June 5, 2026 · 13 min read
Nuclear power was written off a decade ago. Today it is one of the most discussed investment themes on Wall Street — and the catalyst is AI. Microsoft, Amazon, and Google have all signed nuclear power purchase agreements in the last two years. Uranium spot prices have climbed from $30/lb in 2020 to nearly $90/lb. And a new generation of small modular reactors is approaching commercial deployment. Here is how to invest across the full nuclear supply chain.
The fuel supply chain for nuclear plants. Uranium prices have climbed from $30/lb in 2020 to ~$87/lb as demand from new and restarted plants increases and supply from Kazakhstan and Canada faces constraints.
Companies that own and run existing nuclear plants. Most direct beneficiaries of AI data center power purchase agreements. Revenue is predictable, asset-backed, and increasingly commanding clean energy premiums.
Next-generation nuclear technology with smaller, faster-to-deploy reactors. Higher risk but potentially transformative — NuScale has the first NRC-approved design, Oklo targets 2027–2028 for its first Aurora unit.
Four structural forces have converged in 2024–2026 to create the most compelling nuclear investment environment in decades. Each force independently would have been meaningful; together they are reshaping the energy sector.
AI training and inference workloads require 24/7 electricity at massive scale. A single large-scale AI data center campus consuming 1 gigawatt of power needs a reliable, always-on supply that intermittent solar and wind cannot guarantee. Nuclear plants generate dense, dispatchable baseload power — exactly the profile that hyperscalers need for decade-long supply contracts. The Microsoft–Constellation Three Mile Island power purchase agreement (20 years, $1B+) was not an isolated deal: it was the opening move in a wave of tech-nuclear partnerships that has continued with Amazon, Google, Meta, and others.
By 2030, AI data center power demand is projected to add more than 1,000 TWh per year to global electricity consumption — roughly equivalent to adding France's entire annual electricity consumption to the grid. Nuclear is uniquely positioned to supply a meaningful portion of this incremental demand under long-term contracts.
The US Inflation Reduction Act created a Production Tax Credit (PTC) for existing nuclear plants that had previously been uneconomic in low-power-price environments. The nuclear PTC pays approximately $15/MWh for qualifying carbon-free generation from existing reactors — meaningfully improving the economics of plants that otherwise might have been retired. Combined with DOE loan guarantees for new nuclear construction (over $15B committed to SMR and advanced reactor developers), the policy environment in 2026 is the most supportive for nuclear since the 1970s.
After decades of closures, nuclear plant restarts are accelerating. Three Mile Island Unit 1 — shut down in 2019 for economic reasons — restarted in September 2024 to supply Microsoft's data centers. Palisades in Michigan is undergoing the first-ever restart of a fully decommissioned US nuclear plant, with a planned restart in late 2025. The DOE identified approximately 40 GW of previously retired US nuclear capacity that is technically viable for restart with appropriate investment.
As the grid increasingly depends on variable renewable energy, the value of dispatchable carbon-free baseload power has risen significantly. Grid operators in PJM, MISO, and ERCOT are all flagging reliability concerns as coal plants retire faster than firm replacement capacity is added. Nuclear plants — which can run at 90%+ capacity factors for 18-month stretches between refueling — provide the grid stabilization that no other carbon-free technology can replicate at scale.
Uranium has gone from a forgotten commodity to one of the best-performing raw materials of the decade. The spot price tripled from $30/lb in 2020 to nearly $90/lb in 2026, driven by supply disruptions at Kazatomprom, new reactor orders, and speculative buying through uranium trust vehicles like Sprott (SPUT). Elevated uranium prices directly benefit miners' margins and signal tightening long-term supply/demand fundamentals.
Uranium miners are the most leveraged play on higher uranium prices. When spot uranium rises from $65 to $87/lb, a miner with $20/lb production costs sees its per-pound profit rise from $45 to $67 — a 49% margin expansion on a 34% price move. This operating leverage cuts both ways, making miners more volatile than operators but with larger upside in a strong uranium market.
The blue chip of uranium miners. Cameco operates tier-1 mines in Saskatchewan (Cigar Lake, McArthur River) with among the lowest production costs in the industry. It also owns 49% of Westinghouse Electric (uranium fuel services and reactor technology) through a JV with Brookfield. FY2025 revenue of ~$2.4B reflects strong uranium pricing and expanded fuel services. CCJ is the go-to large-cap uranium holding.
