Warren Buffett

See’s Candies, 1972

Documented Bought through Blue Chip for $25 million. Sales about $30 million; pre-tax earnings under $5 million (2007 letter). The seller asked $30 million; Buffett later wrote he almost walked away above $25 million. By 2007 sales were $383 million and pre-tax profit $82 million, on about $40 million of capital, so incremental capital since 1972 was about $32 million. He has said the business produced well over $1 billion, later over $2 billion, of pre-tax earnings. Saw: western boxed-chocolate brand with pricing power and honest operators. Others missed: a high multiple of tangible book was cheap against untaxed cash that did not need reinvestment. Valuation: roughly mid-single-digit pre-tax multiple, but the logic was a reinvestment rate near zero plus price increases. Risk: regional brand, cocoa, key person. Today: separate maintenance capital from growth capital before calling a consumer brand a compounder.

Washington Post, 1973

Documented Cost about $10.6 million. The 1985 letter says the purchase was at not more than one-fourth of then per-share business value of a dominant franchise. Catalyst was time and cash. Miss: the franchise was not perpetual. Later letters treat newspaper economics as impaired by the internet. A correct 1973 valuation can still be a bad 2005 hold if the moat thesis dies.

GEICO, 1951 / 1976 / 1996

Documented 1951: Davidson explained the direct-writer cost advantage. 1976: after a reserving crisis, Berkshire bought convertible preferred and then common while the market treated the franchise as impaired. 1995–96: remaining shares for about $2.3 billion. Others missed in 1976: a low-cost producer with a damaged balance sheet is not a high-cost producer with a damaged balance sheet. Risk: growth ahead of capital and dishonest reserves. Float is an asset only if the combined ratio and reserving are disciplined.

Coca-Cola, 1988–94

Documented Accumulation began after the 1987 crash. Cost basis about $1.3 billion. Saw: global brand, bottler system, pricing, high incremental returns under Roberto Goizueta. Missed by cigar-butt training: a mid-teens multiple can be cheap if owner earnings compound for decades. Risk: valuation and volume, not insolvency. He did not sell when the late-1990s multiple was too high. That non-sale is a hold decision, not proof that price never matters.

BNSF, 2010

Documented Announced November 2009, closed February 2010, $100 a share in cash and stock. The 2010 letter describes a purchase on the order of $34 billion of cash and stock; assumed debt took the commonly cited enterprise cost toward $44 billion. Saw: an irreplaceable western rail whose replacement cost exceeded the price, with a long-haul advantage versus trucks. Risk: regulation, unions, fuel, cyclical volumes, and paying partly with Berkshire stock — the Dexter error in form. Capital intensity is acceptable when incremental returns and replacement barriers are both high.

Bank of America, 2011

Documented $5 billion of 6% preferred plus warrants on 700 million shares at $7.14. Asymmetry was in the contract: income if the bank survives, equity upside if confidence returns. Risk: another crisis in a leveraged balance sheet.

Apple, from 2016

Filings First disclosed in 2016, on the order of 61 million shares by year-end, later expanded. Saw: a consumer franchise with switching costs, an installed base, services attach, and buybacks. See’s/Coke logic, not a venture-technology conversion. Risk: regulation, China supply chain, premium multiple. Large sales in 2024 reduced the stake while it remained a top holding. Interpretation: trimmed for concentration and tax, not a declared false thesis.

Float, from National Indemnity 1967

Documented Premiums held before claims are paid became the funding advantage. Letters show float growing into the tens and then over $100 billion, often at negative cost when underwriting profits. Rule inside the model: float is cheap leverage only while reserves are adequate. Gen Re after 1998 is the counter-example he wrote about.

Intrinsic value, in his usage, is the discounted value of cash that can be taken out during the remaining life. He refuses false-precision DCF in the letters and still requires the idea of one.

• Partnership letters, 1957–1969: generals, workouts, controls, and the decision to shut the partnership.
• 1977 letter: acquisition criteria — size, demonstrated earning power, good returns on equity with little debt, management in place, simple businesses, an offering price.
• 1983 Owner’s Manual, later restated: partnership attitude, per-share progress, reluctance to issue shares.
• 1986 letter: owner-earnings appendix.
• 1989 letter: wonderful company at a fair price; institutional imperative.
• 1992 letter: look-through earnings.
• 2007 letter: Great, Good, and Gruesome, with See’s as the great business. Dexter called the worst deal.
• 2013 letter: Energy Future Holdings — about $2 billion of bonds, $837 million interest received, sold for $259 million, pre-tax loss about $873 million.
• 2020 letter: airline sale described as a mistake in assessing the industry.
• Berkshire 10-K equity and insurance footnotes for position size and float. Letters outrank meeting transcripts when they conflict. Biographies are secondary.

Asian Paints — write the memo before you open the solution

1. Circle: dealer density, tinting machines, working-capital cycle, input pass-through.
2. Moat versus Berger, Kansai Nerolac, and new organised players. What would a rival need to take 500 basis points of volume?
3. Owner earnings: profit after tax + depreciation − maintenance capex you can defend. Separate depot growth capex. Incremental return on that growth capex for five years.
4. Allocation: dividend, buyback, capacity, promoter behaviour.
5. Price: owner-earnings yield versus a 10-year government bond and versus the multiple Berkshire paid for Coke in 1988–89, not the multiple Coke later traded at.
6. Pass condition: the price or fact that makes you refuse a wonderful business.

1. Primary source. Read the 1989 or 2007 Berkshire letter end to end. Extract one principle, one number, one named mistake, and one rule you will use. Ten lines.
2. Owner earnings. For one company you own or follow, compute owner earnings. State maintenance capex as an assumption and a range. Do not use adjusted EBITDA.
3. Sell journal. For your largest holding, write what would make you sell. If the only answer is a lower quote, rewrite it until the answer is a broken moat, broken allocation, or a clearly better use of the capital.

1. Berkshire chairman’s letters, 1977, 1986, 1989, 1992, 2007, 2013, 2020 — berkshirehathaway.com.
2. Buffett Partnership letters, 1957–1969.
3. An Owner’s Manual, as restated on the Berkshire site.
4. Lawrence Cunningham, The Essays of Warren Buffett — a topical index of the letters, not a substitute.
5. Latest Berkshire Form 10-K, float footnote and equity-investment table.

Week 01 is the base layer. Nothing earlier is on the system. Carry four tests into every later investor: Is it a business or a security? What is owner earnings after maintenance capital? Where is the circle’s edge? What did the investor write down as an error?

Week 02 is Charlie Munger. Expected connection: he is the pressure that moved Buffett off cigar butts. Expected disagreement: he will price a great business more willingly, and will treat a lattice of models as the defence against a single accounting ratio. Do not merge them yet. Keep Buffett’s price discipline as the constraint Munger has to clear.

Sequence: Buffett → Munger → Graham → Fisher → Lynch → Templeton → Marks → Klarman → Greenblatt → Pabrai → Li Lu → Nick Sleep → Jhunjhunwala.

Back to series hub