Forget Hyperscalers, These Stocks Are The New AI Plays

Caterpillar, Eaton, and Vertiv are showing strong AI-infrastructure demand, with Caterpillar Q1 revenue up 22.22% and Power Generation revenue up 41% YoY, Eaton’s Electrical Americas revenue up 21% YoY with a 24.9% margin, and Vertiv revenue up 30.13% YoY with backlog up 109% YoY. All three are raising or reinforcing upbeat guidance, while Vertiv lifted 2026 net sales to $13.50B-$14.00B and adjusted EPS to $6.30-$6.40. The article argues the AI capex beneficiaries remain less crowded than chip stocks, supported by heavy institutional demand and strong retail skepticism.

Analysis

The market is underpricing the durability of the AI capex chain outside semis. CAT, ETN, and VRT sit in the part of the stack where demand is less hype-sensitive and more tied to power availability, grid constraints, and multi-quarter project backlogs, which makes earnings less vulnerable to the usual “beats now, resets later” pattern seen in chip suppliers. The key second-order effect is that AI buildout is creating a bottleneck in electrical infrastructure, so pricing power is migrating downstream to the firms that can deliver megawatts, not just wafers.

Among the three, VRT is the highest-beta expression of the constraint trade because it is closest to the data center load center and is still in a phase where backlog can convert into multiple quarters of estimate revision. ETN is the cleaner quality compounder: its mix, margins, and acquisition of thermal capabilities broaden its attach rate as AI racks become power-dense enough that cooling is no longer optional. CAT is more of a capital cycle beneficiary than a pure AI story, but hyperscale power demand adds a new end market to an already leveraged backlog, and buybacks amplify per-share upside if execution holds.

The contrarian risk is that this trade becomes crowded only after the fundamentals have already inflected, meaning valuation could outrun estimate revision if investors migrate from chips into infrastructure all at once. The main reversal catalysts are a pause in data center permitting/interconnects, a hyperscaler capex digestion quarter, or evidence that power delivery timelines—not demand—become the binding constraint, which would defer revenue recognition rather than eliminate it. Near-term, the setup is still better over months than days because these are backlog-driven businesses with visible forward conversion, not headline-sensitive momentum names.

Key Decisions for Investors

• Go long VRT on pullbacks versus semis for a 3-6 month trade; highest earnings torque and backlog visibility, but size modestly because the stock will be most vulnerable if AI capex expectations pause.
• Add ETN as the lower-volatility quality compounder in the AI infrastructure basket; prefer a 6-12 month holding period where margin expansion and thermal attach rate can re-rate the multiple.
• Use CAT as a secondary beneficiary rather than the core AI expression; best risk/reward if entered on weakness after a broad industrial selloff, with upside supported by backlog and capital returns but less pure AI sensitivity than VRT/ETN.
• Pair long VRT / short a crowded AI chip proxy to express the view that incremental value is shifting from compute scarcity to power delivery scarcity; target 2-3 quarters of relative outperformance if capex remains strong.
• If chasing the trade, use call spreads rather than outright longs in VRT because the market may already be rewarding perfect execution; spreads cap theta risk if investor positioning turns congested.