AI Capex Crosses $1 Trillion Next Year. Here Are 2 Industrial Stocks That Will Benefit

AI-driven data center buildout is lifting industrial demand: Vertiv revenue more than doubled from $5.0B (2021) to $10.2B (2025) and adjusted EBITDA rose from $698M to $2.2B, with backlog exceeding $15B at end-2025. Quanta’s year-end backlog grew from $19.3B (2021) to $44B (2025) as it expands grid and data center power systems to meet rising electricity needs. Analysts project 2025-2028 CAGRs of 28% revenue and 38% adjusted EBITDA for Vertiv, and 17% revenue and 19% adjusted EBITDA for Quanta, positioning both as AI-tailwind beneficiaries despite valuation at ~34x and ~29x adjusted EBITDA, respectively.

Analysis

The cleaner trade is not “AI broadly,” but the bottleneck behind AI: power delivery and thermal management. VRT is the more direct torque name, but it is also the more fragile multiple because its upside depends on hyperscalers sustaining unusually high capex intensity and on VRT converting backlog without margin slippage. PWR is the sturdier second-order beneficiary: grid upgrades, substations, and data-center power systems are less tied to any single chip cycle and more likely to persist even if AI hardware spending moderates.

The main losers are the hyperscalers themselves: incremental infrastructure spend is a drag on free cash flow, and if that spend does not translate into faster monetization, markets will eventually punish AI capex for its own sake. That also creates spillover winners in electrical equipment and grid-adjacent names such as ETN, HUBB, and AME, which can capture the same power bottleneck with less single-customer concentration than VRT. Over 1-3 months, the catalyst is order commentary; over 6-18 months, it is whether this becomes a recurring utility-style capex cycle or a one-time buildout.

Contrarian view: consensus is extrapolating “AI power demand” as linear, but efficiency gains, design changes, and interconnection delays can shift spend from hardware urgency to project timing. If backlog grows faster than revenue conversion, that is not a win; it can just mean working capital and execution risk rising before cash flow does. A 100-150 bps gross margin miss or any hyperscaler capex reset would likely compress these multiples quickly.

Key Decisions for Investors

• Long PWR / short VRT for 3-6 months: prefer the broader grid-exposure name with lower valuation and less dependence on one AI hardware architecture; thesis breaks if VRT margin expansion persists while PWR growth stalls.
• Buy PWR on a 3-5% pullback and hold 6-12 months: this is the cleaner structural AI-power beneficiary; risk/reward is strongest if utility capex remains firm and backlog converts without gross margin erosion.
• Do not chase VRT at current levels; only add if the stock de-risks on a 10%+ pullback or if the next print confirms accelerating revenue conversion and stable margins.
• Watch NVDA and hyperscaler capex guides as the falsifier set: any sign of capex moderation or efficiency-driven delay should reduce exposure to VRT first, then PWR.
• Use ETN/HUBB as secondary beneficiaries if you want broader AI-power exposure with less single-name execution risk than VRT.