (LSE: PSN) from 22 percent in 2015 to 31 percent in 2018, and the margin fell to 14 percent in 2023 as the subsidy ended. The buyers of China Evergrande paid before their apartments were built and became its largest creditors.

  6. Part IIIStructure & EconomicsThe profit on a new home goes to whoever holds the scarcest input, the landowner in a boom and the holder of cash in a bust.

    A builder prices the lot from what is left after its costs and its profit: on a $400,000 house the shares in the National Association of Home Builders' 2024 survey leave $55,200 for the lot, and if buyers can pay 10 percent more at the same construction cost, $86,480 is left. A 10 percent rise in the price thus lets the builder pay 57 percent more for the lot, and competing builders pass that gain to the landowner.

  7. Part IIIStructure & EconomicsDemand for new homes is set by the monthly payment, so the cycle runs through credit.

    The average 30-year mortgage rate rose 69 basis points in the year to September 17, 2026, so a household with the same budget can borrow about 7 percent less. In July 2026 the Census Bureau counted 9.6 months' supply of new homes, a level last seen in the bust, while estimates of the national shortfall run from 3.7 million to 5.5 million homes. The report reads the first as a local surplus at today's payments and the second as a national gap at today's prices.

  8. Part IIIStructure & EconomicsHolding little land multiplies the return on a builder's cash, yet homebuilding has no durable moat.

    In the 200-lot community the report works through, buying the lots under option at a margin one point lower cuts the peak cash from $18m to $5.9m and raises the return on that cash from about 25 percent a year to about 60 percent. The ten largest public builders doubled their share of Census new-home sales from 2005 to 2025, yet the two largest, D.R. Horton and Lennar, earned 15 and 8.3 percent on equity in FY2025. The one advantage that has lasted, holding less land and less debt than competitors, depends on the managers who keep to it.

  9. Part IVThe PlayersThe market prices the return a builder's land position earns, and much of the industry now answers to owners without a share price.

    At the September 18, 2026 close NVR, which owns 1.3 percent of the lots it controls, stood at 4.8 times book on a 33 percent return on equity, and Lennar, which owns about 2 percent, at 0.85 times on 8.3 percent. Eight of the fourteen American production builders traded below book. The private and outside-owned builders closed about 94,800 homes in 2025, about 14 percent of Census new-home sales, and the American homebuilding business of Sekisui House (TSE: 1928) lost ¥13bn on ¥414bn of sales in the six months to July 2026.

  10. Part IVThe PlayersThe great records were made by a few owner-operators who held little land with borrowed money at the top, kept cash for the bust and shrank the share count.

    Dwight Schar founded NVHomes in 1980, bought Ryan Homes with more than $450m of debt in 1987 and led NVR out of bankruptcy in 1993. NVR then compounded at 24 percent a year with dividends reinvested to Schar's retirement in May 2022, against 10 percent for the S&P 500 total return, and by June 2026 it had retired 87 percent of all the shares it ever issued. D.R. Horton returned 22 percent a year against 11 percent under David Auld as chief executive, from 2014 to 2023, and PulteGroup (NYSE: PHM) 20 percent against 15 percent under Ryan Marshall since 2016.

  11. Part IVThe PlayersThe leaders who cut their land early or bought it at the bottom kept their equity, and the test of a buyback is the price paid.

    Robert Toll said in December 2005 that the market had softened, and the company walked away from thousands of optioned lots. The shares of Toll Brothers (NYSE: TOL) fell 77 percent in the bust, against about 92 percent for the median measured builder. In February 2009 Berkeley Group (LSE: BKG) sold £50m of new shares to buy land after prices had fallen. Weighted by the dollars spent, Toll Brothers bought back its shares at about 1.3 times book at its year ends and NVR at 5.1 times.

  12. Part VSpeaking the LanguageMost arguments about a builder are arguments about a definition, and the accounts adjust book value in one direction.

    D.R. Horton's FY2025 gross margin ran from about 22 to 24 percent depending on the line read. Under US rules a write-down of land is never reversed, while the British builders reverse theirs, and interest on the money borrowed to carry land reaches the income statement only when the house closes. NVR's $920m of deposits controls at least $9.2bn of lots, none of them on its balance sheet. The filings also show the next year: NVR's backlog fell 15 percent in 2025, so 2026 opened with fewer homes already sold.

  13. Part VIAnalysis & ValuationA builder is priced on its book value and on the return that book earns, and a low earnings multiple misleads when the earnings come from land bought with debt.

    At the September 18, 2026 close a line fitted across fifteen US builders and land developers, NVR excluded, crossed one times book at a return on equity of about 11 percent. At the fiscal 2005 year end D.R. Horton traded at 7.7 times earnings, and its shares then lost 29 percent a year for three years. At the fiscal 2022 year end four of the five large builders traded at 3.7 to 5.5 times earnings with most of their lots under option, and they returned 17 to 47 percent a year over the next three years.

  14. Part VIAnalysis & ValuationBuyers of whole builders pay about book value, and the base rate for a builder bought at random is poor.

    Since 2017 buyers have paid 1.1 to 1.4 times book for a healthy builder, and Berkshire paid 1.1 times for Taylor Morrison, which earned 13 percent on its equity, about the multiple the market's line gives for that return. A buyer of a listed builder in September 2006 had about one chance in nineteen of beating the index by 2026, and that chance lay in the builder that owned the least land. The report's predictors of the winners rest on a handful of companies and one great bust, so it calls the evidence suggestive rather than statistical.

  15. Part VIIThe AI InflectionAI reaches the sales office and the mortgage before the building site, and at the one builder that publishes numbers the cost has come before the saving.

    The phrase "artificial intelligence" appears in 14 of the 19 annual reports for fiscal 2025 filed by the US builders, as a risk factor. Lennar's cycle time, the days it takes to build a house, fell to 116 days in the quarter to August 31, 2026 from 126 a year earlier, while its selling, general and administrative expense rose to 9.2 percent of home-sale revenue from 8.2 percent. Electricians' hourly pay rose 7.4 percent in the year to July 2026, against 5.2 percent across construction, as data centers competed for them.

  16. Part VIIIRisks & DebatesEvery homebuilding failure in the record is land held with borrowed money into a credit contraction.

    The joint venture, the condominium tower and the presale moved the debt to where the parent's balance sheet did not show it. The slower failures pass the loss to shareholders over years: Taylor Wimpey (LSE: TW.) raised about £510m net in May 2009 to repay debt and lost 2.7 percent a year in pounds over the twenty years to September 2026. Texas and Florida issued 28 percent of the country's single-family permits in 2025, so a slowdown in those two states now reaches most of the large builders at once.

  17. Part VIIIRisks & DebatesIn September 2026 the numbers a builder controls, its incentives and its deposit write-offs, were rising, while cancellations and debt stayed near their lows.

    Lennar's incentives ran at about 12 percent of the price, against the 4 to 6 percent Stuart Miller calls normal. NVR wrote off $76m of deposits in 2025 and about $22m more in the second quarter of 2026, while its cancellation rate fell to 15 percent from 17 percent. The report sets out eleven debates with the facts first and the proponents named, and it dates most of the numbers that settle them between October 20 and November 16, 2026.

