CAT Deep Dive: A great franchise priced beyond its modeled economics

The judgment

Caterpillar is two businesses under one ticker, and the price leans heavily on the faster-growing one. Construction Industries and Resource Industries supply dealer-delivered machinery into cycles that cut consolidated revenue by 22.4% to 41.5% in the three downturns we measured in our long record; Power & Energy supplies engines and turbines into a power market in which orders and customer deposits are strong, led this quarter by data-center demand. The second business is real and growing. Our difficulty is the price. At the 11-02-2026 close of $613.62 the shares sat 31.7% above the upside case in our model, and our reference value of $346.89, which we treat as our central estimate of current economics and not as a conservative hurdle, is 43.5% below that close. Across the ranges we tested, a point on the discount rate, two points on the margins and half a point on terminal growth, no single change closes that gap. Holding the rest fixed, the model does match the price when the reference is discounted at 7.01% or when every segment margin is 12.22 points higher in every year, and both of those sit well outside the ranges we tested. The most favorable of the joint sets we ran in the model, the rate one point lower, Power & Energy margins higher by 1.5% of sales and working capital at 10% of the change in sales, leaves the reference 30.8% short of the 11-02-2026 close and the upside case 7% short. That set gives the upside $570.61, which is 4.72% below the 11-09-2026 settled close of $598.88 (the 11-06-2026 close of $558.63 was below it). It is a sensitivity and not one of our three paths. Stacking every favorable end of our sensitivity table and of the independent review's challenges at once goes further: the upside reaches $676.06, 10.18% above the 11-02-2026 close, while the reference reaches $517.55, still 15.66% below it; against the 11-09-2026 settled close the upside is 12.89% above it and the reference 13.58% below. That stack needs, beyond anything in the record, a ninth-year MP&E operating margin of 24.09% on the upside path, above every MP&E margin in our record (FY2024, FY2025, the last twelve months and the third quarter), together with a lower rate, faster terminal growth and a collected tariff receivable, so it is a hypothetical stress and not a fourth path.

That is why the rating is 1/5 and not 2/5. The methodology we use reserves its lowest level for a price that requires economics materially better than we expect, and at the 11-02-2026 close the price asks for more than our upside delivers at the rate we assume. The strongest argument for the level above is set out later, and it is a good one: the upside case matches the price at a discount rate of 8.5%, against the 10.27% we use, and the equity premium and beta behind our rate are judgments and not measurements. We take the argument seriously and still rate the price, not the company.

We do not know why the shares fell as far as they did. The 11-02-2026 close of $613.62 was 28.9% below the $863.44 of 10-06-2026 and 42.4% below the 12-month high close of $1,064.90 on 06-30-2026, and the fall from the $713.38 close of 10-28-2026, the session before the third-quarter release, to 11-02-2026 was 14%. No document we gathered, the latest of them on 11-07-2026, explains the October decline, and the third-quarter Form 10-Q reports no material change to the risk factors in the 2025 annual report, so we have not tuned any assumption to it. A falling price is not evidence that the earlier price was wrong, and it is not evidence that the new one is right.

What moved is the price and, slightly, the yield. The price fell 9.0% by 11-06-2026 and recovered 7.2% by 11-09-2026, and we do not know why it did either, since no document we gathered explains the moves. The last Treasury row we gathered is dated 11-06-2026, when the 10-year yield was 5 basis points lower than on the valuation date, at 4.86% against 4.91%; we gathered no later row. It is an input to our discount rate, which we built from the 11-02-2026 row and have not rebuilt. The stored one-point sensitivity gives the scale: a discount rate one point lower gives the reference $399.19 and one point higher gives $307.24, against $346.89, so a move of five hundredths of a point is small beside it.

What did not move is the value. The three values are those of 11-02-2026. We have not rolled them forward, and we have not recalculated any scenario, break-even or sensitivity. At the 11-09-2026 settled close each of our three paths still sits below the price: the upside 22.2% below it, the reference 42.1% below and the downside 64.4% below. Put the other way, the price is 28.5% above the upside, 72.6% above the reference and 180.6% above the downside. The gap to the upside was 24.1% at the valuation close, 16.6% at the 11-06-2026 close and is 22.2% at the 11-09-2026 close, and each change is the effect of a different price and not of a better value.

