The Caterpillar Correction Shouldn't Last Much Longer

Caterpillar's AI-linked power and construction demand has lifted its backlog to $72 billion, nearly double year over year and up $9.4 billion, or 15%, sequentially. Its power & energy segment grew 17% year over year in Q2, while construction revenue rose 35%, supported by North American AI data-center buildouts. Following a recent correction, the article argues CAT's 1.4x PEG valuation is more attractive versus its prior level above 2x, with hyperscaler capex projected to reach $1.3 trillion in 2027.

Analysis

CAT’s AI optionality is real but should be valued as a distributed-power and site-enablement cycle, not as a direct compute proxy. The highest-margin opportunity is likely reciprocating generation equipment and aftermarket service, where grid-interconnection delays make on-site power economically rational; this favors CAT versus construction-only peers. However, a meaningful portion of order visibility can be exposed to mining, dealer inventory, and traditional non-residential construction, so investors should not capitalize the entire backlog at data-center-like growth multiples.

Over the next 1-3 months, the key catalyst is evidence that Power Generation orders convert into revenue without material supply-chain or warranty-cost dilution. A positive read-through would also support CMI and WCC, while generator-heavy GNRC is more exposed to residential/standby demand and is a less clean AI-power beneficiary. The more important 6-18 month risk is that utilities, gas-turbine suppliers, or hyperscalers solve the power bottleneck through grid upgrades and centralized generation, reducing the premium for temporary/distributed power; an AI capex pause would then expose CAT’s cyclical construction and mining earnings simultaneously.

Consensus may be underestimating CAT’s pricing and service attach opportunity, but is likely overestimating the duration of the construction uplift. The stock should outperform if Power & Energy mix rises and dealer inventories remain controlled; it should de-rate if management’s incremental margin or 12-month order commentary suggests backlog is being filled by lower-quality, lower-margin equipment. The promotional framing and unverified valuation claims in the source are not sufficient evidence of a near-term earnings inflection.

Key Decisions for Investors

• Initiate a 6-12 month long CAT / short DE pair at equal dollar beta after the next earnings print, conditional on Power & Energy orders and segment margin exceeding management’s prior run-rate. CAT has cleaner exposure to behind-the-meter data-center power, while DE remains more tied to agricultural replacement demand; target 10-15% pair return, with exit if CAT power order growth decelerates for two consecutive quarters.
• Use a staged CAT long rather than chase a headline-driven rebound: buy one-third now and add only if the next quarterly backlog conversion and dealer-inventory commentary validate demand quality. Risk-manage with a 10-12% downside stop or on a material cut to full-year segment-margin guidance; expected upside requires multiple expansion plus earnings revisions, not merely AI enthusiasm.
• Monitor CMI as the higher-beta confirmation trade. Go long CMI only if data-center-related power bookings and North American engine margins accelerate concurrently; otherwise CAT is the preferable quality exposure because its dealer/service ecosystem better monetizes installed equipment over multiple years.
• Avoid treating NVDA or META as direct hedges for CAT. A hyperscaler capex slowdown would pressure all three, but CAT’s more relevant falsifier is a decline in power-equipment lead times, canceled construction starts, or a rising dealer inventory-to-sales ratio rather than GPU demand alone.