Caterpillar Doubled on a $72 Billion Backlog. The Duel Is Over What Happens to the Margin Next.

Both camps watched the same annual report card. Caterpillar's second-quarter sales reached $20.5 billion, up 24% and the first quarter over $20 billion in company history; adjusted earnings were $8.17 a share against a consensus around $6.20; the order backlog swelled to a record $72 billion; and adjusted operating margin hit 21.9%, the best in years. The stock has nearly doubled over the past twelve months on the strength of that story, and it trades near 33 times trailing earnings with a dividend yield under 1%.

Bulls read the print as proof that Caterpillar has stopped being a cyclical and become an AI-infrastructure platform: the backlog gives earnings visibility, and the margin is the new durable base. Bears read the identical numbers as the classic peak of a cycle — an overflowing order book and a record margin at a record multiple, right before both revert. The whole argument collapses into one disagreement variable: whether that 21.9% margin is the foundation of a growth story or the outlier print of a machine that always mean-reverts.

Shared facts, as of September 14, 2026. Caterpillar trades at $783.63, down roughly 4% on the day and about 27% below its 52-week high near $1,073, after a pullback from a year-long run that topped an 80% gain. Market capitalization is $360 billion against roughly $399 billion of enterprise value. The second quarter delivered the record backlog — up about 92% from a year ago and $9 billion sequentially — plus record free cash flow, raised full-year guidance to mid-to-high-teens revenue growth, and a roughly $2.2 billion tariff bill that management has flagged as a margin drag. Consensus already assumes the boom cools. The fight is over the numbers behind that consensus.

Round one: Is the backlog visibility, or just the top of the book?

The bull's strongest case is that the backlog is unusually concrete. It grew 92% year over year to $72 billion, roughly a full year of trailing revenue, with growth across all three operating segments and some power-and-energy orders booked into 2030. Power generation sales rose 72% in the quarter on data-center demand, and management says construction growth is being pulled by investment in critical infrastructure and data centers rather than by speculative dealer restocking. An aggregate AInvest signal from analyst ratings labels the stock a Buy, and management has weaned the enterprise long-term sales outlook up to a 6–9% compound annual growth rate from 5–7%.

The bear concedes the scale but disputes what it means. Only about 59% of that backlog is expected to ship within the next twelve months, so part of the "visibility" is a mirror of Caterpillar's capacity constraints, not pure demand — lead times on some power products now stretch into 2028–2029. Order books always look like this at the top of a cycle, before they start coming undone. And one explicit tell is already visible: management says changes in dealer inventory will be a headwind to construction sales volume in the second half of 2026, with a reduction of more than $1 billion expected in the fourth quarter. That is normalization, not scarcity.

Winner: the bull, narrowly. A 92%-grown backlog with orders pushing into the decade is not the same as the order book at the last cyclical peak, and the data-center driver is genuinely new. But the round ends with a warning the bull must carry: a chunk of the backlog is a function of capacity, the rest can unwind, and management itself has written one headwind into the guided numbers.

Round two: Are 21.9% margins a floor or a ceiling?

Here the bull's claim gets bolder. Adjusted operating margin reached 21.9%, up 430 basis points from a year ago, helped by volume, favorable price realization, and lower tariff drag. The bull argues the level is structural: data centers demand the same catalog of high-margin power engines, services attach to a growing installed base, and Caterpillar's pricing power absorbed nearly all of the tariff shock in construction. Customer credit quality is the best in decades — past-due accounts are at their lowest ratio since 1998 — which the bull reads as confirmation that end demand is real, not credit-fueled.

The bear says the margin is precisely the unreliable number. Trailing-twelve-month operating margin is about 17.5%, below the three-year average of 18.7%, which means the 21.9% quarterly print is the outlier, not the new normal. Two costs are set to grind against it: the roughly $2.2 billion full-year tariff bill, and accelerated depreciation on the new engine capacity Caterpillar is racing to build — a drag management has already warned will weigh on power-and-energy margins for several years. That is the bear's heaviest punch: the very expansion the bull celebrates as a growth enabler comes with accounting charges that push the trend margin down.

Winner: the bear, but only on the margin of the margin. The bull's structural story is credible and the credit quality is a real improvement. Yet the bear is correct that the quarter's 21.9% figure is above the recent trend, that tariffs are booked, and that the capacity build-out does not produce margin for free. The duel has genuinely narrowed to this one variable — which is exactly what the valuation round will expose.

Round three: What the price already demands

Now make both stories pay rent. The striking thing about Caterpillar's valuation is that an unusually cheap-looking comparison isn't hiding anywhere. At about 33 times trailing earnings, roughly 23–24 times EBITDA, and with a payout yield near 0.8%, the stock sits alongside — and in some cases above — peers that were themselves re-rated by the same data-center trade: Deere at ~38 times, Eaton near 40 times, Cummins around 27 times. Michael Burry disclosed a short position on June 30 at roughly $1,061 a share, calling Caterpillar one of the market's most overvalued AI beneficiaries.

The reverse-engineering does the real work. Consensus already models revenue growth slowing to roughly 8% a year through 2027 — far below the 24% just reported — and earnings growing instead through continued margin expansion. In other words, the market is not paying for the backlog's revenue; it is paying for the durable margin. Trefis makes the point explicitly: hold the multiple at a stable ~27 times 2027 earnings and the stock is worth roughly $918, about 7–8% above today's price. The price of today's quote buys back roughly today's value, with the entire upside depending on two years of margin holding up against tariffs, depreciation, and a normalizing construction cycle in a company that has historically fallen 60–70% peak to trough.

Winner: the bear at this price. Both cases are now spending the same budget. The bull's differentiated revenue is real but partially capacity-bound and not always capitalized at full value even by consensus; the margin that would justify the multiple is the least durable number on the page.

Ruling: bull on the business, bear on the stock

Caterpillar is a better business than the price gives credit for being worth worrying about as a cyclical — and it is a worse stock than the business deserves at current levels. The backlog is genuine, the data-center tailwind is new, and the company can clearly fund its own build-out. But at roughly 33 times trailing earnings and near peak-cycle margins, the buyer is not being paid to be patient; they are financing the bull's case where it is weakest. The burden of proof sits with the bull to show the 21.9% margin is the base, not the spike, while revenue growth slows as consensus expects.

The ruling flips on one dated number. In late October, Caterpillar reports the third quarter. Watch whether adjusted operating margin holds near the 20% mark despite the tariff bill and the depreciation drag — and whether power-and-energy sales to users grow enough to offset the construction dealer-inventory unwind. A margin print holding above ~20% with power generation still compounding vindicates the bull's structural case and gives the multiple room to stand. A reversion toward the 17–18% trend — with dealer inventories coming down faster than guided and the data-center order flow slowing — is the bear's confirming indicator and the earliest point the stock's peak-cycle multiple loses its justification. Either result lands within a quarter, which is why this is a wager on the margin, decided on schedule, not a story that can wait.