  18. AppendicesThe appendices give the analyst the working tools: the investable universe of 94 listed companies in 17 countries, a checklist of 36 questions in eleven sections in the order of the analysis of Chapter 21, a reading course that opens with the builders' own letters and calls and continues with books, public records, interviews and the data series, and the numbered notes that source the report.
Read the full report and hear the deep dive Premium members get the complete Homebuilding primer, every primer in the archive and each new one as it is published.
Industry Primer No. 4 · October 2, 2026

Aerospace

How aerospace companies make money, why the owners of small certified monopolies in the supply chain compounded far faster than the airframers, and how specialists appraise the companies

Most of the value in aerospace sits with the companies paid over an aircraft's working life rather than at its delivery. At the September 18, 2026 close the forty companies featured in the report were worth about $2.3tn, and the six engine companies among them about $941bn, nearly twice the $499bn of the seven listed airframers. GE Aerospace (NYSE: GE), at $326bn, was worth about as much as Boeing (NYSE: BA) at $157bn and Airbus (Euronext Paris: AIR) at $176bn together. World air traffic, which drives demand, grew 6.2 percent a year from 1970 to 2019, and Boeing and Airbus held a backlog of about 16,100 large jets at August 31, 2026, about 12 years of 2025's deliveries.

Over the long run the industry beat the market, and inside it the owners of small certified monopolies in the supply chain compounded far faster than the airframers. Kenneth French's value-weighted portfolio of US-listed aircraft, engine and parts makers returned 13 percent a year from July 1926 to July 2026 against 10 percent for the market, and beat it in 63 of 80 rolling twenty-year windows, with deeper falls. Over the twenty years from September 29, 2006 to September 18, 2026, with dividends reinvested, TransDigm Group (NYSE: TDG) returned 27 percent a year, HEICO (NYSE: HEI, HEI.A) common stock 21 percent and Safran (Euronext Paris: SAF) 18 percent, against 11 percent for the S&P 500 total return. Airbus returned 12 percent and Boeing 6.5 percent, and five of the seven listed airframers sit in the bottom third of the report's twenty-year board.

Listen to Part One freeThe story of the aircraft industry in twenty sections, from the Wright brothers’ four flights of 1903 and William Boeing’s first seaplane to the families and managers who compounded fortunes at General Dynamics, HEICO, TransDigm and Precision Castparts.

Key findings17 findings across the report. Tap any one to expand.

  1. Part IOrientationAn airplane is sold once and flown for decades, and the three businesses in its chain are paid at different points in its life.

    The 737 first flew on April 9, 1967, and Boeing delivered 447 more in 2025. An airframe program spends tens of billions of dollars before its first delivery, and in 2025 Boeing Commercial Airplanes lost $7.1bn on $41bn of revenue. An engine maker sells the engine thinly and is repaid over the shop visits, the scheduled overhauls of later decades, and services were 75 percent of revenue at GE Aerospace's Commercial Engines & Services. The owner of a proprietary part, one it alone is certified to supply, prices it against the cost of a grounded airplane.

  2. Part IIEvolutionThe industry's structure was set by its first regulator, and it narrowed to a duopoly because each generation of airplane cost more than the last.

    The Air Commerce Act of May 20, 1926 made the government the licensor of every airworthy design, and the Air Mail Act of June 12, 1934 separated the airframer from its engine maker. Pan Am's order of 1966 for twenty-five 747s launched a program that took Boeing's payroll down to 38,690 in 1971. Rolls-Royce (LSE: RR.) went into receivership on February 4, 1971 on a fixed-price engine, and the engine makers then shared each program's risk with partners, so the three Western makers of large turbofans of 1970 are the three of 2026.

  3. Part IIEvolutionIn the century's first quarter Boeing lost its parity with Airbus through a loss of engineering control, and the constraint moved from demand to supply.

    The two makers split 17,140 deliveries almost exactly from 2003 to 2018, and Boeing's share has been between 22 and 43 percent in every year since 2019. The 737 MAX was grounded from March 2019 to November 2020, and after the door plug of January 5, 2024 the FAA froze 737 output at about 38 a month. The top of the chain is now buying back its constrained suppliers: Boeing bought Spirit AeroSystems in December 2025, and GE Aerospace agreed in September 2026 to buy the castings maker Consolidated Precision Products (CPP) for $12bn.

  4. Part IIEvolutionThe report's four cases show the industry's way to compound and its three ways to fail.

    TransDigm bought proprietary parts with large aftermarket content, priced them on their value, financed the purchases with debt and returned what the businesses did not need, and its EBITDA As Defined, its own non-GAAP measure, rose from $164m in fiscal 2005 to $4,760m in fiscal 2025. Its exposed flank is the price, which the Pentagon's Inspector General contested in 2019. Rolls-Royce failed in 1971 on a fixed-price engine sold to one customer, and Boeing's crisis grew from two decades of decisions on where engineering sat, what the regulator was told and where the cash went. Bombardier (TSX: BBD.B) built a good airplane in the CSeries and could not fund it.

  5. Part IIIStructure & EconomicsThe profit in the chain goes to whoever owns certified content that the fleet consumes.

    Three contractual facts decide the split: who buys the part (the airframer once, or the operator one unit at a time), whether the position is sole-source for the life of the aircraft type, and how much aftermarket content the part carries. FY2025 margins on each company's own measure run from a 17 percent loss at Boeing Commercial Airplanes and 10 percent at Airbus's commercial business through 21 to 27 percent at the engine makers' commercial units to 47 percent on GAAP operating income at TransDigm. The engine makers take two-thirds to three-quarters of their commercial revenue from services.

  6. Part IIIStructure & EconomicsTraffic compounds, orders swing far more than deliveries, and the backlog absorbs each shock.

    World revenue passenger kilometers rose from 460bn in 1970 to 8,664bn in 2019, fell 66 percent in 2020 and first exceeded 2019 only in 2024. The combined book-to-bill, orders over deliveries, ran from 0.42 in 2009 to 3.1 in 2005 and 2007, and 33,574 net orders against 23,109 deliveries from 2006 to 2025 built the backlog. The aftermarket carries the cycle, and Rolls-Royce's Civil services revenue rose 21 percent in 2025 on flying hours up 8 percent. On July 29, 2026 Aengus Kelly of AerCap Holdings (NYSE: AER) expected widebody retirements to rise, which lowers the demand for spares on the oldest types.

  7. Part IIIStructure & EconomicsThe three business models earn at opposite ends of an aircraft's life, and the moats that pay belong to the owners of sole-source parts and installed engines.

    Every clean-sheet airliner launched after the 777 that has delivered was late, by one and a half to three and a half years, or never delivered. On one definition the proprietary owners' returns on invested capital never fell below 11 percent from FY2015 to FY2025, while Spirit went from 29 percent in 2015 to an after-tax operating loss of 78 percent of its invested capital in 2024. The check on a sole-source price is the Parts Manufacturer Approval (PMA), a certified copy sold below the original maker's price, on which HEICO's Flight Support Group built $3.1bn of net sales in fiscal 2025.

  8. Part IVThe PlayersAirbus and Boeing hold the same position in the chain and have used it very differently.

    On €73bn of 2025 revenue Airbus earned a 9.7 percent EBIT Adjusted margin and ended the year with €12bn of net cash. Boeing's $4.3bn of GAAP earnings from operations included a one-time gain of $9.6bn, so the year was an operating loss of about $5.3bn without it. The position shows inside the conglomerates too, Pratt & Whitney at a 7.9 percent margin beside Collins at 16 percent within RTX (NYSE: RTX). In materials the Engine Products segment of Howmet Aerospace (NYSE: HWM) earned 33 percent on adjusted EBITDA, and Hexcel (NYSE: HXL), which builds to the airframers' rates, 9.1 percent.

  9. Part IVThe PlayersThe industry's great compounders are the builders and allocators of its supply chain, who bought or built small certified monopolies and priced them.

