Nuclear Energy Stocks to Watch as AI Data Centers Drive the Next Power Boom

Everyone has spent two years asking the same question: who are the real winners of the AI economy? The chip designers. The cloud providers. The model companies. All fair. But there's a second-order answer that most investors haven't fully acted on yet. The AI economy, stripped down to its basics, is a power infrastructure story. And the one energy source that delivers reliable, carbon-free electricity around the clock, regardless of weather or time of day, is nuclear. Global data center electricity consumption hit 448 terawatt-hours in 2025, as per Gartner, and is already forecast to exceed 1,000 TWh by end of 2026, that’s equivalent to Japan's entire annual electricity usage, according to the IEA. 

In the US alone, Goldman Sachs Research projects data center power demand will climb from 31 GW in 2025 to 66 GW by 2027: a doubling in just two years. A typical AI-focused hyperscaler, as per IEA, consumes as much electricity annually as 100,000 households. The larger ones being built right now will use 20 times that. Every single one needs power that never goes dark. In the past 18 months, Microsoft, Meta, Amazon, and Google have committed to over 9.8 gigawatts of nuclear power capacity between them.

Let's break down the structural case for nuclear as an investment theme, what each layer of the nuclear supply chain actually offers investors, and how you can access this from INDmoney.

Why AI Data Centers Are Driving Nuclear Power Demand

A standard data center campus today consumes between 200 MW and 500 MW of power. The next generation of AI hyperscale campuses, the ones being built to train tomorrow's models and serve billions of daily inference requests, will consume between 1 GW and 5 GW each. That's the output of an entire nuclear power plant, for a single campus.

Goldman Sachs Research, in a report published in May 2026, puts the US picture in stark terms: data center power demand reached 31 GW in 2025, is expected to hit 41 GW in 2026, and will nearly double again to 66 GW by 2027. Year-over-year capacity additions are scheduled to reach 36.3 GW in 2027 alone, compared to just 6.4 GW actually added in 2024. S&P Global's October 2025 outlook puts US capacity at 75.8 GW this year, crossing 100 GW in 2028. Globally, Gartner projects data center electricity consumption will grow from 448 TWh in 2025 to 980 TWh by 2030, with AI-optimized servers alone jumping from 93 TWh to 432 TWh over the same period.

The range across forecasters is still wide, for the honest reason that nobody knows exactly how fast AI inference workloads scale. What isn't in dispute is the direction or the order of magnitude.

Why Solar and Wind Alone May Not Be Enough for AI Data Centers

The most common response to the data center power problem is to build more solar and wind. Both are cheap. Both produce zero emissions. Tech companies have spent years signing renewable power purchase agreements. So why are they now signing nuclear deals instead?

Intermittency. Solar panels produce electricity for roughly 6 hours of effective output per day. Wind turbines stop when the air is calm. A data center serving live AI requests to users across 15 time zones simultaneously cannot handle that.

The number that explains the whole situation: nuclear plants run above a 92.5% capacity factor, meaning they produce close to their maximum output almost all the time. Solar capacity factors average around 23-25%. Wind averages 34-35% in the US, as per EIA data. So 1 GW of solar capacity delivers roughly 250 MW of usable output on average. One GW of nuclear power delivers 925 MW, consistently, year-round.

Goldman Sachs put it plainly in a January 2025 research note: "Wind and solar could serve roughly 80% of a data center's power demand if paired with storage, but some sort of baseload generation is needed to meet the 24/7 demand." Nuclear is the only source that meets this bar at the scale hyperscalers actually need, without new transmission infrastructure or battery storage at a scale that doesn't yet exist.

Big Tech Nuclear Power Deals: Microsoft, Meta, Amazon and Google

The clearest way to know an investment thesis is real is to watch who is writing the large checks. As of May 2026, 13 announced projects have committed over 9.8 GW of nuclear capacity to AI data center infrastructure, as per the SMR Intel tracker. Every major US hyperscaler has signed at least one deal.

Sources: SMR Intel Tracker, Latitude Media, Introl Blog (May 2026)

Microsoft's Three Mile Island deal is worth sitting with for a moment. This is the same facility that became synonymous with nuclear failure after its 1979 partial meltdown. Microsoft is paying $16 billion for a 20-year power contract and will receive 100% of the plant's output from 2027. A DOE $1 billion loan to support the restart closed in November 2025.

Meta's 6.6 GW commitment, announced in early 2026, is the single largest hyperscaler nuclear pledge to date. Jefferies analysts noted in a research note that it was roughly equal to the total capacity of Northern Virginia's Data Center Alley, the world's largest data center hub. S&P Global reported in March 2026 that Microsoft had surpassed Amazon as the largest corporate buyer of clean power, with 34.7 GW contracted as of end-September 2025.

