The Nuclear Renaissance: How AI's Insatiable Power Hunger Is Restarting an Industry

Executive Summary

The numbers are stark. A single large AI training cluster consumes as much electricity as a small city. Hyperscaler data centers — the physical backbone of the AI revolution — added more than 50 gigawatts of new power demand globally in 2025, a figure that analysts project will double again by 2028. Renewables are expanding rapidly, but solar panels do not generate power at midnight and wind turbines stop when the air is still. Batteries help, but not at data-center scale. The grid needs firm, always-on, zero-carbon power — and the only proven technology that fits that description is nuclear.

The result is a reversal so sharp it would have seemed implausible in 2015, when Germany was shutting reactors and the Fukushima disaster had poisoned public opinion globally. In 2026, Microsoft has signed the largest corporate nuclear power purchase agreement in history, restarting Pennsylvania's Three Mile Island Unit 1 under a 20-year contract. Google is funding the construction of six small modular reactors with Kairos Power. Amazon has invested $500 million in X-energy's modular reactor programme. The US, UK, France, Japan, and South Korea are all extending reactor lifetimes and accelerating new build programmes.

Uranium prices have responded. The spot price touched $106 per pound in early 2024 — the highest level since the Fukushima spike — and has consolidated in the $85–100 range through 2026 as long-term contract signings absorb near-term supply. Cameco, the world's largest publicly traded uranium producer, has more than tripled in value since 2022. The Global X Uranium ETF (URA) is up over 180% from its 2020 lows.

This article examines the structural drivers behind nuclear's comeback, the investment landscape, the risks that remain real, and how individual investors can think about exposure to this once-forgotten sector.

Why AI Changed Everything for Nuclear

The Scale of Data Center Power Demand

The AI boom has fundamentally altered electricity forecasting models that were built for a world of moderate, predictable demand growth. Consider the numbers:

The International Energy Agency (IEA) revised its global electricity demand forecast in early 2026, attributing 35% of incremental demand growth through 2030 to data centres and AI infrastructure. This is a demand shock of a magnitude not seen since industrial electrification.

Why Renewables Alone Cannot Fill the Gap

Solar and wind are the cheapest new sources of electricity at the point of generation — but they share a structural flaw for data centre operators: intermittency. A hyperscaler running inference workloads 24/7 cannot afford to power down when clouds roll in or the wind drops. The alternative — overbuild renewables dramatically and pair them with grid-scale batteries — works at small scale but runs into fundamental cost and physics constraints at gigawatt-hour scale.

Nuclear, by contrast, operates at capacity factors of 90–93%, meaning the plant produces near-maximum output virtually all the time. A single 1-GW reactor can reliably power approximately 700,000 homes — or one large AI campus — with zero carbon emissions and without depending on weather.

This is why Microsoft's energy team didn't just buy renewable credits for Three Mile Island. They negotiated a 20-year power purchase agreement for actual baseload electricity from a reactor, committing to absorb the full output of the plant in exchange for funding its restart.

The Key Catalysts in 2026

1. The Big Tech Nuclear Land Grab

Every major hyperscaler has now made a direct nuclear investment:

• Microsoft: 20-year PPA with Constellation Energy to restart Three Mile Island Unit 1 (Pennsylvania). First power delivered in late 2024. Additional SMR commitments with TerraPower.
• Google: Partnership with Kairos Power to build six Hermes-2 fluoride salt-cooled reactors by 2035, with an option for additional units. Total committed capacity: ~500 MW.
• Amazon Web Services: $500 million investment in X-energy's Xe-100 high-temperature gas-cooled modular reactor. Also acquired a nuclear-powered data campus in Susquehanna, Pennsylvania.
• Meta: Announced an open request for proposals seeking 1–4 GW of nuclear capacity by 2030 for its AI infrastructure expansion.
• Oracle: Chairman Larry Ellison has stated publicly that the company is planning a data centre campus to be powered by three small modular reactors.

The pattern is consistent: tech companies are not merely buying nuclear-generated electricity through the grid. They are contracting directly with reactor operators and, in several cases, funding the construction of next-generation reactors. This goes well beyond ESG signalling — it is a hard-headed infrastructure decision driven by grid reliability concerns.

2. Government Policy Reversals

The political winds on nuclear have reversed sharply. After a decade of phase-outs and moratoriums following Fukushima, governments are now:

• United States: The Nuclear Power Purchase Act and the Accelerating Deployment of Versatile, Advanced Nuclear for Clean Energy (ADVANCE) Act (signed 2024) streamlines NRC licensing, incentivises SMR deployment, and funds a $900 million SMR programme. The White House has set a goal of tripling US nuclear capacity by 2050.
• United Kingdom: The government has reversed its position and committed to 24 GW of nuclear by 2050, including the new Sizewell C plant and a £300 million SMR competition won by Rolls-Royce.
• France: President Macron announced a programme to build six new EPR2 reactors, reversing the prior phase-down policy. EDF's recapitalisation removed uncertainty about reactor lifetimes.
• Japan: Following a decade of near-total shutdown post-Fukushima, Japan has approved the restart of 17 reactors and announced plans to replace ageing plants with next-generation models.
• South Korea: Reversed the prior administration's phase-out policy and resumed construction of Shin-Hanul 3&4, while targeting nuclear exports to Poland, Czech Republic, and the Philippines.