    Nicholas Howley's TransDigm returned 27 times with dividends reinvested from March 2006 to April 2018, 31 percent a year against 8.3 percent for the S&P 500 total return. At HEICO, Laurans Mendelson and his sons Eric and Victor built a company with fiscal 1990 sales of "just over $26 million" into one with about $4.5bn, and $100 invested on October 31, 1990 became $128,481 by October 31, 2025 on the company's own table, 23 percent a year. Mark Donegan's Precision Castparts returned 21 percent a year from March 2002 to March 2015, with earnings per share up 16-fold, before Berkshire Hathaway (NYSE: BRK.A, BRK.B) bought it for $235 a share in cash.

  10. Part IVThe PlayersThe best recent records belong to leaders who rebuilt companies around their installed base, and the levers behind the top records are few.

    John Plant separated Howmet from its aluminum rolling business in April 2020, cut its long-term debt by about half by 2025 and reduced its diluted share count from 463m to 406m, and Howmet has returned 42 percent a year since February 2019 against 16 percent for the S&P 500 total return. Larry Culp's GE returned 25 percent a year from September 2018. TransDigm alone publishes its hurdles: an acquisition must earn about 20 percent on the equity, and options vest fully only on 17.5 percent a year of growth in operating performance per share. Plant's letters print the price paid in every buyback.

  11. Part IVThe PlayersMuch of what sets the chain's terms has no share price.

    The French and German states each held 11 percent of Airbus at March 31, 2025, and the United States and the European Union fought over launch aid at the World Trade Organization from 2004 until arbitrators authorized countermeasures in 2019 and 2020. The FAA's freeze of 737 output in January 2024 made the regulator the judge of an airframer's output as well as its designs. Precision Castparts, inside Berkshire, reported $11bn of revenue in 2025, more than Howmet's $8.3bn, and the lessors own just over half of the commercial fleet by industry estimates.

  12. Part VSpeaking the LanguageMost arguments about an aerospace company are arguments about a definition, and the cash is the audit of the estimates behind the earnings.

    Each company labels its margin its own way, from Airbus's EBIT Adjusted to TransDigm's EBITDA As Defined, and an air-show tally is not a firm order: at Farnborough in July 2026 Reuters counted 173 aircraft for Boeing, against firm July orders of 38. Boeing still uses program accounting, which spreads a program's cost over the units it expects to deliver, and Rolls-Royce's 2017 operating profit was £1,175m under the old accounting standard and £321m under IFRS 15, with no change in cash. Boeing returned $64bn to shareholders from 2013 to 2019, 133 percent of its $48bn of free cash flow, then burned $38bn from 2019 to 2025 as its diluted shares rose from 565m to 762m.

  13. Part VIAnalysis & ValuationAn aerospace company is analyzed from the installed base upward, and the market pays for proprietary aftermarket margin and scarce capacity more than for aftermarket revenue.

    At the September 18, 2026 close the featured companies traded at a median of about 20 times FY2025 EBITDA, from 14 times for the airframers to 29 times for the proprietary-component makers. Across sixteen companies the aftermarket share explains little of the multiple, with an r-squared of 0.10, and a US listing counts about as much, a median of 26 times against 14 elsewhere. Strategic buyers paid 13 to 17 times trailing EBITDA from 2011 to 2025, and GE Aerospace's price for CPP, about 26 times expected 2027 EBITDA before synergies, breaks that band.

  14. Part VIAnalysis & ValuationThe base rates sort the industry by position in the chain, and the suppliers compounded on growth while the airframers rode margin recoveries that the next crisis reversed.

    French's portfolio beat the market by 2.4 points a year over the century but lost five of eleven decades. Across thirty-six listed companies over the twenty years to September 18, 2026 the median returned about 12 percent a year, from 17 to 27 percent at the proprietary-component makers to a median of 6.5 percent at the airframers. Revenue growth added 13 points a year to HEICO's return from FY2009 to FY2025, while Boeing's 25 percent a year from FY2009 to FY2018 turned into minus 5.1 percent a year to FY2025.

  15. Part VIIThe AI InflectionIn aerospace, artificial intelligence is so far a risk factor in the filings rather than a business line, and its largest effect has come through power demand for data centers.

    Across 27 FY2025 annual filings of the featured companies the median names artificial intelligence once, against about 44 uses of "cyber," and Curtiss-Wright (NYSE: CW) writes that it does not currently use AI in the design and development of its products. Certification keeps machine learning off flight-critical functions until about the end of the decade. The data-center build-out pulls on the same foundries as the jet engine, and Howmet's Gas Turbines revenue rose 25 percent to $944m in 2025 and 38 percent in the second quarter of 2026.

  16. Part VIIIRisks & DebatesNearly every aerospace failure in the record is a program bet larger than the balance sheet, a loss of engineering control or leverage carried into a traffic shock.

    The RB211 and the CSeries were bets of the first kind, and each failure showed first in the inventory, the customer advances, the regulator's letters and the production rate. Of the eleven debates the report sets out, the one the market waits on is the aftermarket: in the first half of 2026 the engine and parts makers grew by a fifth to a quarter, and on September 14, 2026 Melius Research forecast growth, as published reports of its note quote it, in "the high-single-digit percentage range next year." RTX's chief executive, Chris Calio, said of the engine business model on July 23, 2026, "I don't think that, that is the best model going forward."

  17. AppendicesThe appendices give the analyst the working tools: the investable universe of 82 listed companies along the chain, with the forty featured companies marked and the unlisted forces beside them; a due-diligence checklist of twenty-six questions in the order of the analysis of Chapter 21; a reading course of the companies' letters, proxies and investor days, the accident and regulatory reports, books, hearings and calls; and the numbered notes that source the report.
Read the full report and hear the deep dive Premium members get the complete Aerospace primer, every primer in the archive and each new one as it is published.
Industry Primer No. 3 · September 25, 2026

US Regional Banks

How regional banks make money, why the gap between the best operators and the industry’s own index is the widest in financial services, and how specialists appraise the companies

A regional bank buys money from depositors in its territory and lends it back there at about ten times leverage, so its value sits in the price it pays for that money. At June 30, 2026, 4,238 federally insured institutions held $26tn of assets and $21tn of deposits, and each filed the same quarterly Call Report whether it held $4bn or $700bn. The 109 listed regional banks and thrifts held $6.4tn of those assets, from U.S. Bancorp (NYSE: USB) at $726bn to Arrow Financial (Nasdaq: AROW) at $4.5bn, and were worth $871bn at the September 11, 2026 close. Across the forty-eight banks the report features, the cost of total deposits in fiscal 2025 ran from 0.26 percent at Westamerica Bancorporation (Nasdaq: WABC) to 3.5 percent at Live Oak Bancshares (NYSE: LOB), against a median of 1.9 percent.

The industry’s average return has trailed the market for twenty years, and the case for studying it rests on the gap between the best banks and the rest. Kenneth French’s value-weighted portfolio of American bank stocks compounded at 11 percent a year from July 1926 to July 2026 against 10 percent for the market, but at 6.5 percent over the twenty years to December 2025 against the market’s 11 percent. Over that window eight owner-operator records compounded tangible book value per share with dividends at a median of 11 percent a year, led by First Citizens BancShares (Nasdaq: FCNCA) at 15 percent. At the bottom the record ends at zero: SVB Financial, the parent of Silicon Valley Bank, compounded at 17 percent a year for thirteen years and went to nothing on March 10, 2023. Every variable behind that gap was observable in advance in free public documents.

Listen to Part One freeThe story of American banking in fourteen scenes, from a Philadelphia bank in 1782 to the Federal Reserve’s decision of September 2026, told for anyone.

Key findings19 findings across the report. Tap any one to expand.

  1. The report covers the listed American banks below the four money-center charters, from the ten super-regionals through the regionals and mid-caps to the specialists, and every figure in it is dated and sourced in the notes.

    These are the main findings of each part.

