Caterpillar is usually pictured on a construction site. Its June-quarter filing points to another customer: a data-centre operator looking for power. The company says growth in large reciprocating engines, turbines and related services is being driven primarily by data-centre applications. The industrial supplier participates in the build-out before the computing customer can switch on its new capacity.
By 13 October 2026, CAT had gained 35.7% on an adjusted-price basis from the end of 2025. SPY had gained 15.0% over the same window. We see an interesting business behind that gap, but the price chart cannot establish why investors bought the shares. The company’s revenue mix provides a better starting point.
Caterpillar has outpaced the broad market
The electricity order inside the AI order
Power Generation external sales reached $3.098bn in the June quarter, against $2.407bn in the same quarter of 2025: an increase of 28.7%, calculated from the reported values. In the March quarter, the comparable figures were $2.817bn and $1.996bn. Those March amounts are derived by subtracting June-quarter sales from the first-half totals, rather than mixing a six-month figure with a quarterly one.
The mechanism is practical. A building full of computing equipment needs a dependable supply of electricity. A delayed grid connection, a need for backup capacity or demand for an alternative prime-power solution can translate into engine, generator and service orders. Caterpillar’s filing specifically describes increasing demand for prime power as data-centre customers look for alternative solutions.
That link is narrower than saying every generator sale is an AI sale. The application also serves other customers, and the filing does not disclose the fraction of Power Generation revenue attributable to AI. We can identify the transmission mechanism without inventing that missing share.
Power generation reached $3.098bn in the June quarter
The diggers still matter
The attractive power story sits inside a broader industrial upswing. Construction Industries sales increased from $6.190bn to $8.346bn, or 34.8%, in the June quarter. Resource Industries rose from $3.886bn to $4.648bn, or 19.6%. Within Power & Energy, Oil and Gas application sales grew 9.5% and Industrial application sales grew 8.8%.
Construction therefore grew faster than Power Generation in this comparison. That matters for attribution: an investor buying CAT owns a diversified industrial company, with dealer inventories, construction demand, commodity investment and equipment finance alongside the power business. Deere’s 40.7% adjusted-price gain over the same year-to-date window is another reason to avoid assigning the whole Caterpillar rally to one theme.
Our comparison puts reporting segments and application-level sales on the same growth-rate axis to show the breadth of demand. The categories are not additive, and the chart is not a revenue-share breakdown.
Construction grew faster than power generation
Trade provides context, not a Caterpillar order book
The selected customs category offers a second window on physical equipment demand. US imports of HS 850213 diesel generating sets were $271m in 2023, $177m in 2024 and $507m in 2025, rounded from the stored customs values. The rebound is substantial, but the observations are annual and describe nominal import spending.
These are not Caterpillar sales, and they do not identify data-centre buyers. They also exclude domestic production and generators outside the selected product code. A change in value can reflect equipment mix or prices as well as the number of machines. We use the series to frame the equipment market, not to claim that a customs total predicts CAT’s next quarter.
The economic backdrop is more restrained than the company growth rates. The latest stored US industrial-production observation is 1.4% year-on-year in August 2026. Company revenue and economy-wide output measure different things, but the contrast reinforces the case for analysing specific applications instead of treating a manufacturer as a simple readout of national growth.
US generator import values rebounded in 2025
A physical bottleneck can create a profitable niche
The principle is that spending can migrate toward the part of a system that is hardest to expand. In a computing build-out, that constraint can sit outside the chip: electricity availability, installation, delivery schedules and reliability all matter. A supplier with relevant equipment and servicing capabilities can participate even when it does not sell computing hardware.
Whether that becomes an attractive investment depends on what the share price already assumes. A rising stock is not evidence of cheap valuation, and a higher sales figure does not by itself establish higher cash generation. On earnings day, we would read the application table alongside segment profitability, backlog commentary, capacity investment and dealer inventory. Broad construction strength also needs to persist.
The economy-wide production backdrop is more modest
Our call
We assign a 65% probability that Caterpillar reports at least 20% year-on-year growth in third-quarter 2026 Power Generation external sales in its 29 October 2026 results. This is a dated operating-business forecast, not a share-price target. We will resolve it using the third-quarter Power Generation application-sales figure against the comparable prior-year figure.
The case rests on the June-quarter growth rate, the first-half increase and the company’s explicit description of data-centre and prime-power demand. The odds are our editorial judgment; they are not a statistically calibrated probability or a conclusion drawn from the generator-import series.
- Growth slows below 20% but stays positive — 25%. The application expands, but installation timing, delivery constraints or a tougher comparison limit the pace.
- Power-generation sales stop growing — 10%. A flat or negative year-on-year result would challenge the near-term momentum case even if longer-term demand remains.
- The test. Read the application-sales table and its comparable prior-year figure on 29 October. Check whether any reporting reclassification changes the comparison before resolving the call.
Methods and limits
The information cut-off is 14 October 2026. Market observations end on 13 October; the filing comparison is the June quarter; the selected trade observations end in 2025; the latest industrial-production reference month is August 2026. No third-quarter result is treated as known.
Stock comparisons use adjusted closes in US dollars and omit transaction costs and taxes. The selected-company comparison is not a reconstructed historical index or an unbiased stock-selection test. Macro and trade data can be revised. All factual figures are traceable to saved database queries or cells in filing HTML returned by a saved query; derived changes and unit conversions are retained separately. The prediction does not revise any earlier Geoanalytic call.
Geoanalytic builds private data systems for advisory firms.
Visit the homepage →