Energy Stocks and AI Data Center Demand: Utilities and Power Companies That Benefit

---

Section 1 — The Power Demand Equation

The AI infrastructure buildout has created a structural demand shock for electricity that the utility sector has not experienced since the electrification of U.S. manufacturing in the 1940s. A modern AI training cluster running 10,000 Nvidia B200 GPUs consumes approximately 40 megawatts of power continuously — enough electricity to power 30,000 average American homes. Scale this to the dozen or more hyperscale AI training campuses currently under construction across the U.S., and the demand implications become staggering.

The Department of Energy's latest load growth forecast, published in January 2026, projects U.S. data center electricity consumption will grow from 200 TWh in 2024 to 290-320 TWh by 2027, representing an increase of 45-60% in just three years. For context, U.S. total electricity generation is approximately 4,300 TWh annually, meaning data centers will consume 7-8% of national electricity by 2027 — up from 4.7% in 2024.

The geographic concentration of this demand matters for investors. Data center construction is heavily concentrated in Northern Virginia (the world's largest data center market), Central Texas, Georgia, and the Pacific Northwest. This creates localized grid stress: PJM Interconnection (which covers the Mid-Atlantic and Midwest) has already issued warnings about potential power shortfalls in the 2028-2032 timeframe if AI-related load growth continues at current rates.

For utilities serving high-density data center markets, this demand is a revenue windfall. A large data center with 100MW+ of power demand — contracted for 15-20 years under a Power Purchase Agreement (PPA) — represents the most attractive utility customer in history: stable, high-volume, creditworthy, and growing.

---

Section 2 — Nuclear Power: The Preferred AI Energy Source

Hyperscalers have rapidly discovered that renewable energy, while cheap per MWh, cannot meet the 24/7 baseload requirements of AI clusters. Solar panels produce nothing at night; wind turbines are intermittent. AI training runs — which may execute continuously for weeks — require uninterrupted power that only two sources can reliably provide at scale: natural gas and nuclear.

Nuclear power has a decisive advantage for AI-focused hyperscalers: carbon-free generation. Microsoft, Google, and Amazon have all committed to 24/7 carbon-free energy matching for their operations. Nuclear is the only firm, dispatchable, carbon-free power source available at gigawatt scale. This has driven a wave of nuclear Power Purchase Agreements (PPAs) that would have seemed inconceivable five years ago.

The landmark deal was Microsoft's September 2024 agreement with Constellation Energy to restart the Three Mile Island Unit 1 reactor in Pennsylvania (renamed Crane Clean Energy Center). The 20-year PPA covering 835 MW of nuclear power is priced at approximately $100/MWh — well above current market rates of $65-70/MWh — reflecting the scarcity premium for firm, carbon-free power. Constellation Energy's stock has appreciated 180% since the deal was announced.

Vistra Energy (VST) signed a similar deal with an unnamed hyperscaler (widely reported to be Google) for output from its Comanche Peak nuclear plant in Texas. Vistra's nuclear portfolio of 6.4 GW across four plants has become a core asset in a way that was not anticipated when the company emerged from bankruptcy in 2016.

---

Section 3 — Natural Gas and the Peaker Plant Opportunity

Natural gas generation benefits from AI data center demand in a more nuanced way than nuclear. While hyperscalers prefer carbon-free power for their long-term commitments, the grid needs flexible gas generation to balance the intermittency of renewables and manage peak demand periods. In markets with high renewable penetration (ERCOT/Texas, California), gas peaker plants are increasingly valuable despite low capacity factors.

Vistra's natural gas fleet — 26 GW across ERCOT and other markets — provides exactly this flexibility. The combination of nuclear baseload and gas peaking within a single company makes Vistra uniquely positioned as a "full stack" power provider for the AI economy.

Kinder Morgan (KMI) and Williams Companies (WMB) are natural gas pipeline operators that benefit indirectly from increased gas-fired generation. Pipeline throughput grows as gas demand from power generators increases. Williams Companies, with its Transco pipeline system serving the high-density Northeastern data center markets, is particularly well-positioned. At 21x earnings for 7% EPS growth, KMI offers a more modest but stable return profile with a 5.8% dividend yield.

The small modular reactor (SMR) sector represents the long-horizon optionality in the energy/AI intersection. NuScale Power (SMR) and X-energy (private, backed by Amazon) are developing reactors specifically sized for data center campuses. These remain 5-8 years from commercial deployment at best, but Amazon's $500 million investment in X-energy signals hyperscaler willingness to fund novel power solutions.

---

Section 4 — Investment Framework

The energy/AI intersection offers a relatively unusual combination in today's market: reasonable valuations (utility multiples of 12-20x versus tech at 25-40x) with a genuine, quantifiable demand catalyst. The challenge is that utility earnings are regulated and do not directly capture power price upside — unregulated generators like Constellation Energy and Vistra are better positioned to monetize premium AI PPAs.

The framework for this sector divides into three tiers. Tier 1 (highest conviction) includes Constellation Energy — the largest nuclear operator in the U.S. with 21.5 GW capacity, premium AI PPAs already signed, and a reasonable 18x forward multiple given the structural demand shift. Vistra is our second-highest conviction name with similar nuclear optionality and a more attractively priced Texas gas portfolio.

Tier 2 includes Dominion Energy and Pacific Gas & Electric — regulated utilities with nuclear assets where AI demand may drive above-average rate base growth but where regulatory approval processes slow monetization. These are appropriate for income-oriented investors seeking 3-4% dividend yields with modest growth.

Tier 3 is pure optionality: NuScale, X-energy (when public), and uranium miners (Cameco, Uranium Energy Corp) that benefit if the nuclear renaissance drives sustained uranium price appreciation. Cameco at 40x earnings is pricing in significant upside and carries execution risk on its uranium production ramp.

---

Verdict

---

Data as of March 2026. Not financial advice.

— iBuidl Research Team