The lost decade of underinvestment is over. With AI datacenters demanding 24/7 carbon-free power, nuclear is the only viable solution. The physical supply deficit is undeniable.
The uranium market is in a structural supply deficit. Primary mine production delivers approximately 130 million pounds of U3O8 per year. Global reactor demand consumes approximately 180 million pounds per year — and is growing toward 220+ million pounds by 2030 as new reactors come online in China, India, and the West restarts. The 50 million pound annual gap has been covered by secondary supply: drawdowns of utility inventories, government stockpile sales, and underfeeding of enrichment facilities. But these buffers are depleting rapidly.
New uranium mines take 10–15 years from discovery to first production. Even with current elevated prices, the supply response is painfully slow. No amount of money can accelerate geological timelines.
COP28 commitment to triple nuclear capacity by 2050. Microsoft, Amazon, and Google signing nuclear PPAs. 60+ reactors under construction globally. The demand side is as certain as it gets.
The AI revolution has a dirty secret: it is extraordinarily power-hungry. A single GPT-4-class training run consumes 50+ GWh of electricity — enough to power 4,500 American homes for a year. Hyperscaler data center power demand is growing at 30–40% annually. By 2030, data centers could consume 8–10% of total US electricity, up from ~3% today.
The question is not whether this power is needed, but where it comes from. The answer, increasingly, is nuclear.
Intermittent. Solar produces power 25–30% of the time (capacity factor). Wind: 30–35%. AI training runs 24/7 at 100% load and cannot tolerate interruptions — a power dip during a multi-week training run can corrupt days of compute. Battery backup for multi-day outages is prohibitively expensive.
Reliable (85%+ capacity factor) and flexible, but produces significant CO2 emissions. Tech companies have net-zero pledges. Microsoft, Google, and Amazon cannot build gas plants without violating their ESG commitments. Regulatory backlash is intensifying.
90%+ capacity factor (highest of any power source). Zero carbon emissions during operation. 60+ year operational lifespan. Proven technology with 70+ years of operating history. The only source that is simultaneously 24/7, carbon-free, and scalable to GW-level.
Excellent baseload (50%+ capacity factor), zero carbon during operation, but geographically constrained. Most prime hydro sites are already developed. New large-scale dams face environmental and regulatory opposition. Not scalable to meet AI demand growth.
Signed a 20-year PPA to restart Three Mile Island Unit 1 (837 MW). The plant will be renamed Crane Clean Energy Center. Expected restart: 2028. This was the signal that changed the market's perception of nuclear forever.
Amazon purchased the Cumulus Data Campus adjacent to Talen's Susquehanna nuclear plant (960 MW) for $650M. Direct co-location with a nuclear plant — the shortest possible distance between power generation and compute.
Google signed the first-ever corporate PPA for Small Modular Reactors with Kairos Power. Initial 75 MW by 2030, scaling to 500 MW by 2035. This validates the SMR pathway and signals a multi-decade commitment from big tech to nuclear.
AI training runs are the most demanding electricity consumers ever created. A large training cluster draws 100–200 MW of continuous power for weeks or months without interruption. Any power fluctuation, even for milliseconds, can corrupt a training checkpoint and force a restart — wasting millions of dollars in compute time.
This is why intermittent sources like solar and wind are fundamentally incompatible with AI training at scale. You need a power source that delivers constant, uninterrupted megawatts, 24 hours a day, 365 days a year. Nuclear is the only carbon-free source that achieves this — with a 90%+ capacity factor, no seasonal variation, and no weather dependence.
The math is simple: if the world is building hundreds of GW of AI compute capacity over the next decade, and each GW of compute needs roughly 1 GW of continuous power, and the only carbon-free baseload source is nuclear — then nuclear capacity must triple. That means dramatically more uranium demand.
The uranium supply-demand imbalance is one of the most well-documented in commodities. Primary mine production has not covered reactor demand since 2005. The gap has been bridged by secondary sources: government stockpile sales (the US-Russia HEU agreement, which ended), utility inventory drawdowns, and underfeeding at enrichment facilities. These secondary sources are finite and depleting.
