Should You Buy This Sneaky AI Dividend Stock Up 330% In The Last 5 Years?

CAT is being valued less like an industrial cycle name and more like a scarce-capacity proxy for AI buildout. That usually works until the market realizes the real bottleneck is not excavation equipment, but electrical interconnect, transformers, switchgear, and service capacity; those are the businesses with more persistent pricing power and less earnings volatility. The second-order winners are therefore the grid-heavy names such as ETN and PWR, plus engine/power competitors like CMI if data-center backup generation keeps expanding.

The near-term setup is supported by backlog visibility, which should keep estimates sticky for a few quarters and limit outright downside. The medium-term risk is multiple compression, not demand collapse: if AI capex growth slows or shifts from on-site generation toward utility hookups, CAT can still grow EPS while the market de-rates the stock from a tech-like multiple back toward a cyclical industrial multiple. That is the key falsifier: continued upward revisions and sustained order growth would invalidate the fade.

Consensus appears to be missing that CAT’s AI exposure is real but indirect, and the best economics may accrue to the picks-and-shovels around power delivery rather than the earthmoving layer. For holders, the asymmetry is poor here: you are paying for a long runway already. The cleaner expression is to own the more direct grid winners and use CAT as a source of funds or a relative-value short against them.