CAT Learning Companion: How to read the Caterpillar research

The CAT research family (page counts are those of the approved PDF editions, which this site does not host):

• Deep Dive: A great franchise priced beyond its modeled economics (44 pages). Public preview; the Deep Dive page says how to read the full report.
• Research Note: The business has moved up a level, the price has moved further (6 pages).
• Technical Appendix: How the valuation was built and tested (20 pages).
• Learning Companion: How to read the Caterpillar research (10 pages).

---

The Deep Dive says what we think of Caterpillar. This edition explains how the reasoning works, using the same numbers, so that you can check our logic and disagree with it where you have better information. Nothing here is a recommendation to trade. The Blockhaven Rating is our own assessment, on our own scale, of risk and reward at a stated price and horizon: for Caterpillar it is 1/5 POOR at $613.62, set against intrinsic value at the 11-02-2026 valuation date, and it is not a forecast of twelve-month price performance.

One ticker, two kinds of business

Caterpillar's reports split the company into segments, and the split carries most of the analysis. Construction Industries and Resource Industries supply machines through dealers into construction and mining, and their earnings rise and fall with the cycle. Power & Energy supplies engines and turbines, and in the third quarter of 2026 it earned a segment margin of 24.4%, against 13.48% in Resource Industries. Financial Products is a captive lender that finances customers and dealers. The company also shows a subtotal, Machinery, Power & Energy (MP&E), that leaves the lender out, and that is the measure we value as an industrial business.

A first lesson in reading segment tables: segment profit is not operating profit, and segment sales include sales between segments. The third quarter shows how the pieces fit. Segment profit was $2.14 billion in Power & Energy, $1.83 billion in Construction and $622 million in Resource Industries; subtract the All Other segment's small loss ($2 million) and the charge for corporate items and eliminations ($450 million) and you reach MP&E profit of $4.14 billion. Adding Financial Products ($261 million) and subtracting the consolidating adjustments ($191 million) gives consolidated operating profit of $4.21 billion. When a table does not add up on a first look, the missing line is usually one of those two.

From profit to cash

Profit is an opinion that accounting has formed about a year; cash flow is what is left after the business has paid to keep itself running. We build free cash flow from operating profit by taxing it, adding back depreciation (a cost that does not use cash today), and subtracting the capital spent on new plant and the cash tied up as sales grow. The first forecast year of our reference path shows the mechanics. MP&E sales of $78.0 billion at a margin of 19.26% give operating profit of $15.0 billion; after tax at 23% that is $11.6 billion. Depreciation adds back $1.79 billion, capital expenditure takes away $3.59 billion and extra working capital takes away $607.4 million, leaving free cash flow of $9.16 billion. The company's own measure of MP&E free cash flow was $9.48 billion in FY2025, and we measure the same thing over the last twelve months at $13.2 billion.

That last number is a good example of reading a figure for what it contains. It is higher than our forecast years partly because customers paid advances on large orders, which is cash that arrives before the work is done, and partly because the cash taxes the company paid ran below the tax charged in its accounts. A model that carried that boost forward forever would overstate the business, and one that ignored it would understate the present, so we show both and say which one we use.

What a discount rate does

A dollar in nine years is worth less than a dollar today, and the discount rate says how much less. We use 10.27% for the whole company's cash flows, built from a cost of equity of 10.41% and a small weight on debt of 2.19%. One input underneath it is observed: the risk-free yield of 4.91%, which is the 10-year Treasury par yield on the price date. The others are assumptions we made, namely an equity risk premium of 5%, a beta of 1.1 and a cost of debt set 0.5% above the Treasury yield, and that is a limit you should keep in view because the rate and the margins move value more than any other input, and which of the two moves it more depends on how far each is changed. Half a point less on the premium gives the reference $374.87 a share and half a point more gives $322.75. In the reference, a discount rate one point lower gives $399.19 a share and one point higher gives $307.24, against $346.89 at our rate. The cash flow for each forecast year is multiplied by a discount factor, which for the first year in our reference is 0.960732 and gets smaller every year.

Why the terminal value is more than half

We forecast nine years and then need to say something about everything after. The standard shortcut is to take the first year after the forecast, grow it forever at a modest rate (3% for us) and divide by the discount rate minus that growth rate. In the reference that terminal year's cash flow is $14.1 billion, which gives a terminal value of $194.3 billion that we then discount back to the valuation date. It ends up being 55.33% of the business's value, which is typical and also uncomfortable: more than half of what we say Caterpillar is worth depends on a margin and a growth rate in a year we cannot see. The discipline is to check that the terminal year is not a flattering one. Ours spends $3.66 billion on capital against depreciation of $3.05 billion, because a business that grows has to keep adding plant.