The policy pivot reflects a growing recognition that climate goals and energy security are incompatible with shutting down the world's largest source of low-carbon firm power.

3. Small Modular Reactors: The Technology That Could Change Everything

Traditional nuclear plants — the kind at Three Mile Island or Sizewell — cost $10–20 billion to build, take 10–15 years, and routinely run over budget. This construction risk is the primary reason nuclear has struggled to attract private investment.

Small Modular Reactors (SMRs) are designed to solve this problem. Defined as reactors generating under 300 MW (versus 1,000+ MW for conventional designs), SMRs offer:

• Factory manufacturing: Standardised designs built in factories and shipped to site, reducing construction time and cost variability
• Passive safety systems: Walk-away safe designs that shut down without operator intervention or external power
• Flexible siting: Small footprint allows placement near demand centres, reducing transmission losses
• Faster deployment: Estimates of 5–7 years from groundbreaking to first power, versus 12–15 for large reactors

The SMR sector is pre-revenue for most players — these are 5–10 year construction timelines — but the combination of Big Tech backing, government contracts, and DoE loan guarantees has shifted the risk calculus meaningfully.

The Uranium Supply Squeeze

Nuclear power plants need uranium fuel. And the uranium market has a structural problem: years of underinvestment have left supply unable to keep pace with accelerating demand.

Why Uranium Supply Is Constrained

After Fukushima (2011), uranium prices collapsed from over $70/lb to under $20/lb by 2016. At those prices, mines were unprofitable. Cameco, the world's largest publicly traded uranium miner, placed its flagship McArthur River mine in care-and-maintenance. Kazakhstan's state producer Kazatomprom cut production targets. Global uranium mining capacity atrophied for nearly a decade.

Meanwhile, existing reactors continued consuming uranium, running down global inventories. When the nuclear demand story began to change in 2021–2023, the supply response was slow: mines take 7–12 years to develop from discovery to production. The result was a price spike that drove the spot price to $106/lb in February 2024.

Current Market Dynamics (2026)

• Spot price: $85–100/lb range, having pulled back from the 2024 peak as spot buying by utilities slows and long-term contracting picks up
• Long-term contract price: $70–80/lb, with most utilities now contracting for multi-year supply at these levels
• Supply deficit: The World Nuclear Association estimates a supply gap of 50+ million pounds per year by 2030 as new reactor demand outpaces new mine supply
• Kazakhstan risk: Kazatomprom (50%+ of global production) has repeatedly missed production targets due to sulphuric acid shortages and permitting issues — a structural vulnerability in global supply chains
• Secondary supply exhaustion: The drawdown of US government and utility inventories (a key supply buffer since the 1990s) is largely complete

The Investment Landscape

Uranium Producers

Cameco Corporation (CCJ) is the investment most directly exposed to uranium prices. Operating McArthur River (the world's largest high-grade uranium mine) and Cigar Lake in northern Saskatchewan, Cameco has long-term contracts covering a significant portion of production at locked-in prices. The company's 49% stake in Westinghouse (acquired 2023) adds nuclear services revenue — fuel fabrication, reactor components, and maintenance — providing exposure to the broader nuclear cycle beyond raw uranium.

Kazatomprom (KAP) — listed in London and on the Astana stock exchange — is the world's largest uranium producer but carries political risk as a Kazakhstani state entity. Its production challenges have paradoxically been bullish for uranium prices.

NexGen Energy (NXE), Denison Mines (DNN), and Uranium Energy Corp (UEC) offer higher-risk, higher-reward exposure to uranium development assets in the Athabasca Basin (Canada) and Wyoming (US), respectively.

ETFs for Diversified Exposure

The Sprott Physical Uranium Trust is a distinctive instrument: it buys and holds physical uranium (yellowcake), creating direct commodity exposure without operating company risk. The trust has been an active buyer in the spot market, and its purchases have been a meaningful factor in the recent price run.

Nuclear Utilities

Constellation Energy (CEG) is the largest US nuclear operator with 21 reactors across 12 sites. The company has been a primary beneficiary of AI-driven nuclear demand — its Three Mile Island restart is fully contracted to Microsoft, and the company is in active discussions for additional corporate PPAs. Constellation's stock has been one of the best-performing large-caps in the S&P 500 since 2023.

Vistra Corp (VST) operates a diverse generation portfolio including nuclear assets (Comanche Peak in Texas), and has been an aggressive participant in the data centre power contract market.