One qualification belongs beside that. The most favorable of the joint sets we ran (the rate one point lower, Power & Energy margins higher and a lighter working-capital charge) gives the upside case $570.61, which was 7% 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, while the same set leaves the reference 29.1% below it. That set is a sensitivity and not one of our three paths, and the refresh did not change it. It fell short of the price at the 11-02-2026 close, did not fall short at the 11-06-2026 close and falls short again at the 11-09-2026 close, so whether it reaches the price depends on the close it is compared with and on nothing we recalculated. The full stack of favorable ends, set out under how hard we tested the reference, puts the upside at $676.06, above each of the three closes and 12.89% above the 11-09-2026 settled close, and the reference at $517.55, 13.58% below it. We state it plainly because what we say about the gap covers the cases we ran and the close each was compared with, and not every setting a reader might try.

The comparison does not change our reading. The rating is a statement about the 11-02-2026 close and valuation date, a rating is tied to the price it is given at, and at the 11-09-2026 settled close the price is still above each of our three paths, which is the relation the rating rests on, and the gap to each is still wide. The break-even discount rates move with the price, and the table of what the price fits sets the 11-09-2026 figures beside the 11-02-2026 ones. The new company information we read, the third-quarter Form 10-Q and the Fabick transaction, is set out under subsequent events below and moves no forecast.

The rating also depends on a path: the reference needs backlog to convert, the power premium to persist for several years, and machinery margins to stay well above their long-run average, and it still falls short. The long-run average there is the consolidated record's, FY2007 to FY2025; the segment record behind the machinery margins runs only from FY2023 to the last twelve months. The balance sheet is not the issue, since MP&E has $7.10 billion of cash against modest net debt, so what we are rating is the price paid for the earnings and not the safety of the business.

The strongest competing reading. A sophisticated shareholder would say that the shares had already fallen 42.4% from their high by the 11-02-2026 close, that the power backlog is an unusually large order book with a longer life than our reference allows, and that a lower equity premium or beta than the ones we assume matters more than any margin debate. The arithmetic supports part of this: at a discount rate of 8.5% the upside case equals the price, and the premium and beta behind our rate are judgments, while the Treasury yield under it is observed. If power margins stay above their current level for most of a decade and the discount rate is nearer the low end of what we think plausible, the price is not irrational. We maintain that the price is asking for more than our upside, and that is a statement about the price and not a prediction about the stock.

We would revisit the rating on any of these: evidence that Power & Energy margins are staying at their third-quarter level through a period of higher capacity; machinery margins staying near the level of FY2023 and FY2024 through a destocking quarter; evidence for a lower equity premium or beta, which would lower our rate materially; an explanation of the October decline in a primary document; or a materially lower price, since the rating is tied to the price and would change even if our economics did not.

What we measured, what we assumed and what we do not know

The figures for history come from Caterpillar's annual report, its 2026 quarterly filings and its third-quarter earnings release; where we compute a figure, such as a last-twelve-month total or a margin, it is arithmetic on those reported numbers. The price the rating is set at is the settled regular-session closing price of 11-02-2026, split-adjusted and not adjusted for dividends, and the 11-06-2026 close comes from a later series of regular-session closing prices of the same kind. The latest settled close, 11-09-2026, is the November 9, 2026 regular-session closing price, one market snapshot taken after that session settled; it is a closing price and not a live quote. The risk-free rate is the 10-year yield on the official daily Treasury par yield curve for 11-02-2026, as published, and the 11-06-2026 yield is from the same table. The valuation date and the rating price are 11-02-2026, the latest price observation is the settled close of 11-09-2026, and the evidence runs through 11-09-2026. What is new since 11-07-2026 is that market snapshot only: the filings and other documents we gathered, the model, the valuation and the disclosures are as of their earlier dates. The forecasts, the equity risk premium, the beta, the spread of our cost of debt over the Treasury yield, the margin paths, the stresses and the mechanical one-year roll-forward of value are Blockhaven assumptions and judgments and are labeled as such. We did not obtain the third-quarter earnings call, segment margins before FY2023, how the tariff recoveries are allocated between segments, or any outside estimate, and we found no document that reports machine or engine unit volumes, the amount of services revenue, market share, or the results of the non-US makers. The eight-year history and the peer comparison are read from the SEC company-facts series for Caterpillar, Deere, Cummins, GE Vernova and Terex, retrieved on 11-07-2026, consolidated and as first reported; a peer's fiscal year is named by the calendar year in which it ends, and a figure a company did not report for a year is shown as not available and never as zero. The price-realization bridges, the incremental margins and the allowance figures are read from Caterpillar's own tables and sentences, and the review's assumption table is read from the independent review's record. We read the third-quarter Form 10-Q and the company's analyst slides for the dated facts set out above, and we did not rerun the model on either. The filing says where the customer advances sit: current Customer advances and Other liabilities, as the subsequent events describe. We do not know why the shares fell in October.