Scenarios, and why they are not averaged

We show a downside, a reference and an upside: $213.44, $346.89 and $466.04 a share. They are paths with stories behind them and not a probability distribution. We attach no probabilities, so the middle value is not an expected value, and nothing in them says the share price will stay between the ends. It is reasonable to ask which of them is most likely. We rate against the reference because it is our best reading of current economics, and we describe it against three yardsticks, which is more useful than calling it the base case. Its ninth-year MP&E margin of 18.14% is conservative compared with the reported 18.85% of the last twelve months, level with that margin once the tariff recoveries are taken out, and generous compared with the long record, while its first four years, at 19.26%, 19.3%, 19.16% and 18.99%, are above both twelve-month readings.

Reading a sales bridge: volume, price, currency and other

A company's sales change for four reasons, and Caterpillar's filings break them apart. Volume is how much more or less was sold, price realization is what was charged for it, currency is the effect of exchange rates, and other covers the remainder. The bridge for the third quarter of 2026 in Machinery, Power & Energy starts from $16.7 billion a year earlier, adds volume of $2.21 billion, price of $663 million and currency of $58 million, and arrives at $19.7 billion. Dividing each item by the starting sales gives its size: volume 13.20%, price 3.96%. Now compare FY2025, when the same bridge shows volume of $3.39 billion and a price effect that took away $817 million. The lesson is to ask of any growth figure how much was price, because price is the part that can be reversed and the part that tells you about pricing power. What the bridge cannot tell you is mix, which the company does not break out.

Incremental margin: what an extra dollar earns

An incremental margin is the change in profit divided by the change in sales between two periods. In the third quarter, Machinery, Power & Energy sales rose $2.93 billion and MP&E profit, which carries corporate items and the small All Other segment as well as the three primary segments, rose $1.14 billion, so each extra dollar of sales brought 39.00% of profit; the three segments alone brought 35.59%. Both are far above the 17.95% average margin of a year earlier. That is what operating leverage and rising prices look like together. The same measure turned negative for the full year FY2025, when sales rose and profit fell. Incremental margins swing widely, which is why we use them to understand a year and not to forecast the long run.

Why a long history matters

A single good year tells you little about a cyclical business. Caterpillar's consolidated operating margin was 1.78% in FY2009, 10.91% in FY2020 and 20.17% in FY2024, and the mean from FY2018 to FY2025 is 15.53%. A mid-cycle margin is an estimate of what the business earns on average through ups and downs, and it is a range: we show the reference's own path, and beside it the path with every margin lowered to the long-record mean, which is worth $294.56 a share. The limits are as important as the number. The series includes the lending arm and every restructuring charge, so it is not a machinery margin, and we could not measure segment margins before FY2023.

Invested capital and what a multiple of it implies

A different way to check a value is to ask how much capital the business ties up and what it earns on it. At 09-30-2026, Machinery, Power & Energy held $23.1 billion of net property, plant and equipment and working capital, and its last-twelve-month profit after tax earned 46.37% on that. Our reference values the business at 6.67 times the capital and the price at 11.95 times. A steady-state identity connects the two: the multiple of capital equals the return minus growth, divided by the discount rate minus growth. Turn it around and it gives the return at which a multiple fits, 51.51% for the reference and 89.85% for the price. Neither is a forecast. Both show what a value assumes about returns, which is a different question from what it assumes about cash.

Comparing companies carefully

The Deep Dive sets Caterpillar beside Deere, Cummins, GE Vernova and Terex, and the comparison is useful only if the measures match. We compare consolidated operating margin and research spending, which the five report in the same consolidated form, with any missing year marked as not available, and not segment margins, which differ in definition. Even then a figure can mislead. GE Vernova's profit in fiscal 2025 was $4.88 billion on an operating profit of $1.39 billion, helped by an income tax benefit of $2.05 billion. A reader who compared profit margins would have concluded the business was far more profitable than its operations were. When a number looks strong, look for the line that made it so, and when a number is not in the filing, say it is not available instead of leaving it as zero.

How a challenge is tested

An independent review of our model tested each assumption by changing it alone and recording what happened to the value. For terminal growth, our base is 3%; the review tried 2%, which lowered the reference by $24.78, and 3.5%, which raised it by $15.43, on the model as it stood at the review. A model changes after a review, so an effect is always dated to the model it was measured on: rerun on the current model, the same two values move the reference by $19.90 down and $12.15 up. The point of the exercise is to find out which disagreements matter. A disagreement that moves value by a few dollars and does not change the conclusion is recorded and left alone, and one that changes how we describe the model, as the review's challenge to the through-cycle margin did, is answered in the text.

A related lesson is the difference between a bear case and a stress. Our downside has a mild trough: its worst year is 8% of sales below the year before. History is harsher, because consolidated revenue fell 22.4%, 36.9% and 41.5% in three past downturns. So we add stresses that sit below the downside: a deep trough whose margins and growth recover but whose ninth-year sales stay 15.90% below the downside's ($175.00), and a trough with no recovery within the forecast ($105.52). A stress is labeled hypothetical because nobody has measured how likely it is.