EDF (France, listed Paris) and E.ON (Germany) offer European exposure to the nuclear utility space, though both carry more complex regulatory and political risk profiles.

The SMR Plays

Most SMR developers remain private, but a few offer public market exposure:

• Oklo Inc (OKLO) — NYSE listed via SPAC merger, backed by Sam Altman. Pre-revenue but has the highest profile in the sector. Trades at a large premium to book value.
• NuScale Power (SMR) — The first SMR company to receive NRC design certification. Has faced project cancellations but secured new DoD and international contracts. Highly speculative.
• BWX Technologies (BWXT) — Manufactures nuclear components and operates the US Naval Nuclear Propulsion programme. More defensive exposure to the SMR build-out as a supplier to multiple developers.

Risk Factors: Why Nuclear Is Still Not For Everyone

Despite the compelling narrative, nuclear investing carries risks that demand clear-eyed assessment:

Construction and timeline risk: Large nuclear projects have a well-documented history of cost overruns and delays. Even Vogtle Units 3&4 (the only new large reactors completed in the US in 40 years) came in at $35 billion versus an initial $14 billion estimate, and years late. SMRs promise to solve this, but the first commercial units are still unproven at scale.

Regulatory risk: Nuclear licensing remains one of the most complex regulatory environments globally. Even with ADVANCE Act reforms, NRC licensing timelines are long and uncertain. A single high-profile accident anywhere in the world could re-trigger political backlash.

Uranium price volatility: The spot market is thin and volatile. Inventory releases, production surprises from Kazakhstan, or demand revisions can move prices sharply. Miners carry significant operational leverage to these swings.

Technology obsolescence: The SMR sector has multiple competing designs and no clear winner yet. Some will fail to reach commercial scale. Investors in individual SMR developers face genuine binary outcomes.

Long project timelines: Nuclear is a long-cycle business. The Kemmerer, Wyoming Natrium plant funded by Bill Gates and Warren Buffett won't produce power until 2030 at the earliest. Capital tied up in these projects faces long J-curves before generating returns.

Geopolitical supply risk: Russia supplies approximately 35% of global uranium enrichment services through Rosatom. US legislation to phase out Russian enriched uranium is in place, but alternative capacity is still being built. Until it comes online, this remains a vulnerability in the fuel supply chain.

Putting It Together: Who Should Consider Nuclear Exposure?

Nuclear is not a trading idea. It is a structural, multi-year thesis anchored in two durable realities: AI electricity demand is real and accelerating, and nuclear is the only firm low-carbon power source available at scale. The investment case does not depend on uranium hitting $150/lb or every SMR company succeeding — it depends on the simple observation that nuclear plants are being restarted, new ones are being built, and long-term contracts are locking in demand.

A sensible approach for individual investors:

• Core exposure via ETF: URNM or URA provides diversified exposure to the uranium miner complex without single-company risk. These are appropriate for investors with a 3–5 year horizon who believe in the demand thesis.
• Quality single-name via Cameco: For those willing to do the work, Cameco's combination of high-quality uranium assets, Westinghouse exposure, and disciplined contract book makes it the highest-quality single-stock expression of the theme.
• Utility angle via Constellation Energy: For a lower-risk, dividend-paying approach, Constellation offers nuclear exposure with regulated cash flows and a growing corporate PPA book.
• Speculation via SMRs: Oklo, NuScale, or BWXT for those with higher risk tolerance and a genuine conviction in SMR timelines — but size these as options, not core positions.

Position sizing matters: Nuclear is a high-conviction, long-duration theme — but it has been a volatile sector. A 3–8% portfolio allocation depending on risk tolerance gives meaningful exposure without concentration risk.

Key Takeaways

• AI data centres are driving an electricity demand surge that solar and wind cannot fully address due to intermittency — creating structural demand for firm, baseload nuclear power
• Microsoft, Google, Amazon, Meta, and Oracle have all made direct nuclear investments, signalling long-term corporate conviction beyond ESG signalling
• Uranium prices are structurally supported by a supply deficit built up during a decade of underinvestment — the McArthur River care-and-maintenance period has permanently tightened supply
• Small Modular Reactors are the most consequential technology development in the sector, promising faster, cheaper, safer construction — but remain pre-commercial for most developers
• Investment exposure ranges from conservative (Constellation Energy, Cameco, ETFs) to speculative (Oklo, NuScale)
• Risks are real: construction overruns, regulatory timelines, uranium price volatility, and geopolitical supply chain dependencies all warrant position sizing discipline
• The thesis is structural, not cyclical — the best investors in this space are thinking in 5–10 year horizons, not quarters

Sources: International Energy Agency · World Nuclear Association · Goldman Sachs Global Investment Research · Cameco Corporation IR · US Department of Energy · NRC ADVANCE Act Summary · Constellation Energy IR · Sprott Asset Management Uranium Report 2026 · Wood Mackenzie Nuclear Power Outlook