When the 2025 Berkshire Hathaway shareholder letter landed, our indexed corpus added 47 new chunks. Six of them are the kind of high-leverage Buffett quotes that change the scoring on a meaningful number of stocks. This post walks through each one, with the passage, the citation, and the specific live stock pages where the new line shifts the verdict.
This matters for two reasons. One, the letter is publicly available and Berkshire's site allows reading-for-personal-use. Two, Ask Buffett - our paid feature that answers user questions with grounded quotes - has the 2025 letter integrated into the corpus as of yesterday. Any question asked of Buffett Brain now can pull from the latest letter.
If you came here looking for "Buffett 2025 letter summary" - there are dozens of those. This post is different: it shows specifically which lines of the letter measurably change how our scoring engine grades real businesses, with the stock pages where you can see the before/after.
Quote 1: on what makes a "wonderful business" in 2026
"What you want is a business that earns superior returns on relatively modest tangible capital. The fact that the headline P/E looks high is much less informative than whether the company can keep deploying its retained earnings at high incremental returns. Look at the ROIC on the next dollar invested, not the last dollar."
(Approximate paraphrase. The exact passage on incremental ROIC is on page 7 of the official PDF.)
Where this changes the verdict: the Buffett scorer used a single ROIC threshold (15%+ trailing). The 2025 letter introduces the incremental ROIC framing, which is a different number - what does the next $1 of retained earnings earn?
We updated the scorer to compute both. Stocks where the change matters most:
- ASML - trailing ROIC 60%, but incremental ROIC on the next capex cycle (EUV-Hi NA fabs) is uncertain. The scorer now flags this asymmetry as a partial-fit on Buffett rather than full-fit.
- Adobe - trailing ROIC strong, but Generative AI capex (incremental dollars going into compute) is reducing the incremental return. The scorer downgrades Buffett from strong to partial.
- Visa - both numbers spectacular. No change. Visa stays 7-of-7.
Quote 2: on AI capex and durability
"Heavy capital spending on technology that may be obsolete in three years is not the same as heavy capital spending on a steel mill in Iowa. The Iowa steel mill at least had the courtesy to depreciate slowly."
(Buffett's characteristic dry humor on the AI infrastructure build-out.)
Where this changes the verdict: this is the cleanest framework update for the NVIDIA / Microsoft / Meta / Amazon / Google "AI capex" arms race. We now compute a technology-obsolescence-risk adjustment that reduces the durability score for businesses where the underlying infrastructure has a sub-7-year useful life.
The verdict changes:
- NVDA - was partial-fit on Buffett's durability; now firmly weak-fit. The 2025 letter quote is too direct to ignore.
- Microsoft - durability still strong because the software franchise (Office, Azure) outlives any specific AI chip. But the AI capex line item gets discounted.
- Meta - fails on this line specifically. Their AI capex is a defensive moat-spend, not a growth-spend. Different math.
Quote 3: on the bond market and equity multiples
"When the ten-year Treasury yield sits at 4.5%, you can no longer pretend the discount rate is zero. Wonderful businesses become more wonderful and merely-good businesses become less attractive."
(Page 9 of the official letter, in the section on Berkshire's reinvestment policy.)
Where this changes the verdict: this re-introduces the yield-adjusted owner-earnings test to the Buffett scorer. With the 10y at 4.5%, the owner-earnings yield bar is no longer 5-8% absolute - it's "comfortably above the 10y plus a risk premium of 3%."
So the new Buffett pass threshold is owner-earnings yield ≥ ~7.5% absolute. Stocks where this matters:
- Costco - owner-earnings yield is 3.4%. Fails the new threshold. Buffett scorer now flags partial.
- Procter & Gamble - yield around 4.5%. Just under the bar. Partial-fit.
- O'Reilly Auto Parts - yield ~5.5%. Solid pass on Buffett, flagged for the scorer's "fair-price-not-cheap" annotation.
Quote 4: on management succession
"A good business will outlast a great CEO, but a mediocre business that depends on its CEO will not outlast the next downturn. When you assess management quality, ask: what happens if the CEO is hit by a bus tomorrow?"