The supply side is dominated by three countries: Kazakhstan (43% of global production via Kazatomprom), Canada (15% via Cameco), and Namibia/Australia (~20% combined). New mine development takes 10–15 years from discovery to first production. Even at $90/lb, the supply response is glacially slow.
Source: World Nuclear Association (WNA), TradeTech, Cameco, Kazatomprom. M lbs = million pounds U3O8.
| Producer | Country | Production (M lbs/yr) | Global Share | Reserves (M lbs) | Cost Curve Position |
|---|---|---|---|---|---|
| Kazatomprom (KAP) | Kazakhstan | ~56 | ~43% | 800+ | $25–35/lb (ISL) |
| Cameco (CCJ) | Canada | ~20 | ~15% | 450+ | $35–45/lb |
| Orano (gov. owned) | France (Niger ops.) | ~10 | ~8% | 300+ | $40–50/lb |
| Uranium One (Rosatom) | Russia / Kaz. JVs | ~12 | ~9% | 400+ | Sanctioned risk |
| BHP / Olympic Dam | Australia | ~8 | ~6% | 2,000+ (by-product) | By-product (low cost) |
| NexGen Energy (NXE) | Canada | 0 (pre-prod.) | 0% | 300+ (Rook I) | $15–25/lb (world class) |
Source: WNA, company annual reports, UxC, TradeTech.
Kazakhstan produces 43% of the world's uranium, almost entirely via In-Situ Leach (ISL) mining. In 2024–2025, Kazatomprom repeatedly lowered production guidance due to acid supply shortages, construction delays, and geological challenges. If the world's largest producer cannot meet its own targets, the supply deficit widens faster than anyone's models predict. This is the single most important variable in the uranium market.
The cost curve is the key to understanding where the uranium price needs to go. Currently operating mines (like Kazatomprom's ISL operations and Cameco's McArthur River) have production costs in the $25–50/lb range. But current production does not cover demand. To bring new mines online, the uranium price must exceed the incentive price for greenfield development: approximately $70–80/lb.
At the February 2026 spot price of ~$90/lb, we are above the incentive price — but only marginally for the higher-cost projects. And term contract prices (~$75/lb) are below the incentive threshold for many projects. The key insight: prices need to stay elevated for years to incentivize the 10–15 year development timelines for new mines.
Source: UxC, TradeTech, company filings. Cumulative production capacity plotted against estimated all-in sustaining cost (AISC). Current spot: ~$90/lb.
A cost curve ranks all producers from lowest to highest production cost, plotted against their cumulative output. It tells you the "marginal cost of production" — the cost of the most expensive pound of uranium needed to meet total demand.
In uranium, the cost curve has a distinctive shape: the first ~100M lbs/yr come from low-cost ISL operations in Kazakhstan and by-product mines in Australia at $25–40/lb. The next ~30M lbs come from conventional underground mines (Cameco, Orano) at $40–60/lb. To reach 180M+ lbs to cover demand, you need greenfield projects (NexGen's Rook I, Denison's Wheeler River, Peninsula's Lance) at $60–80/lb.
The critical point: the spot price must stay above the cost of the marginal producer ($70–80/lb) for a sustained period to incentivize the investment needed to close the supply gap. Any dip below this level risks delaying new projects and worsening the deficit.
Thesis: Cameco is the largest Western uranium producer and co-owner of Westinghouse Electric (nuclear fuel and services). It is the closest thing to a "blue chip" in the uranium space. McArthur River/Key Lake is the world's largest high-grade uranium mine. Long-term contracts provide revenue visibility. Westinghouse ownership gives downstream exposure to nuclear fuel cycle.
