Morgan Stanley analyst Angel Castillo upgraded Caterpillar to equal weight from underweight and more than doubled his price target to $915 from $430, after the industrial giant's revenue surged more than 20% to $17.4 billion on the back of data center power generation demand.
The most telling detail in Caterpillar's recent earnings story is not the revenue figure. It is who changed their mind. Angel Castillo was Morgan Stanley's biggest bear on the stock, and he just tore up his thesis. When the analyst who was most convinced a company was overvalued does a public about-face of that magnitude, it is worth pausing to understand what he saw that he did not see before. The answer, according to Bloomberg's reporting on May 1, is that Caterpillar has quietly become one of the most direct beneficiaries of the AI infrastructure buildout, and the market had not fully priced that in.
Caterpillar makes engines, turbines, and industrial power generation equipment. It has made them for construction sites, mining operations, and oil fields for over a century. That business does not change. What has changed is who else needs large-scale reliable power generation, and how urgently. Data centers running GPU clusters for AI workloads consume electricity at a scale that would have seemed implausible to most infrastructure planners five years ago. A single large AI training facility can require hundreds of megawatts of continuous power. When the grid cannot guarantee that supply with sufficient reliability, operators turn to backup and supplemental generation. Caterpillar builds exactly that equipment, and the order books are reflecting it.
Caterpillar reported revenue of approximately $17.4 billion in its latest quarter, a rise of more than 20 percent. That kind of growth rate is not typical for a century-old industrial manufacturer operating in mature markets. Construction cycles move slowly. Mining capex follows commodity prices. Neither explains a 20 percent jump with the consistency and forward visibility that would prompt an analyst to more than double a price target rather than simply nudge it upward.
Data center related power demand is the variable that breaks the old model. It is not cyclical in the way construction spending is. Hyperscalers and colocation providers are signing multi-year capacity commitments and building to those commitments regardless of broader economic conditions, because the competitive cost of falling behind on AI infrastructure is higher than the cost of the capital expenditure. That creates a more durable demand signal for Caterpillar's power generation business than anything in its traditional end markets, and it is durable in a way that shows up in forward order data, not just in a single quarter's revenue.
Castillo's previous underweight rating was built on a reasonable reading of Caterpillar's traditional exposure: construction equipment demand tied to housing and infrastructure cycles, mining equipment tied to commodity prices, and a valuation that had stretched during the post-pandemic capital spending boom. None of that logic was wrong on its own terms. What it missed was the emergence of a new demand driver large enough to change the growth trajectory of the entire energy and transportation segment.
The AI trade is moving into unexpected corners of the industrial economy
Nvidia remains the most discussed AI infrastructure beneficiary, and rightly so. The chip layer is where the most direct AI spending flows. But the buildout required to run Nvidia's hardware at scale involves far more than semiconductors. It involves cooling systems, power distribution equipment, backup generation, fiber networks, and the physical construction of the facilities themselves. Each of those categories has a set of incumbent industrial suppliers who are now seeing demand patterns they have never encountered before.
Caterpillar is the clearest example to surface in public markets so far, but it is unlikely to be the only one. Eaton, which makes electrical distribution and power management equipment, has seen similar dynamics. Vertiv, which specializes in data center thermal management, has been rerating for months as its order backlog expanded. What Caterpillar's earnings and Castillo's upgrade confirm is that this is not a theme limited to a handful of purpose-built data center infrastructure companies. It is spreading through the broader industrial supply chain in ways that traditional sector analysis was not built to capture.
For investors, the practical implication is that screening for AI exposure purely through software and semiconductor lenses misses a growing portion of where the capital is actually flowing. The physical infrastructure required to keep AI systems running, power, cooling, backup generation, and grid interconnection, represents hundreds of billions of dollars in spending over the next decade, and that spending lands on balance sheets that most technology-focused analysts do not cover closely.
For founders and operators building in the AI infrastructure space, Caterpillar's quarter is a useful reminder that the supply chain constraints most likely to create operational headaches are not always the obvious ones. GPU availability gets the headlines. Power availability at the site level is increasingly the binding constraint that determines whether a data center project moves on schedule or sits waiting for equipment that Caterpillar and its competitors are now struggling to deliver fast enough. That bottleneck is not going away soon, and the companies positioned to relieve it are worth watching closely regardless of which sector bucket they happen to fall into.
Also read: The Pentagon Just Put Frontier AI on Its Most Classified Networks • Nebius acquires Eigen AI for $643 million as the inference bottleneck becomes the new GPU war • Retail traders are handing their portfolios to AI agents and the early results are as mixed as you would expect