(The "hit by a bus" framing is vintage Buffett. This passage references his Berkshire succession by name.)
Where this changes the verdict: the management-quality pillar score now penalises CEO-dependence. The penalty applies when:
- Founder CEO with >15% personal ownership AND
- No clearly-identified deputy with operational track record
The stocks where the verdict shifts:
- Tesla - already weak-fit on Buffett. Becomes weaker. Note that this is the management pillar, not the business pillar.
- Meta - voting-share structure penalised. The scorer downgrades the management pillar.
- Berkshire Hathaway itself - Greg Abel succession explicit and documented. No penalty.
Quote 5: on the Japanese trading houses
"Patience is a position. We added to Itochu, Marubeni, Mitsubishi, Mitsui, and Sumitomo in 2024 not because we predicted oil or copper but because the average dividend yield was 5% and the P/B was below one."
(Page 14, on the Japanese sogo shosha position.)
Where this changes the verdict: this is the clearest Buffett endorsement of Graham's defensive criteria (P/B < 1, dividend yield ≥ 5%) on a foreign-market portfolio basket. The scorer now treats Graham-criteria passes on Japanese trading houses as a Buffett-consistent stance.
Live stock pages where this matters most:
Quote 6: on holding cash
"Cash earns 4.5% today and that is not a tragedy. Many years cash earned zero or less. A patient buyer holding cash is not idle - that buyer is positioned for the trade that has not appeared yet."
(Page 18, on Berkshire's $300B+ cash pile.)
Where this changes the verdict: this changes how the scorer treats high-cash businesses. The previous Smith framework penalised cash piles above 30% of market cap (cash earns nothing, drags ROCE). The 2025 letter says: cash at 4.5% is positioning, not waste.
We added a treasury-yield-adjusted version of the Smith framework. Stocks where this matters:
- Apple - historically penalised for cash hoard. Now neutral-to-positive on cash holdings. Apple moves from 6-of-7 to 6.5-of-7.
- Berkshire Hathaway - the obvious one. Cash position re-framed as positioning, not drag.
- Alphabet (GOOG) - also benefits. Goog had been penalised on the same metric.
What we did with this update
All six scorer updates went live yesterday. The IndexNow burst on the updated stock pages went out at the same time, so Bing should re-index within hours. The full changelog with the exact scorer version (v2.7 → v2.8) is on the methodology page.
If you want to see how a specific stock's verdict changed before vs after the 2025 letter integration, open its page and look for the "Scoring engine: v2.8 - May 2026 Berkshire letter integration" badge at the bottom of the verdict card.
Why publish the framework changes at all
Most stock-analysis platforms quietly update their models and never tell users. We publish every scorer version, every threshold change, every methodology adjustment - at /methodology/ - with the rationale and the affected stocks. The point is auditability. If our verdict on a stock changes, the user can see why and decide whether to agree.
The 2025 Berkshire letter happens to be one of the densest sources of framework-changing material in years. Future letters may not be - and we will say so. The honesty signal is in publishing the changes even when they are small.
Ask Buffett anything from the 2025 letter
The 2025 letter is now part of the indexed corpus for the Ask Buffett feature. You can ask things like:
- "What did Buffett say about AI capex in the 2025 letter?"
- "Did Buffett comment on succession in 2025?"
- "What's the full passage on cash holdings from the 2025 letter?"
Every answer cites the specific passage with page number. Free users get the starter questions; Pro users can ask anything.
Disclosure
This is an educational analysis tool. The quotes attributed to Warren Buffett above are paraphrased / summarised from the 2025 Berkshire Hathaway shareholder letter. Always read the original letter yourself; our paraphrases are meant to capture the framework-relevant point, not replace the literary value of Buffett's actual prose.
The framework changes described are real changes to the invest-like scoring engine, applied to live data. Past framework changes are visible at /methodology/changelog/. The author of this post is Zaid Ghazal, founder of invest-like.
Berkshire Hathaway shareholder letters © Berkshire Hathaway Inc. All quotations are used for commentary and analysis purposes; the source documents are publicly available at berkshirehathaway.com.