The Nuclear Investment Stack: Three Different Bets of Operators, Uranium and SMRs

Nuclear energy stocks are not one trade. There are three separate plays here, with meaningfully different risk and return profiles. Here's a framework we call the Nuclear Stack.

Layer 1: The Operators

These companies run nuclear plants today and are signing long-term contracts directly with hyperscalers. Constellation Energy (CEG) runs 21 nuclear reactors, the largest commercial nuclear fleet in the US. In January 2026, CEG completed its $22 billion acquisition of Calpine Corporation, adding 23 GW of natural gas, geothermal, and solar capacity and transforming the company from a nuclear-focused operator into the largest competitive power generator in the country. 

Q1 2026 revenue came in at $11.1 billion, up 64% year-over-year, with GAAP EPS of $4.49. Management affirmed full-year 2026 adjusted EPS guidance of $11-12. Crucially, the nuclear fleet maintained a 92.3% capacity factor in Q1 2026. They've also continued building the hyperscaler pipeline: a 760 MW power agreement with CyrusOne in Texas was secured in Q1 2026, adding to over 5,650 MW of existing long-term clean energy agreements. Vistra Corp (VST) acquired Energy Harbor in 2024, adding over 4,000 MW of nuclear generation, and followed that with PPAs with both Amazon and Microsoft. 

Layer 2: The Fuel Chain

Uranium is to nuclear what lithium is to batteries. Without it, none of this works. The World Nuclear Association projects uranium demand will grow 28% by 2030 and more than double by 2040. The supply situation is structurally constrained: global uranium production in 2025 was approximately 173 million pounds while primary demand was around 204 million pounds, as per INN analysis citing industry data. 

Cameco (CCJ), one of the world's largest uranium producers with a 49% stake in Westinghouse Electric, reported Q1 2026 adjusted EBITDA of C$509 million on May 5, 2026. The uranium spot price, which touched $106.75 per pound in early 2024 before retreating, has stabilized at around $84-86 per pound as of June 2026 as per TradingEconomics, with Citi analysts projecting a move to $100-125 per pound through the rest of the year.

Layer 3: The Future Builders (SMRs)

Small Modular Reactors are the industry's bet on making nuclear deployable faster and closer to where the demand actually sits. A conventional nuclear plant takes a decade or more to build, costs $10-20 billion, and needs a major grid connection. An SMR is factory-built, modular, and can go up near a data center campus.

To understand it simply, a traditional nuclear plant is like building a centralized telephone exchange to serve an entire city. An SMR is like deploying self-sufficient modular towers exactly where you need coverage. You get the same output, in less time, with less upfront capital commitment.

The companies in this space backed by hyperscalers include Oklo (Meta), X-energy (Amazon), and Kairos Power (Google). The publicly listed ones are Oklo (OKLO) and NuScale Power (SMR). The catch is that both are pre-revenue or near-pre-revenue. Oklo posted zero revenue for full-year 2025 and an operating loss of $139.3 million, as per 24/7 Wall Street. NuScale generated just $31.5 million in revenue in 2025, down from $37 million the year before, with a net loss of $355.8 million. They're bets on a future that's real but not yet commercially proven.

Nuclear Energy Stocks and ETFs to Watch 

Risks of Investing in Nuclear Energy Stocks and Uranium

The nuclear re-rating is real. But it doesn't make the risks disappear.

• Construction timelines are the most consistent risk in this sector's history. Nuclear plants have always run late and over budget. Even with regulatory support from the Trump administration (four executive orders signed in 2025, targeting growth from 100 GW of US nuclear capacity today to 400 GW by 2050), actual timelines slip. Wood Mackenzie expects US nuclear generation to stay roughly flat until 2035, then grow 27% through 2060. The hyperscaler SMR commitments mostly deliver power in the 2030-35 window. That's a long runway for pre-revenue companies still burning cash.
• Uranium price volatility matters. The spot price swung from around $106.75 per pound in early 2024 down to the low $70s through much of 2025 before recovering to around $84-86 per pound in June 2026, as per TradingEconomics. Citi analysts project a move to $100-125 per pound through the rest of 2026. That recovery is encouraging, but the swings are real: a 30% decline in the spot price can cut uranium miner stock prices by significantly more, even if the company's long-term contract book is intact. 
• SMR companies carry a specific risk that is easy to underestimate. They need to be funded continuously from now until commercial operations, which may be five to ten years away. Oklo burned $82.2 million in operating cash in 2025. If capital markets tighten or sentiment turns, the cash runway gets precarious well before a reactor ever turns on.

The historical parallel worth keeping in mind is the US fiber optic buildout of 1999-2001. The premise was right: the internet would consume enormous bandwidth. Many individual telecom companies still went bankrupt because they ran out of cash before the demand materialized at the scale that justified their cost base. The AI electricity thesis is probably correct, but not every company in the theme will survive to see the payoff.