Reading a price against a value

The price of $613.62 on 11-02-2026 is a fact; the values are our conditional estimates. The gap between them can be described in two ways that a good analyst keeps apart. The first is a forward-looking description: our reference value is 43.5% below that price, and the upside is 24.1% below it. The second is a reverse calculation. With everything else fixed, the reference would match the price at a discount rate of 7.01%, or with every segment margin 12.22 points higher, and the upside would match it at a discount rate of 8.5%. Across the ranges we tested on the rate, the margins and terminal growth, no single change closes the gap, and these break-evens sit well outside those ranges. They are statements about the model. They do not say what investors believe, and they do not say the market is wrong.

Prices keep moving after a valuation date, and a good habit is to date every comparison. On 11-06-2026 the shares closed at $558.63 and on 11-09-2026 at $598.88, 2.4% below the valuation close, so against the 11-09-2026 settled close the reference is 42.1% below the price instead of 43.5%. The value did not change, because we did not recalculate it. Only the price we compare it with moved, which is why a rating is always tied to a price. A settled close is the price at the end of a session, not a live quote.

The same habit applies to what a statement covers. The most favorable of the joint sets we ran, three settings together, gives the upside case $570.61, which was below the 11-02-2026 close, was 2.14% above the 11-06-2026 close and is 4.72% below the 11-09-2026 settled close, so which side of it the price falls on depends on the date. Applying every favorable setting we tested at once gives the upside $676.06, above all three closes, and the reference $517.55, below all three. Each is a sensitivity and not one of the three paths. A sentence such as "no combination reaches the price" is true only of the combinations it names, and a sentence such as "the values sit below the price" is true only of the cases it names and the close it uses, and a careful reader asks which cases and which close.

A second method is a useful check on the first. Taking the last twelve months of MP&E operating profit, assuming it neither grows nor shrinks, and capitalizing it at the same rate gives $238.73 a share; using the FY2018 to FY2025 average margin instead gives $198.96. We never average a method like this with the cash-flow model, because they answer different questions: one asks what the cash flows are worth, the other what today's earnings are worth if nothing improves.

How the rating works

The same company can move up or down this scale as its price moves, without any change in our view of the business. It is not a trading signal and it is not a price target. It compares the price with intrinsic value at the valuation date, which comes from nine forecast years and a terminal value, so it says nothing about how the shares will perform over the next twelve months or how long a gap would take to close. We judge it against several things at once without turning them into a formula: the gap between price and value at that date, the discount rate at which the model would match the price, the dividend ($6.52 a year at the current rate, a yield of 1.06% and covered 4.39 times by MP&E free cash flow), the stresses, the balance sheet and how many things must go right.

What we do not know, and how to use that

We do not know why the shares fell 28.9% between 10-06-2026 and 11-02-2026; no document we gathered explains it, and a good habit is to say so and not to invent a story. We assumed the equity premium, the beta and the debt spread behind the discount rate. We did not have the third-quarter call, and we read the quarterly filing only for dated facts, without rerunning the model on it. We keep the long margin record apart from the forecast because it includes the lender and every one-off charge. The right way to use all of this is as a set of questions to ask of the next quarterly report: do power margins stay high, do machinery margins stay up when dealers destock, and does the backlog turn into sales on the schedule the company implies. If the answers are better than ours, the values rise, and the rating should be reviewed.

Assumptions and disclosure

This edition is for information and education. It is not individualized advice. Scenarios are conditional, unweighted and not forecasts.

Important disclosures

Blockhaven research is provided for general informational and educational purposes. It is not individualized investment advice and does not take into account any reader's financial circumstances, investment objectives, risk tolerance, tax situation or portfolio.

Nothing in this report is an offer or a solicitation to buy or sell a security or other financial instrument. Reading or accessing Blockhaven research does not by itself create an investment-advisory, fiduciary, brokerage or client relationship.

Investing involves risk, including the possible loss of principal. Scenario values, terminal values, valuation ranges, sensitivities and forward-looking assumptions are analytical tools, not guarantees or promises of future performance, and actual results may differ materially.

Blockhaven uses information from company filings, company publications, market-data providers and other sources it believes to be reliable, but that information may be incomplete, delayed, revised or incorrect.

Opinions, ratings and interpretations reflect Blockhaven's research judgment as of the report date and may change as new evidence becomes available.

Positions and conflicts

As of November 6, 2026, Blockhaven's owner held no shares, options, or other direct derivatives in Caterpillar Inc.; had no material positions in Caterpillar Inc. through entities the owner controls; and had no paid financial relationships with Caterpillar Inc.

© 2026 Blockhaven LLC. All rights reserved.