A pure-play US uranium miner focused on in-situ recovery (ISR) — the lowest-cost uranium extraction method. UEC has no debt and a portfolio of permitted ISR projects in Wyoming, Texas, and South Dakota. The company acquired Uranium One Americas in 2022, dramatically expanding its resource base to 140M lbs. Production costs of ~$18/lb position it attractively against current spot prices.
Denison's Wheeler River project in Saskatchewan's Athabasca Basin contains Phoenix (high-grade ISR) and Gryphon (underground). Phoenix received environmental assessment approval in 2023 and is moving toward construction. With 109M lbs in reserves and a low-cost ISR extraction plan, Denison represents a pre-production leveraged bet on sustained high uranium prices through construction and into production.
A small-cap US ISR uranium producer operating the Lost Creek mine in Wyoming. Ur-Energy has a Department of Energy contract to supply enriched uranium to the US strategic reserve — providing a portion of revenue at fixed, above-spot prices. The Lost Creek expansion and Shirley Basin project represent organic production growth potential. High-risk, high-leverage play on US uranium supply security.
Comparing the major investable uranium vehicles across revenue, production cost, resource base, and 1-year return gives a sense of where each sits on the risk/reward spectrum.
Nuclear operators own and run the generating plants that produce electricity from uranium fuel. They are the most direct beneficiary of AI data center power demand — hyperscalers need to sign PPAs with entities that actually operate always-on generating assets, and nuclear operators are the only large-scale carbon-free baseload suppliers in the US market.
The largest nuclear operator in the United States, with ~32 GW of generating capacity across 21 reactors in Illinois, Pennsylvania, New York, Maryland, and other states. Constellation is 100% nuclear and the purest way to invest in the AI data center nuclear power theme. The landmark 20-year PPA with Microsoft to restart Three Mile Island Unit 1 (September 2024) established a template for tech-nuclear deals. Amazon, Meta, and others have since signed Constellation PPAs. Forward P/E of ~28× reflects the scarcity premium on pure-play nuclear capacity.
Vistra's acquisition of Energy Harbor added ~4 GW of nuclear capacity to its existing natural gas, coal, and solar fleet. Total capacity of ~41 GW makes Vistra one of the largest independent power producers in the US, though nuclear now represents ~40% of its portfolio. VST has signed multiple data center PPAs and its diversified generation mix provides revenue stability nuclear-only operators lack during plant outage or refueling periods. Strong buyback program returning capital to shareholders.
NextEra is the world's largest generator of wind and solar power and also operates the FPL regulated utility in Florida. Its nuclear fleet (~10% of total capacity) includes Turkey Point and St. Lucie in Florida plus the Point Beach plant in Wisconsin. NEE's nuclear exposure is meaningful but not dominant. Its primary investment case centers on regulated utility earnings growth and its renewable development pipeline — the nuclear component is a stable baseload complement to variable renewables.
How do the major nuclear-exposed utilities stack up on capacity, nuclear concentration, valuation, and PPA activity? The table highlights why CEG commands a premium multiple as the only pure-play nuclear operator at scale.
| Ticker | Company | Total Capacity | Nuclear % | Fwd P/E | Key PPA Activity |
|---|---|---|---|---|---|
| CEG ⭐ | Constellation Energy | ~32 GW | 100% | ~28× | Microsoft TMI (20yr), Amazon, Meta |
| VST | Vistra Corp | ~41 GW | ~40% | ~22× | Multiple data center PPAs 2025–2026 |
| NEE | NextEra Energy | ~70 GW | ~10% | ~20× | Long-term regulated utility PPAs |
Small modular reactors represent the next generation of nuclear technology — factory-built, standardized reactor units of 50–300 MWe that can be deployed in modular configurations. The SMR market is projected to reach $18B by 2030. The key investment appeal: SMRs can be sited closer to demand centers (including industrial campuses and data centers), require less capital per unit than gigawatt-scale traditional plants, and can be deployed incrementally as power demand grows.
All current SMR stocks are pre-revenue or early revenue companies — they are technology bets rather than operating business investments. Risk is substantially higher than uranium miners or nuclear operators.
NuScale's VOYGR SMR design was the first small modular reactor design to receive NRC design certification, cleared in 2022. Each VOYGR module produces 77 MWe, and multiple modules can be combined into a single power plant (6-module plant = 462 MWe). NuScale is partnering with the DOE's Carbon Free Power Project at the Idaho National Laboratory site. The challenge: the first CFPP project was cancelled due to rising cost estimates; NuScale is pivoting to industrial and international markets. Stock is highly speculative, reflecting the gap between approved design and operating plant.