Thesis: NexGen's Rook I deposit in the Athabasca Basin (Saskatchewan) is the largest undeveloped, high-grade uranium deposit in the world. At 2.37% U3O8 grade and 300M+ lbs of reserves, it is a world-class asset. Pre-production status means no current revenue, making it a pure option on the uranium price. If Rook I gets built (targeting 2029 first production), NXE re-rates massively. The risk is permitting and construction execution.
Thesis: URNM provides diversified exposure to uranium miners AND physical uranium (via Sprott Physical Uranium Trust). Top holdings: Cameco (~15%), Kazatomprom (~13%), NexGen (~8%), Sprott Physical (~10%). Higher beta than CCJ but with built-in diversification. Ideal for investors who want the sector thesis without single-stock risk.
Thesis: URA casts a wider net than URNM, including nuclear fuel companies, reactor builders, and enrichment firms alongside miners. Lower beta, broader value chain exposure. Top holdings include Cameco, NexGen, Paladin Energy, CGN Mining, and nuclear services companies. Best for conservative investors seeking nuclear sector exposure without concentration risk.
Horizon: Medium to long-term (12–36 months). The uranium bull cycle is structural and multi-year. Unlike HBM (which may rebalance by 2028), uranium's supply deficit is expected to persist through 2035+.
Key catalysts: Kazatomprom production guidance (quarterly), US nuclear policy updates, new reactor restart announcements, SMR permitting milestones, term contract pricing updates.
Sizing: CCJ = 3–5% of portfolio (core). URNM or URA = 2–3% (satellite). NXE = 1–2% (speculative). Total uranium exposure should not exceed 8%. These are volatile, commodity-linked equities.
Scaling in: Enter 40% at initial entry zone. Add 30% if spot uranium dips below $80/lb (buying opportunity). Reserve 30% for catalyst-driven entries (post-earnings, policy announcements). This is a multi-year trade — patience is the edge.
Small Modular Reactors (SMRs) are the next frontier of nuclear energy. Unlike traditional 1,000+ MW reactors that take 10–15 years to build and cost $10B+, SMRs are designed to be factory-built, truck-transportable, and deployed in 3–5 years. Capacities range from 50 to 300 MW — perfect for data center co-location, remote communities, and industrial applications.
While traditional reactors remain the backbone of the nuclear renaissance (60+ under construction globally), SMRs represent the growth catalyst that could dramatically expand nuclear's addressable market — and with it, uranium demand.
| Company | Design | Capacity | Target Timeline | Key Backer | Status |
|---|---|---|---|---|---|
| NuScale Power (SMR) | Light Water (PWR) | 77 MW per module | 2030–2031 | Fluor Corp, US DOE | NRC Certified |
| Kairos Power | Fluoride Salt-Cooled (KP-FHR) | 75 MW (initial) | 2030 (demo) | Demo construction | |
| X-energy (Xe-100) | High-Temp Gas (HTGR) | 80 MW per module | 2030–2032 | Dow Chemical, US DOE | Licensing review |
| TerraPower (Natrium) | Sodium-Cooled Fast | 345 MW | 2030 (demo) | Bill Gates, US DOE | Construction started |
| GE Hitachi (BWRX-300) | Boiling Water (BWR) | 300 MW | 2029 (OPG) | Ontario Power Gen. | Site prep underway |
Source: WNA, NRC, company announcements. Timelines are targets and subject to regulatory approval.
If even 10% of planned SMR deployments materialize by 2035, they could add 20–30 GW of new nuclear capacity, requiring an additional 40–60 million pounds of U3O8 annually. This is on top of the 62 traditional reactors already under construction and the hundreds planned for the 2030s.
The key takeaway: current uranium supply projections do not adequately account for SMR demand. If SMRs succeed, the supply deficit grows dramatically. If they fail, traditional reactor construction still ensures growing demand. Either way, uranium demand goes up.
Nuclear energy carries unique risks that other commodity investments do not. The uranium thesis is compelling, but investors must soberly assess the tail risks that could derail the trade. These risks range from low-probability / high-impact (nuclear accident) to high-probability / moderate-impact (regulatory delays).
