Every week, we see another headline about which AI model is winning.
Gemini vs. ChatGPT. Anthropic vs. Grok. The model war dominates the conversation, the coverage, and most investors’ portfolios.
For two years now, we’ve been focused on a different question: who keeps the lights on?
Because it doesn’t matter which model wins; they all need power to run. And right now, the U.S. grid is nowhere near ready to deliver it.
U.S. data center power demand is expected to climb from 41 gigawatts in 2026 to 66 gigawatts by 2027, according to Goldman Sachs Research. That’s a near-doubling in roughly two years. The grid can’t keep up.
That gap is where the real investment opportunity lives.
1. The transmission play
Everybody’s fighting over who generates the power. Almost nobody’s asking how it gets from the source to the data center.
You can build all the plants you want. But without transmission lines, the data center stays dark.
That’s where Quanta Services (PWR) comes in. The company builds and maintains the electric transmission and distribution infrastructure that connects power generation to end users. No other company in the country does this at Quanta’s scale.
Quanta raised its 2026 revenue guidance, implying roughly 40% growth, and lifted its diluted earnings-per-share (EPS) guidance to a midpoint around $11.66. The backlog is enormous, and the pipeline keeps growing as utilities scramble to keep pace with data center load requests.
If you want a second way to play this same trade, MasTec (MTZ) does similar transmission and power delivery work (albeit at a smaller scale). MasTec reported Q2 2026 revenue of $4.37 billion, up 23% year over year, with its Power Delivery segment delivering over $1.25 billion in quarterly revenue alone.
Both companies win as long as AI demand keeps growing.
2. The cooling play
Getting power to the data center is step one. Keeping $40 million worth of graphics processing chips (GPUs) from melting is step two.
AI chips run hot, and the denser the workload, the more heat they generate. Cooling is what keeps the whole system from shutting down.
That’s Vertiv’s (VRT) job. The company makes the cooling systems, power distribution units, and equipment racks that go inside data centers.
Last month in the Curzio Alpha portfolio, we locked in a gain of ~200% on Vertiv. But the growth story is far from over. The company reported Q1 2026 revenue of $2.65 billion, up 30% year over year, with Americas growth up 53%.
And as data center density increases—and it will, because each new generation of AI chips draws more power per rack—Vertiv’s products will become even more critical.
3. The generation play
Bloom Energy (BE) makes fuel cells—power generation units that run on natural gas or hydrogen and produce electricity on-site, without connecting to the broader grid. That’s important because grid interconnection can take years. A data center that needs power now can’t wait.
Bloom Energy guided for 2026 revenue of $3.1–$3.3 billion, representing roughly 58% growth at the midpoint.
We got into Bloom early and ultimately locked in gains of ~1,000% in under two years… while most investors were still arguing about chatbots.
The stock has pulled back from its June highs—shares closed around $204 in late August, down roughly 40% from a June peak near $351—but the underlying demand story hasn’t changed. AI data centers need power that can come online fast. Bloom’s fuel cells can make that happen. The grid, in most cases, cannot.
The bigger picture
Every dollar that Big Tech pours into AI training and inference has to flow through physical infrastructure first: transmission lines, cooling systems, on-site power generation. That’s not changing. If anything, it accelerates as the models get bigger and the data centers get denser.
The average investor’s instinct is to buy the “winning AI.” The smarter trade—one we’ve been making for over two years—is to buy the companies that every model depends on, regardless of who comes out on top.
Power isn’t a supporting character in the AI story. It’s the foundation everything else is built on.
For more analysis on where the real AI infrastructure opportunities are—and which names are best positioned right now—tune in to Wall Street Unplugged each week.