Oklo is developing the Aurora Powerhouse — a sodium fast reactor using spent nuclear fuel as feedstock. The 15 MWe design targets remote industrial and data center applications. Oklo is backed by Sam Altman (OpenAI CEO, chairman of Oklo board), giving it a high profile in the AI power discourse. First operational Aurora units are targeted for 2027–2028, pending NRC licensing. The sodium fast reactor design can consume used nuclear fuel, addressing waste concerns that have historically impeded nuclear expansion. Extremely early stage, very high risk.
BWX Technologies is not a pure-play SMR developer — it is the industrial backbone of the nuclear industry. BWXT manufactures nuclear propulsion systems for US Navy submarines and aircraft carriers, produces medical radioisotopes, and is a key supplier for commercial reactor components. As the US nuclear industry expands, BWXT benefits as a critical supplier regardless of which reactor designs win. This makes BWXT the most conservative way to gain SMR and nuclear expansion exposure, with existing revenue and earnings vs pre-revenue SMR developers.
Understanding why AI data centers specifically need nuclear power requires understanding the unique electricity requirements of large-scale AI compute.
AI model training runs continuously for weeks or months. A training run that completes 80% and then loses power faces significant setbacks — checkpointing helps, but interruptions are costly. More importantly, hyperscalers sign long-term PPAs to power their campuses and need a supply guarantee that matches the consumption profile: continuous, large-scale, and reliable over a 20-year horizon. Wind and solar cannot contract to guarantee 24/7 supply — they can only contract for average generation across time. Nuclear plants running at 90%+ capacity factor can guarantee near-continuous supply.
Microsoft, Google, Amazon, and Meta have all made net-zero electricity commitments. Data center electricity must be matched with carbon-free generation under their corporate accounting frameworks. For 24/7 carbon-free power at scale, nuclear is the only commercially available option. Gas (even with carbon capture) is not carbon-free. Large hydro is geographically constrained. Wind and solar require storage to achieve 24/7 matching, which is expensive and still limited at the required scale.
These deals are templates, not one-offs. Every major hyperscaler is now actively negotiating nuclear power agreements. The supply of available nuclear capacity is finite — existing US reactors are largely contracted, which is why Microsoft had to restart a shut plant and why SMR development has become commercially urgent.
For investors who want uranium exposure without single-stock risk, uranium-focused ETFs provide diversified access to the supply chain. Each ETF has a different portfolio construction approach and risk profile.
The nuclear investment opportunity spans a wide risk spectrum. A thoughtful portfolio allocates differently based on an investor's risk tolerance and time horizon. Here is a framework for sizing each bucket:
The nuclear investment thesis in 2026 is more substantiated than any point in the last 30 years. Three concrete things are happening that were not happening five years ago: major technology companies are signing decade-long nuclear power contracts; uranium spot prices have nearly tripled from post-Fukushima lows; and the first new SMR design has received NRC certification. These are real, not speculative, developments.
The primary risk is valuation — many nuclear stocks have already re-rated significantly, and the current prices require continued execution. CEG at 28× forward earnings is not cheap in absolute terms; the bull case requires continued PPA announcements and plant uptime. Uranium miners at today's prices assume $80+/lb uranium persists — if spot prices revert toward $60, margin compression would be significant.
The investment approach that makes the most sense for most investors: anchor the nuclear theme with CEG (pure-play operator, most direct AI demand beneficiary) and CCJ (blue chip uranium miner with fuel services diversification), add uranium ETF exposure (URNM or URA) for broader supply chain coverage, and allocate a small speculative position to one SMR developer (OKLO or BWXT depending on risk appetite) for technology optionality. This combination captures the full nuclear opportunity across all three sub-sectors without over-concentrating in the highest-risk plays.
Nuclear's second renaissance is real. The question is not whether nuclear plays a larger role in the 2030s energy mix — it will — but whether current stock prices already reflect enough of that future. The answer depends on how aggressively you believe AI power demand will scale, how durable uranium supply constraints are, and how quickly SMRs move from design to operating plant. Each of those factors has a meaningful range of outcomes. Size positions accordingly.
Compare the leading nuclear operators and miners side by side using live AI scoring and financial data.