2026 Top 10 AI Power & Utilities: The Real AI Trade Nobody's Talking About

Investment Thesis: LONG (Sleeve Allocation) | Time Horizon: 18-36 months Target: Diversified AI infrastructure exposure through utilities and industrial "picks-and-shovels" Part 1: The Grid Is the New GPU This is Part 2 of our 3-part AI investment series. In Part 1: 2026 Top 10 AI Stock Picks: The Vulcan Systematic Framework , we covered the core AI plays driving the compute revolution. Now we turn to the infrastructure that makes it all possible. Part 3 will explore Power, Pipes, and Cooling: the physical backbone picks. Last month, a transformer manufacturer in Ohio got a call they couldn't fulfill: 150-week lead time. The customer was a hyperscaler trying to connect a 500 MW data center campus. That's three years just to get the equipment to plug into the grid, before a single server rack goes live. Here's the trade most people are missing: while everyone's chasing semiconductor stocks and AI software plays, the real bottleneck isn't chips. It's megawatts. It's copper. It's steel in the ground. I've spent the last six months building what I'm calling the "AI Power Spine" portfolio. Not because I think utilities are sexy, but because I've learned the hard way that the most obvious trades aren't where you make money. When I started digging into data center power demand projections, the numbers were staggering. S&P Global projects U.S. data center grid load jumping from 62 GW in 2025 to 108 GW by 2028. That's not incremental growth; that's a near-doubling in three years. The infrastructure to deliver that power? It doesn't exist yet. And that's the opportunity. Why These 10 Names From 900+ Candidates I started with roughly 900 utilities and power-industry stocks in the Vulcan screener. Applied thematic filters for AI relevance, fundamental cutoffs for quality and safety, and ran everything through the mk5 scoring model. The result: 10 names that passed every gate, from regulated utilities with data center exposure to the industrial companies building the physical backbone. The screening criteria looked like this: Quality score above 5 (stable cash flows, reasonable capital allocation), Growth inflection tied to electrification or data center load, Safety metrics indicating manageable leverage and coverage ratios, Valuation that doesn't require heroic assumptions, and most importantly, a direct line to AI-driven power demand. What you won't find here: pure speculation, single-catalyst bets, or names trading at nose-bleed multiples with no margin of safety. I've paid too much tuition on those trades to repeat the mistake. The Contrarian Insight Here's what made me lean into this thesis: everyone knows AI needs power. But they're playing it through the obvious names, the ones already priced for perfection. Meanwhile, regulated utilities with 3% dividend yields and 17x P/E multiples are quietly signing multi-decade power purchase agreements with hyperscalers. When Duke Energy extended coal plant operations specifically to serve "new data center and AI infrastructure," that told me something. When WEC raised its 5-year capex plan by $8.5 billion for a single AI data center project, that confirmed it. The money is already moving, but the stocks haven't fully reflected it. My positions: I own NEE as the largest utility weight, complemented by positions across the regulated group. I'm building into ETN and watching VRT for better entry points. Full disclosure matters here because I'm asking you to consider the same thesis. How I Could Be Wrong Let me be clear about the risks upfront, because this trade isn't guaranteed. If AI buildout slows materially, say hyperscalers pull back capex by 30% or more, these utilities won't see the load growth they're planning for. If interest rates stay elevated for another two years, utility multiples compress further. If regulatory pushback prevents cost recovery on data center infrastructure investments, margins get squeezed. I've been wrong on utilities before.