  2. Part IOrientationA bank’s return on equity is decided mostly by the choices a management makes on the liability side, because regulation sets the leverage.

    Equity was 10 percent of assets across insured commercial banks at December 31, 2025, against 6.1 percent at the end of 1984. Cullen/Frost Bankers (NYSE: CFR) held $43bn of deposits at that date, a third of them paying no interest, and earned a 3.7 percent margin and a 15 percent return on equity. The return breaks in three ways: credit through the provision, rates through the margin and the securities book, and liquidity when deposits leave. Statute draws the map, since at $10bn of assets a bank loses its uncapped debit interchange income and the $100bn and $250bn lines set the stress-testing and capital regime.

  3. Part IIEvolutionThe thousands of small American banks are a product of law, and deregulation in the wrong order produced the failures of the 1980s.

    A federal tax on state banknotes in 1866 pushed state banks onto deposits, and the McFadden Act of 1927 froze the branching map for sixty-seven years. The price of deposits was freed in 1980 and the powers to invest them in 1982, while the guarantee rose to $100,000. Continental Illinois lost its funding in a week in May 1984. From 1980 to 1994, 1,617 banks and 1,295 thrifts were resolved, and nine of the ten largest Texas holding companies failed or were sold, though in 1980 to 1982 they had been better capitalized than their peers by more than a quarter.

  4. Part IIEvolutionConsolidation since 1994 has been the middle of the industry absorbing the bottom, and each crisis has charged its cost back to the surviving banks.

    Ten super-regionals absorbed 490 charters between them. The crisis of 2008 to 2014 took 507 banks and drove the insurance fund to a negative $21bn. More than $40bn left Silicon Valley Bank on March 9, 2023; Signature Bank and First Republic Bank followed, and the industry paid a $16bn special assessment across 114 banking organizations. In the latest wave approvals take four to six months, and Banco Santander (NYSE: SAN) completed its purchase of Webster Financial for about $12bn on August 20, 2026.

  5. Part IIEvolutionThe case studies show banks dying of funding, underwriting and duration, and one survivor that set down in advance why it would avoid all three.

    Washington Mutual grew to about $300bn of assets through acquisitions and then the option adjustable-rate mortgage, and its failure cost its shareholders everything. Silicon Valley Bank reported a 16 percent total risk-based capital ratio at December 31, 2022, beside an unrecognized held-to-maturity loss of $15,158m against $11,847m of tangible common equity. M&T Bank (NYSE: MTB) under Robert Wilmers refused the loans everyone else made in 2006, bought distressed franchises in 2009 to 2011, and compounded tangible book plus dividends at 12 percent a year from 2005 to 2017 against 3.2 percent for the bank index.

  6. Part IIIStructure & EconomicsThe deposit is a bank’s product, and the speed at which its price follows the policy rate is a lasting trait of each franchise.

    Across the featured banks the cost of interest-bearing money ran from 0.50 percent at Westamerica to 4.0 percent at Flagstar Financial (NYSE: FLG) in fiscal 2025. In the tightening of 2022 to 2023 the industry passed half of the policy move into that price, a cumulative deposit beta of 0.50, while individual banks ran from 0.02 to 0.80. The same eight banks have sat in the bottom quartile across three cycles and twenty years. At the industry’s price Westamerica’s interest expense would have been about $77m a year higher before tax, against $116m of net income.

  7. Part IIIStructure & EconomicsThe average credit loss is survivable by design, and concentration is what turns a downturn into a failure.

    The industry’s net charge-off rate has averaged 0.71 percent a year since 1967 and peaked at 1.6 percent in 1991 and 2.6 percent in 2009. The interagency guidance of December 2006 flags construction lending above 100 percent of total risk-based capital. Bank OZK (Nasdaq: OZK) is the only featured bank above that line, at 120 percent, and Westamerica makes no construction loans at all. In 2026 the lines to watch are office loans, 3.1 percent delinquent at banks above $250bn against 1.3 percent at $10bn to $250bn, and lending to non-depository financial institutions, up 22 percent in a year to about $1.5tn.

  8. Part IIIStructure & EconomicsBanks reach similar returns by opposite routes, and the securities book is where the accounting and the economics part.

    Four banks worked in the report earn 15 to 16 percent on tangible common equity: Cullen/Frost on cheap deposits, Westamerica on a 40 percent efficiency ratio, Western Alliance Bancorporation (NYSE: WAL) on volume and fees, and U.S. Bancorp on fee income. Industry unrealized securities losses peaked at $690bn in the third quarter of 2022, and nearly every regional bank keeps them out of regulatory capital. At June 30, 2026 they stood at $327bn, measured before the rate increase of September 16, 2026.

  9. Part IIIStructure & EconomicsBanks are best sorted by what they pay for money and what they lend against, and only two competitive advantages can be measured from public data.

    One family pays 0.5 to 2.0 percent for money and takes little credit risk, the other pays 3.0 to 4.0 percent and earns it back on a concentrated book; both make 14 to 18 percent on tangible common equity in a good year, and only the first does so in a bad one. The two measurable advantages are a low cost of deposits and the community lender’s knowledge of local borrowers, which the FDIC’s 2020 study credits with 36 percent of the industry’s small-business loans from 15 percent of its lending.

  10. Part IVThe PlayersThe featured banks differ so widely that the industry cannot be analyzed as one business, and the market prices them by their funding.

    Across the forty-eight on one date, the net interest margin runs from 1.7 to 5.1 percent, the efficiency ratio from 32 to 93 percent and price to tangible book from 0.75 to 2.5 times. The tier predicts size and regulation and little about the model. Of the ten banks with the cheapest deposits, eight of the nine with a quoted multiple trade above the median, and of the ten that pay the most, seven of the eight trade below it.

  11. Part IVThe PlayersThe best records came from refusing to overpay for deposits, to concentrate or to grow in a hot market, and from buying failed banks from the government.

    Seven of the seventeen leaders profiled bought failed institutions from the receiver, the only transaction that adds tangible book at a discount to tangible book by construction. First Citizens bought about $72bn of Silicon Valley Bridge Bank’s assets at a $17bn discount on March 27, 2023. The group’s unpriced risk is succession: George Gleason has run Bank OZK for forty-seven years and David Payne Westamerica for thirty-seven, and neither has an announced successor.

  12. Part IVThe PlayersMeasured on one basis, the owner-operators beat their industry by a wide margin, though the panel shows the size of the gap among known compounders and not the odds of finding one.

    From December 31, 2005 to December 31, 2025, David Zalman compounded Prosperity Bancshares (NYSE: PB) at 13 percent a year and Johnny Allison Home BancShares (NYSE: HOMB) at 12 percent, and all eight records beat the bank portfolio’s 6.5 percent. Dividends carry much of it: on ending book alone Cullen/Frost compounded at 7.3 percent and Westamerica at 6.7 percent. Silicon Valley Bank’s pay plan held no metric that worsens as risk rises, and Westamerica’s holds two.

  13. Part IVThe PlayersThe balance sheets that set the industry’s prices file no proxy statement.

    The Deposit Insurance Fund held $161bn at a 1.5 percent reserve ratio at June 30, 2026. Federal Home Loan Bank advances rose from $827bn at December 31, 2022 to more than $1tn at March 31, 2023. The 4,250 tax-exempt credit unions hold $2.5tn and bought sixteen banks in 2025. After the 2024 merger policies were rescinded in 2025, approval times fell from about fifteen months to about four.

  14. Part VSpeaking the LanguageNo bank figure can be compared until its basis is named, and reported capital and book value are the product of three accounting choices.

