Every Hyperscaler Raised Capex Again. These 3 Boring Industrials Get Paid No Matter Who Wins the AI Race.

AI data-center buildouts are driving a “show me” moment for capex, but industrial power/cooling suppliers are showing results: Caterpillar Q2 revenue rose 24% to $20.5B with EPS $7.77 vs $4.62 a year ago, and its power generation segment revenue surged 72%. Eaton reported Q2 revenue of $8.53B (+$1.51B YoY) and Vertiv Q2 revenue of $3.27B (+24% YoY) with net income $497.9M (+$173.7M YoY). Power demand growth is cited as expanding from ~5MW to 500+MW per data center, supporting continued capex spend by hyperscalers (e.g., Alphabet raising capex $185B→$200B, Amazon $200B→$220B, and Meta issuing debt for AI).

Analysis

This is less a pure AI trade than a bottleneck trade: the value is shifting from compute bragging rights to the unglamorous equipment that turns capex into usable uptime. That makes ETN and VRT the cleaner expressions because their pricing power is tied to grid density, thermal management, and delivery lead times, not to model adoption. CAT participates, but it is the most cyclical of the group and therefore the easiest to disappoint if broader industrial demand softens.

Second-order effects matter more than the headline AI spend. As power density rises, hyperscalers are forced to pre-buy generation, switchgear, and cooling capacity, which improves backlog visibility and supports margin expansion for the suppliers with long qualification cycles. The flip side is that AI capex becomes a direct drag on free cash flow at AMZN/GOOGL/META; if monetization doesn’t keep pace, the market can force a slower spend path, which would hit the industrial beneficiaries with a 1-2 quarter lag.

Contrarian view: this trade is already crowded in quality names, so the next leg likely depends on backlog surprise, not narrative. ETN/VRT can still work if order growth stays ahead of revenue, but CAT looks more vulnerable to multiple compression because part of the move is probably sentiment-driven. Falsifiers over 1-3 months are a capex guide-down, easing lead times, or commentary that AI buildouts are shifting to lower-power architectures; over 6-18 months, liquid cooling efficiency could reduce the hardware intensity per MW and cap the upside.

Key Decisions for Investors

• Go long ETN on a 3-5% pullback into hyperscaler earnings; target 10-15% upside if order growth reaccelerates, with a 6-7% stop if backlog or pricing commentary weakens.
• Use VRT call spreads 3-6 months out as a defined-risk way to play power-density normalization; the trade works best if the market starts marking backlog conversion higher.
• Pair trade: long ETN / short CAT into strength. ETN has the cleaner electrical bottleneck exposure; CAT carries more traditional industrial-cycle risk and is more vulnerable if AI-related demand is not enough to offset softness elsewhere.
• Treat AMZN, GOOGL, and META capex updates as the primary catalyst, not the industrial press release cycle. If any of the three signals a slower spend curve, reduce ETN/VRT exposure immediately.
• If you want a broader expression, own ETN/VRT versus XLI rather than outright long the whole industrial basket; the relative-value trade isolates the AI infrastructure premium and reduces exposure to the rest of the cycle.