Three Capex Cycles Reveal What Caterpillar Does Best

Between 2016 and 2019, commodity prices recovered from their post-2015 lows. Mining companies, which had slashed exploration and development budgets, began rebuilding. Orders for Caterpillar's largest machines, the 400-ton haul trucks and hydraulic excavators that move mountains in the Pilbara and Atacama, started climbing.

The temptation to expand was significant. Caterpillar's mining segment had been devastated by the 2012-2015 downturn, with revenue falling by more than 50%. The recovery offered a chance to rebuild capacity and capture market share from Komatsu and Hitachi. Management resisted. Instead of adding manufacturing lines, they increased utilisation at existing facilities and focused on aftermarket services, which carry margins roughly double those of new equipment sales.

Revenue grew from $51 billion in 2021 (the tail end of the cycle overlap) at controlled pace. Operating margins expanded from 13.5% to above 16%. Free cash flow generation accelerated precisely because the company refused to deploy capital into capacity that might sit idle when the cycle turned. That restraint looked conservative at the time. In hindsight, it was the foundation for the stock's re-rating from $150 to $300.

The 2020-2023 cycle was different in character but identical in management response. The US Infrastructure Investment and Jobs Act, combined with similar programmes in Europe and across Asia, unleashed the largest wave of public construction spending in decades. Caterpillar's Construction Industries segment, which accounts for roughly 40% of group revenue, was the primary beneficiary.

Revenue surged from $51 billion in 2021 to $67.1 billion in 2023, a 32% increase over two years. The remarkable part was the margin trajectory. Operating income grew from $6.9 billion to $12.7 billion over the same period, implying operating margins expanded from 13.5% to nearly 19%. That margin expansion, on revenue growth of 32%, tells you that essentially all the incremental revenue dropped through at substantially higher margins than the base business.

How? Pricing power and operating leverage on a fixed cost base. Caterpillar raised equipment prices by 15-20% cumulatively over 2022-2023, citing supply chain costs and materials inflation. Dealers accepted the increases because equipment availability was constrained and project timelines were fixed. When costs subsequently moderated, the pricing stuck. The pattern is identical to what we see at Coca-Cola and Deere: industrial companies with dominant market positions convert cost inflation cycles into permanent margin expansion.

The 2011-2014 cycle, by contrast, saw Caterpillar add roughly $3 billion in manufacturing capacity that became severely underutilised when Chinese infrastructure demand collapsed. Management's refusal to repeat that mistake during the 2020-2023 boom is the single most important factor in the current valuation.

The most recent narrative driver is, frankly, unexpected for a 100-year-old heavy equipment manufacturer. Caterpillar is being reframed as an AI beneficiary. The thesis: hyperscaler data centre construction requires massive earthmoving, site preparation, and power generation equipment. Caterpillar supplies all three.

The data centre power angle is particularly interesting. Caterpillar's Energy & Transportation segment manufactures large-format diesel and natural gas generator sets used for backup and prime power at data centres. As hyperscalers struggle with grid connection timelines (often 3-5 years for utility-scale power), on-site generation using Caterpillar equipment fills the gap. Orders for large generator sets have reportedly accelerated through 2025.

We are cautious about overstating this thesis. Data centre construction, while growing rapidly, represents a single-digit percentage of Caterpillar's total addressable market. The company's fortunes still depend primarily on mining capex, general construction, and energy infrastructure. But the AI narrative has expanded the investor base and contributed to the multiple expansion from 20x to 42x trailing earnings over two years.

The risk, which the consensus is not adequately pricing, is that the AI construction thesis is already fully reflected in the multiple. If data centre construction spending plateaus or grid connections catch up to demand (reducing the need for on-site generation), the narrative premium could compress quickly.