    Cullen/Frost’s fiscal 2025 cost of deposits is 1.89, 1.87 and 1.85 percent on three defensible bases. The three choices are the securities classification, the size of the reserve and the treatment of the marks: the industry’s allowance went from 1.2 percent of loans at the end of 2019 to 2.2 percent at the end of 2020 and 1.6 percent at the end of 2025, while realized losses moved the other way. First Citizens’ $9.8bn bargain-purchase gain is a transfer from the insurance fund, not earnings a multiple can be applied to.

  15. Part VIAnalysis & ValuationPrice to tangible book reflects the quality of a bank’s funding more than its return on equity.

    The analysis runs liability side first, because every failure in the record began with funding. Across ninety-five listed banks at the September 11, 2026 close, the fitted line is 0.94 plus 0.0635 times the return on tangible common equity in percent, with an R² of 0.26. Glacier Bancorp (NYSE: GBCI) traded at 2.0 times tangible book on an 8.1 percent return, while Western Alliance earned 13 percent at 1.2 times. The 2025 and 2026 deals were struck at 1.5 to 1.9 times tangible book and 6.6 to 8.6 percent of deposits.

  16. Part VIAnalysis & ValuationWhat a bank did with its capital explains little of what its owners earned over twelve years, and most exits from the industry were sales rather than failures.

    Over the twelve years to December 31, 2025 the median featured bank grew tangible book plus dividends at 9.8 percent a year and returned 7.6 percent to its shareholders, as an entry multiple of 2.1 times tangible book compressed. Of roughly 18,463 insured charters alive at the end of 1984, 3,451 still operated in September 2026, and most of the exits were sales at a premium. The winners shared cheap and stable deposits, no concentration, a moderate entry multiple and long tenure.

  17. Part VIIThe AI InflectionArtificial intelligence has not yet shown up in the featured banks’ costs, and its largest risk sits in their deposits.

    The median featured bank’s efficiency ratio is 1.4 points better than in 2019, while the ten super-regionals are 1.3 points worse. JPMorganChase (NYSE: JPM) reported $11bn on its tagged technology line in fiscal 2025, against $11m at Westamerica, so regional banks rent their systems. No bank publishes what its deposits would cost if every customer used a tool that moved balances to the best insured rate each week. Regions Financial (NYSE: RF) is the one featured bank that discloses data-center lending, which rose from 2.2 to 6.0 percent of its investor real estate book in 2025.

  18. Part VIIIRisks & DebatesBanks fail through credit, rates or liquidity, each visible in the filings two to eight quarters ahead, except for fraud.

    The warnings are loan growth far above peers, concentration ratios, unrealized losses against common equity tier 1, and uninsured deposits beside brokered funding. At least three failures since the March 2023 cohort were frauds. In the second quarter of 2026 insured institutions earned $90bn at a 1.4 percent return on assets, with forty-seven banks on the problem list. On September 16, 2026 the Federal Open Market Committee raised the target range to 3.75 to 4.00 percent, against $327bn of unrealized securities losses. Of eleven debates, the report finds three will move returns: the margin as rates rise, commercial real estate, and private credit.

  19. AppendicesThe appendices give the analyst the working tools: the investable universe of 109 listed banks and thrifts by tier and region, a due-diligence checklist of forty-two questions in seven sections with a free source for each, a reading course of letters, transcripts, regulators’ post-mortems, books and data sources, and the 312 numbered notes that source the report.
Read the full report and hear the deep dive Premium members get the complete US Regional Banks primer, every primer in the archive and each new one as it is published.
Industry Primer No. 2 · September 18, 2026

Property and Casualty Insurance & Reinsurance

How property and casualty insurers make money, why most celebrated (re)insurance owner-operators trailed the index over the last decade, and how specialists appraise the companies

Most of the industry’s capital sits in balance sheets that cannot be bought, behind a listed universe that one company dominates. The US industry wrote about $1.1tn of direct premiums in 2024, about $514bn of it in personal auto and homeowners, and closed the year with policyholders’ surplus of $1,132bn. Behind the primary carriers stood $800bn of global reinsurer capital at June 30, 2026. The largest American writer, State Farm, has no shares; it wrote $94bn of direct premium in 2023, 9.9 percent of the market. Of the forty-four listed majors the report features, Berkshire Hathaway (NYSE: BRK.B) alone was about 38 percent of the market value at the September 4, 2026 close, and without it Continental Europe’s four reinsurers, worth about $172bn between them, form the largest bloc.

The listed survivors beat the index by about a point a year, and the companies that did not survive pull the industry’s true base rate lower. Nineteen listed US insurers returned a median 12 percent a year with dividends reinvested to September 4, 2026, against a median 11 percent for the S&P 500 over the same windows, and fourteen of the nineteen were ahead. Kinsale Capital (NYSE: KNSL) compounded at 35 percent a year from its July 2016 listing against 15 percent for the index, while AIG (NYSE: AIG) returned minus 1.8 percent a year over about 34 years. Of roughly twenty companies formed in the Bermuda classes of 1993, 2001 and 2005, four were still independently listed on that date, and the insolvencies paid their shareholders nothing.

Listen to Part One freeThe story of property and casualty insurance and reinsurance in fourteen scenes, from a London coffee house to February 2026, told for anyone, in about an hour and a half.

Key findings18 findings across the report. Tap any one to expand.

  1. Part IOrientationUnderwriting results in this industry diverge more widely than in almost any other financial business, and the advantages behind them can be read in public documents.

    In December 1951 a twenty-one-year-old Warren Buffett pointed out that GEICO had earned an underwriting profit of 28 percent of premiums in 1949, against 6.7 percent for 135 stock casualty and surety companies, because it sold direct and had no agents. In FY2025 GEICO, part of Berkshire since 1996, ran an 85 percent combined ratio on a 12 percent expense ratio, against a statutory 92 for the US industry. Kinsale reported 76 percent, RLI (NYSE: RLI) 84 percent in its thirtieth consecutive year of underwriting profit, and Travelers (NYSE: TRV) 90 percent.

  2. Part IOrientationAn insurer collects premiums before it pays claims and invests the money in between, so the combined ratio sets the price of its borrowing.

    The report calls an insurer a leveraged investment vehicle whose leverage is priced by underwriting. Berkshire held about $176bn of this float at December 31, 2025 and earned $9,460m of pre-tax underwriting profit on it, a cost of minus 5.5 percent a year. Return on equity is the underwriting margin times premium leverage plus the investment yield times asset leverage, less tax. Progressive (NYSE: PGR) earned 35 percent in FY2025 from a 13 percent margin at 2.9 times premium leverage, and RenaissanceRe (NYSE: RNR) 26 percent from a similar margin at 0.96 times, its return coming from the investment side.

  3. Part IIEvolutionThe industry’s history repeats one event, in which a correlated loss arrives, carriers learn they wrote more exposure than they knew, and capital is destroyed and then returns at a higher price.

    The Chicago fire of 1871 broke 68 of the 182 companies carrying its risks. San Francisco in 1906 wiped out the American industry’s profit of the preceding forty-seven years, while Munich Re (Xetra: MUV2) paid its 11 million marks in full. The professional reinsurers grew out of such fire losses: Cologne Re in 1846, Swiss Re (SIX: SREN) in 1863 and Munich Re in 1880. The liability century repeated the event slowly, and when incumbents stopped quoting in the mid-1980s, corporate buyers capitalized ACE with $285m and X.L. with about $410m in Bermuda.

  4. Part IIEvolutionEach modern catastrophe destroyed capital and drew in new capital in a newer form, and few of the new companies lasted.

    Hurricane Andrew left eleven insurers insolvent in 1992, and new Bermuda classes raised $4.8bn in 1993, $8.5bn in 2001 and $13bn in 2005, nearly $5bn of the last in sidecars that could be wound up after one year. Most of their companies were later acquired, taken private or wound down. The survivors wrote less when prices fell: National Indemnity‘s premium fell from $366m in 1986 to $55m in 1999, and RLI has made an underwriting profit for thirty consecutive years at an average combined ratio of 88.

  5. Part IIEvolutionThe case studies show that a cost advantage does not prevent a solvency crisis, and that a published underwriting rule outlasts the people who wrote it.

    GEICO nearly failed in 1976 through its reserving and pricing, and Berkshire’s $46m for a third of the company by 1980 became $2.3bn for the other half in 1995. Progressive has published a 4 percent underwriting-profit objective since its 1971 public offering and still calls the resulting 96 combined ratio a constant. AIG failed three ways in 2008, only one of them in insurance, and its insurance companies stayed solvent throughout. Lloyd’s took twenty-one years to move from the first loss of its reinsurance spiral in 1988 to legal finality in 2009.

  6. Part IIIStructure & EconomicsEvery intermediary between the insured and the reinsurer is paid whether or not the business makes money, and the profit pool has moved toward distribution.

    A premium can pass through seven stations, from the retail agent to the investor in insurance-linked securities. The US industry paid 11 points of net written premium in commissions and brokerage in 2023 and ran a 25 percent statutory expense ratio. In FY2025 the distributors earned margins of 23 to 37 percent on revenue that carries no underwriting risk, while the carriers earned 7.8 points of underwriting margin in a decade-best year, after losing money underwriting in 2022 and 2023.

  7. Part IIIStructure & EconomicsThe cycle exists because an insurer sets its price before it knows its cost, and the record results of 2025 came on falling prices.

    The US industry ran a statutory combined ratio of 92 in 2025 against 102 in 2023, with a net underwriting gain of $61bn against $22bn in 2024. Guy Carpenter’s property catastrophe rate-on-line index, which is not risk-adjusted, rose 28 percent at the January 2023 renewals and fell 16 percent for 2026, to 23 percent below its 2024 peak. Reserve releases from the hard years subsidize the prices of the soft ones, and US other liability booked $10bn of adverse development in calendar 2024 and $7.3bn in 2025.

  8. Part IIIStructure & EconomicsSegment mix explains more of the dispersion in returns than management skill does, and few structural moats survive the record.

    In FY2025 the same return identity gave W. R. Berkley (NYSE: WRB) 21 percent on 16 points of investment income per dollar of equity and Chubb (NYSE: CB) 15 percent at 0.77 times premium leverage. The US industry ran at or above a 100 statutory combined ratio in four of the nine years to 2025, and of the four candidate moats the report keeps, only underwriting culture and specialization has repeatedly lasted a full cycle. Low-cost distribution is worth the spread between GEICO’s 12 percent expense ratio and 31 percent in the Business Insurance segment of The Hartford (NYSE: HIG).

  9. Part IVThe PlayersThe forty-four listed majors report under three accounting regimes, and none of their combined ratios can be compared with another’s until its basis is named.

    US GAAP filers strike the ratio on net earned premium and discount nothing, while IFRS 17 filers strike it on insurance revenue and usually discount. Munich Re’s 74 percent is struck on net insurance revenue, Sompo Holdings (TSE: 8630) reported 82 percent and Lancashire (LSE: LRE) 84 percent on a discounted basis, and Fairfax Financial (TSX: FFH) reported 93 percent explicitly undiscounted.

  10. Part IVThe PlayersThe long-tenured owner-operators beat the index decisively over forty and sixty years, but the median owner-operator trailed it over the decade to 2025.

    Berkshire compounded per-share market value at 20 percent a year from 1965 to 2025 against 11 percent for the S&P 500, and Prem Watsa compounded Fairfax’s book value per share plus dividends at 19 percent a year over forty years. Over the decade to 2025, on that basis, three of ten allocators beat the index’s 15 percent: Kinsale at 27 percent, Progressive at 23 percent and Arch Capital (NASDAQ: ACGL) at 17 percent. Four of the twelve owner-operated companies the report compares pay the long-term award on per-share book or tangible book value.

  11. Part IVThe PlayersThe largest US writer has no shares, and the mutuals and the state set the price the listed companies must meet.

    Four of the ten largest US groups are mutual or reciprocal, and their surplus has no cost of equity to meet, so they return it as price: State Farm began paying a $5bn cash-back dividend on more than 49 million vehicles on July 31, 2026. Florida’s Citizens fell from 1,407,805 policies in September 2023 to 266,093 on August 28, 2026, while California’s FAIR Plan levied a $1bn assessment in February 2025. Private capital enters mostly for fees, through managing general agents, fronting carriers and $145bn of third-party reinsurance capital at June 30, 2026.

  12. Part VSpeaking the LanguageThe calendar-year combined ratio is the figure management can shape, and the accident-year triangles that every US filer publishes are the honest record.

    Prior-year development lands in the year an estimate changes, not in the year the business was written. Travelers reported net favorable development in nine of the ten calendar years to 2025, yet its general liability triangle shows $1,593m of adverse development on accident years 2016 to 2024, netted in the headline against workers’ compensation releases. Discounting opens a further gap between regimes: about 9.5 points of Munich Re’s second-quarter 2026 reinsurance combined ratio of 69 percent was discount benefit.

  13. Part VIAnalysis & ValuationPrice to book is the industry’s native multiple, and a low one has usually been a verdict either on the return or on the reserves.

    AIG, at 0.98 times book on September 4, 2026, reported a 7.5 percent return on equity for FY2025. James River Group Holdings (NASDAQ: JRVR) traded at 0.37 times on a 93 percent excess and surplus lines combined ratio, a verdict on its reserves. At the other end, Kinsale’s book value per share compounded at 34 percent a year from December 2022 to June 2026 while its price to book fell from 8.1 to 4.2 times, so its shares returned 11 percent a year.

  14. Part VIAnalysis & ValuationThe variables that separated the listed winners were visible in advance, while over four- and five-year windows the change in the multiple was the largest term in the return.

    From 2021 to 2025 Travelers earned 10 points a year from a rising multiple against 6.0 points of book value growth. The highest multiples on September 4, 2026 belonged to companies with the lowest cost of acquiring a policy, Admiral (LSE: ADM) at 7.7 times book, Kinsale at 4.2 and Progressive at 3.7. Buybacks below book added value only where the return on that book exceeded the cost of equity, and AIG, which bought back heavily, finished last on the leaderboard.

  15. Part VIIThe AI InflectionThe countable effects of AI in this industry are operating facts, and none has yet moved an underwriting ratio.

    AIG reported that its AI tool raised quoted submissions 30 percent and cut time to quote 55 percent in Lexington‘s middle-market property business, and Travelers made more than half of its claims eligible for straight-through processing. The expense ratio, where the change was expected, rose instead: Progressive’s from 20 to 22 percent over the decade to 2025, and GEICO’s from 9.7 percent in 2024 to 12 percent in 2025. Claims savings land in loss adjustment expense, inside the loss ratio, and no carrier publishes the loss ratio of an AI-underwritten cohort, the test that would settle the argument.

  16. Part VIIIRisks & DebatesInsurers fail on three timescales, and each failure has shown first in published disclosures.

    Under-reserved long-tail casualty takes five to twenty-one years, as at Mission Insurance, liquidated in 1987, and Reliance, liquidated in 2001. Catastrophe concentration takes about two years: seven Florida property carriers were ordered into liquidation between March 2022 and February 2023. The third mechanism, a holding company’s borrowing while its insurers stayed solvent, has happened once, at AIG. Everest Group (NYSE: EG) ran the casualty sequence in 2025 and survived it, with $1.7bn of reserve strengthening announced in January, a further $478m of adverse development in October, and a $1.2bn adverse development cover over $5.4bn of reserves.

  17. Part VIIIRisks & DebatesThe industry entered September 2026 on its best results in a generation, earned on falling premium and a quiet weather year.

    No hurricane made landfall in the US in 2025, and Europe’s four large reinsurers averaged a 20 percent return on equity. Fitch’s non-life reinsurer cohort improved its reported combined ratio in the first half of 2026, but its accident-year ratio excluding catastrophes worsened 2.3 points while premium fell 6 percent. Of the twelve debates the report sets out, the cycle divides S&P Global Ratings, which expects sector returns on equity of 12 to 15 percent in 2026, from Howden Re, which warns that another leg down would take large segments below the cost of capital by 2027.

  18. AppendicesThe appendices give the analyst the working tools: the investable universe of ninety-four listed underwriters grouped by segment, with tickers, listing venues, market values at the September 4, 2026 close and one-line descriptions, and the six brokers beside them; a due-diligence checklist of 44 questions in ten sections, in the order of the analysis of Chapter 21; a reading list of books, shareholder letters and investor materials, interviews and podcasts, and data sources; and the sources for the report.
Read the full report and hear the deep dive Premium members get the complete Property and Casualty Insurance & Reinsurance primer, every primer in the archive and each new one as it is published.
Industry Primer No. 1 · September 11, 2026

Exchanges & Financial Market Infrastructure

Research report and companion slide deck on how exchanges and financial market infrastructure businesses make money, what their long-term returns show, and how to appraise their value.

A few dozen listed companies run the machinery through which the world’s securities and derivatives trade, clear and settle, and their combined value is small beside the flows they carry. Roughly 45 listed companies were worth about $700bn in aggregate at the end of August 2026, less than one mega-cap technology company. CME Group (Nasdaq: CME), the most valuable at $103bn, and Intercontinental Exchange (NYSE: ICE), at $90bn, hold the main derivatives franchises, while London Stock Exchange Group (LSE: LSEG), at about $59bn, is by revenue a data company. The Depository Trust & Clearing Corporation (DTCC), a user-owned utility, processed more than $3.8 quadrillion of securities transactions in 2024. CME’s fee of about $0.61 on an E-mini S&P 500 future carrying about $300,000 of notional is about one-fiftieth of a basis point, and CME earned a 65 percent operating margin in FY2025.

Since these businesses became investable, the median major venue has beaten the stock market by a wide margin, but the best and the worst outcomes differ by a factor of about 80. Fourteen major venues returned a median of about 17 percent a year since their IPO or demutualization, against about 11 percent for the S&P 500 from CME’s December 2002 listing. At the first day’s close, $10,000 invested became about $1.2m at Hong Kong Exchanges and Clearing (HKEX: 388) but only about $15,000 in dollars at the predecessors of Brazil’s B3 (B3: B3SA3). The two best performers, Euronext (Euronext Paris: ENX) and Japan Exchange Group (TSE: 8697), were the cheapest at listing, at about 11 and 7 times earnings. An unscreened buyer earned high single digits, while owners of captive clearinghouses bought at moderate multiples under capable allocators earned 16 to 20 percent or better.

Listen to Part One freeA narrative history of how the world’s exchanges, futures markets and clearinghouses came to work the way they do, told through the people who built them.

Key findings19 findings across the report. Tap any one to expand.

  1. Part IOrientationAn exchange group earns from three activities on one platform, and the durable profit sits in clearing trades and selling data rather than in matching them.

    The group matches buyers and sellers for a fee, clears the trades through a central counterparty that guarantees both sides and holds their margin, and sells the record of prices. CME held $160bn of clearing-member cash at the end of 2025, up $61bn in a year. In US cash equities, where regulation separated the three activities, 56 percent of volume executed off-exchange on August 31, 2026, and the $4.7bn of FY2025 gross revenue at Cboe Global Markets (Cboe: CBOE) shrank to $2.4bn after payments to liquidity providers, pass-through fees and index royalties.

  2. Part IIEvolutionFor most of their history exchanges were member clubs that protected brokers’ commissions, and screens and new owners turned them into companies that outside shareholders could own.

    The Buttonwood Agreement of 1792 fixed a minimum commission of a quarter percent that lasted 183 years, until May Day 1975. Chicago invented nearly every franchise of today’s derivatives business between 1972 and 1982. From 1993 to 1998 the electronic DTB, later Eurex, took the Bund future from the London floor of LIFFE, rising from about 24 percent of volume to above 99.9 percent. Demutualization followed, and NYSE seats that traded at $975,000 in January 2005 reached $4.0m that December.

  3. Part IIEvolutionMergers between exchanges in different countries almost always fail, while vertical, data and software acquisitions close.

    CME acquired the CBOT for $12bn and NYMEX for $8.9bn, ICE acquired NYSE Euronext for about $11bn in 2013, and LSEG acquired Refinitiv for $27bn in 2021. In 2026 ICE agreed to acquire MarketAxess (Nasdaq: MKTX) for about $6bn, and Deutsche Börse (Xetra: DB1) agreed to acquire Allfunds (Euronext Amsterdam: ALLFG) for €5.3bn. Regulators or politics stopped four cross-border mergers between 2011 and 2017. That record is why the industry remains divided into national and product groups rather than one global exchange.

  4. Part IIEvolutionA futures franchise whose open positions sit in its own clearinghouse has not lost its market since 1998, while regulation took most of an equity exchange’s trading away by design.

    The Bund moved because the challenger had order flow sponsored by its owners, an all-in cost advantage of about 3 percent and open access, a combination no later challenger has had. ICE grew from Jeffrey Sprecher’s 1997 purchase into a $90bn company by buying networks for their data at disciplined prices. Regulation took the NYSE from about 80 percent of trading in its own listings to a family share of about 23 percent. A trader’s default at the clearinghouse of Nasdaq (Nasdaq: NDAQ) in 2018 and the nickel cancellation at the London Metal Exchange (LME) in 2022 showed that members bear the losses and shareholders the reputational cost.

  5. Part IIIStructure & EconomicsMatching trades earns thin margins wherever a product can trade elsewhere, while clearing and data earn wide ones.

    Cboe’s exclusive SPX option earns $0.92 a contract against $0.066 for an option listed on many venues, about 14 times as much. Clearing is paid twice, through fees and through interest on margin balances, and LSEG paid £1.2bn in 2025 only to extend its share of SwapClear’s economics through 2045. Cboe’s Data Vantage produced $636m of FY2025 revenue against $13m of segment cost of revenue. Acquirers now pay 8 to 16 times EBITDA for matching businesses and 20 times or more for data.

  6. Part IIIStructure & EconomicsVolatility drives exchange volumes, and four decades of records confirm that volumes rise in market stress.

    CME’s volume in January 2008 ran 65 percent above the prior year, and 2022 brought record volume, up 19 percent, in a year when both stocks and bonds lost money. Rate per contract follows product mix rather than a price management sets: CME’s runs from $0.49 for interest rate contracts to $1.4 for agricultural ones, and volume rather than pricing drove about 90 percent of its FY2025 fee growth. India’s SEBI showed that regulation can outweigh volatility by cutting the income of the National Stock Exchange (NSE) 2.4 percent in FY2026.

  7. Part IIIStructure & EconomicsAn exchange group’s costs are small and mostly fixed, so its earnings move much faster than its revenue, in both directions.

    ICE converted incremental net revenue into operating income at 95 percent in FY2025, HKEX at 93 percent, Nasdaq at 89 percent and CME at 76 percent. The leverage works in reverse as well: MarketAxess’s operating margin fell from 48 percent in 2021 to 41 percent in 2025 as growth stalled against a cost base built for more. Capital spending runs at 4 to 6 percent of revenue and free cash flow often exceeds 90 percent of net income, although a clearinghouse’s capital cannot be paid out.

  8. Part IIIStructure & EconomicsThe lasting advantage in this industry is open interest held in an exchange’s own clearinghouse, and every well-funded attack on a benchmark futures franchise since 1998 has failed.

    A trader who moves a position elsewhere must close and reopen it, post margin twice and give up netting. Six challengers backed by dealers or other exchanges attacked benchmark futures with fee cuts, and all six failed. The live test is FMX, backed by BGC Group (Nasdaq: BGC), which clears US rates futures at LCH so that they offset against SwapClear’s swaps, and whose futures open interest has grown fast from a very small base. CME carries $17bn of trading-product intangibles that it never amortizes.

  9. Part IVThe PlayersThe listed companies sort by whether their clearing is captive, how much of their revenue recurs, and whether a policy grant protects the franchise.

    CME took 81 percent of its $6.5bn of FY2025 revenue from transactions, while about 60 percent of ICE’s revenue and more than 70 percent of LSEG’s recurs. HKEX rests on its exclusive Stock Connect gateway to mainland China, and India, the largest derivatives market by contracts, is the most exposed to its regulator: the NSE filed in June 2026 for an IPO at up to about $60bn, while BSE (NSE: BSE) and MCX (NSE: MCX) traded at 48 and 56 times trailing earnings.

  10. Part IVThe PlayersThe industry’s two best capital allocators followed opposite strategies and earned nearly the same return, because neither let its clearinghouse leave the company.

    Jeffrey Sprecher’s ICE spent about $55bn on acquisitions and compounded at about 18 percent a year from its 2005 IPO. Terry Duffy’s CME made essentially two large deals in twenty years, refused to clear rivals’ contracts, and returned more than $28bn since 2012 through a variable dividend, compounding at 20 percent a year since 2002. At Deutsche Börse, Werner Seifert lost his job in 2005 when TCI and Atticus, holding under 10 percent between them, forced him to drop a £1.4bn bid for the London exchange.

  11. Part IVThe PlayersThe largest institutions in market infrastructure cannot be bought, and their charters decide where the listed companies’ profits form.

    DTCC is the sole clearer of US equities, LCH’s SwapClear cleared $1,941tn of swap notional in 2025, and Euroclear holds more than €43tn in custody and sits on about €190bn of immobilized Russian central-bank assets. Clearing the same trade supports margins of 60 to 80 percent where the exchange owns the clearinghouse and roughly zero where a user-owned utility clears at cost, as DTCC and the options clearer OCC do.

  12. Part VSpeaking the LanguageExchange accounts mislead the casual reader in three places, and monthly disclosures let a careful one estimate revenue before it is reported.

    Cboe reports $1.9 of revenue for every dollar it keeps, so a model starts from net revenue. Clearing-member collateral sits on both sides of the balance sheet, which is why CME’s total assets grew from $137bn to $198bn in one year, and it comes out before any return is computed. In FY2025 adjusted earnings exceeded GAAP earnings by 20 percent at ICE, 13 percent at Nasdaq and 76 percent at LSEG, almost all of it amortization of acquired intangibles. Transaction revenue is average daily volume times trading days times rate per contract, and the volumes are published monthly.

  13. Part VIAnalysis & ValuationIn August 2026 the sector traded near the middle of its twenty-year valuation range, and entry price has decided ten-year outcomes as much as the quality of the franchise.

    The median of eighteen companies was 25 times trailing earnings and 16 times EV/EBITDA, and S&P Global (NYSE: SPGI), at 17 times EV/EBITDA, traded below CME at 22 times. MarketAxess closed at $590 on December 22, 2020, about 75 times FY2020 earnings inflated by a burst of credit volatility. Its revenue then grew every year, yet a buyer at that close who holds to ICE’s $167 cash offer of July 2026 loses about 20 percent a year over 5.6 years.

  14. Part VIAnalysis & ValuationCME’s long return came from growth in revenue per share, a one-time rise in margins and the payout, with little from a higher multiple.

    CME’s 20 percent a year since 2002 splits into roughly equal thirds: per-share revenue growth of 8.4 percent a year, the margin build and the payout, with the change in multiple adding only about two points a year. ICE returned 17 percent a year from a first-day close at about 53 times earnings, because its revenue per share compounded at 18 percent. Captive clearing was the best single predictor of returns, and the entry multiple was worth as much as the choice of franchise. From mid-range multiples, the report’s base rate for the median franchise is about 8 to 12 percent a year.

  15. Part VIIThe AI InflectionSo far AI brings exchanges new channels of order entry and internal efficiency, and the orders that software agents place still pass through the exchanges.

    As of mid-2026 there is no credible public evidence that language models earn excess returns as autonomous traders. Robinhood Markets (Nasdaq: HOOD) opened its platform to outside agents on May 27, 2026, and their orders route through its brokerage to the exchanges. Nasdaq’s Verafin is the only AI business in the sector with disclosed results, with revenue up 22 percent in FY2025 and 750 clients using its agent features by July 2026. The matching engines stay closed to AI by design, and the costs come first: ICE raised its 2026 capital spending to about $850m.

  16. Part VIIThe AI InflectionExchange data is the one input to financial AI that cannot be synthesized, and the most countable gain so far is hedging demand from the AI economy.

    No exchange had licensed a dataset for model training as of August 2026. Google (a unit of Alphabet, Nasdaq: GOOGL) holds $1bn of CME convertible preferred stock and Microsoft (Nasdaq: MSFT) took about 4 percent of LSEG. ICE’s US power futures volumes rose 30 percent to a record 7.8bn MWh in 2025, CME planned to launch futures on GPU rental rates on October 5, 2026, and ICE describes its $1.6bn investment in Polymarket as a data acquisition.

  17. Part VIIIRisks & DebatesInvestors in market infrastructure rarely lose money to bankruptcy, and the report finds six ways they do lose it.

    The first is a loss at a clearinghouse: in 2018 one trader’s default consumed €107m of a €166m default fund at Nasdaq Clearing, and in March 2022 the LME, with nickel at $101,365 a tonne, canceled every trade since midnight. The others are regulation that opens matching to competition, long calms and windfall earnings priced as permanent, technology failures such as the abandoned CHESS rebuild at ASX (ASX: ASX), acquisitions at about 30 times EBITDA, as Nasdaq paid for Adenza, and clearing without intermediaries, the model of Hyperliquid, which cleared $633bn offshore in the first quarter of 2026.

  18. Part VIIIRisks & DebatesThe industry entered late 2026 with record results on volumes driven by retail trading rather than a market shock, and seven debates will shape the next decade.

    Prediction markets, where Kalshi was valued at $22bn in May 2026, await a Supreme Court ruling on whether event contracts are federally regulated swaps or state-regulated gambling. DTCC, ICE and CME compete for about $4 trillion a day of Treasury clearing, and the SEC proposed in June 2026 to rescind the trade-through rule with 56 percent of US equity volume already off-exchange. Same-day options reached a record 66 percent of SPX volume in July 2026 without a test through a discontinuous move. Tokenization, consolidation in Europe and the volume booms in India and Hong Kong complete the seven.

  19. AppendicesThe appendices give the analyst the working tools: the investable universe of roughly 45 listed companies by segment with tickers, listing venues, market capitalizations and one-line descriptions, a sixty-question due-diligence checklist ordered to match the analytical procedure of Chapter 18, a reading list, and the sources for the report.
Read the full report and hear the deep dive Premium members get the complete Exchanges & Financial Market Infrastructure primer, every primer in the archive and each new one as it is published.

Get the full library