The single biggest tail risk. A major nuclear accident anywhere in the world would cause governments to freeze reactor programs, potentially for years. Fukushima (2011) wiped 80% off uranium miners and created a 13-year bear market. Modern reactor designs are dramatically safer, but perception matters as much as reality. Risk: low probability, catastrophic impact.
NRC licensing for new reactor designs takes 3–5+ years. Environmental reviews add more time. NIMBY opposition is fierce. SMR timelines have already slipped multiple times (NuScale's UAMPS project was cancelled). Delays do not kill the thesis but they defer demand growth. Risk: high probability, moderate impact.
Nuclear energy remains politically divisive. Germany shut down its entire fleet in 2023. Some political parties campaign on anti-nuclear platforms. A shift in government policy (especially in the US after 2028 elections) could slow the renaissance. However, bipartisan support for nuclear is currently at its highest in decades. Risk: moderate probability, moderate impact.
Russia controls ~44% of global uranium enrichment capacity (via Rosatom/TENEX). The US and EU are working to reduce this dependency, but Western enrichment capacity (Urenco, Orano) will take years to scale. The 2024 US ban on Russian uranium imports creates near-term disruption but is bullish long-term for Western producers. Risk: moderate (geopolitical wildcard).
On March 11, 2011, a magnitude 9.0 earthquake and subsequent tsunami struck the Fukushima Daiichi nuclear plant in Japan. The resulting meltdown of three reactors was the worst nuclear accident since Chernobyl (1986). The consequences for the uranium market were devastating: spot uranium prices fell from $73/lb to below $20/lb over the next five years. Japan shut down all 54 reactors. Germany committed to a full nuclear exit.
The lesson for investors: the uranium bull thesis can be destroyed overnight by a single event. This is why position sizing matters enormously. Even if you are 95% confident in the structural supply deficit, the 5% tail risk justifies keeping total uranium exposure below 8% of your portfolio.
The counterpoint: modern reactor designs (Gen III+, SMRs) incorporate passive safety systems that make Fukushima-style meltdowns physically impossible. The AP1000 (Westinghouse) and EPR (EDF) can safely shut down without any operator intervention or external power. The industry is dramatically safer than 2011. But markets respond to headlines, not engineering.
Our thesis is that the uranium supply deficit (8/10 severity) will persist and deepen through 2030+, driven by AI power demand and reactor construction. Here is our scorecard.
| Date | Event | Significance | Impact |
|---|---|---|---|
| Mar 2026 | Kazatomprom Q4/FY2025 Results | Production guidance for 2026, acid supply update, expansion progress | High |
| Apr 2026 | Cameco Q1 2026 Earnings | McArthur River production, term contract book update, Westinghouse synergies | High |
| May 2026 | Cameco Investor Day | Long-term strategic outlook, new project announcements, demand forecasts | Medium |
| Jun 2026 | NexGen Rook I Environmental Assessment | Key permitting milestone for the world's largest undeveloped deposit | High (for NXE) |
| Sep 2026 | World Nuclear Association Symposium | Industry-wide supply/demand updates, new contract announcements | Medium |
| 2028–2029 | TMI Unit 1 Restart (Crane Clean Energy Center) | Symbolic milestone: nuclear powering AI. First Big Tech nuclear PPA operational. | Critical |
Uranium is in a structural supply deficit. Primary production (130M lbs) covers only ~72% of reactor demand (180M lbs). The gap is growing. Secondary supplies are depleting. New mines take 10–15 years.
AI needs nuclear, and nuclear needs uranium. Microsoft, Amazon, and Google are signing nuclear PPAs because no other power source delivers 24/7 carbon-free baseload at scale. This is a multi-decade demand driver.
The cost curve supports sustained high prices. Incentive price for new mines is $70–80/lb. Spot at $90/lb is above this, but term contracts at $75/lb are borderline. Prices need to stay elevated for years to incentivize new supply.
CCJ is the core holding; NXE for convexity; URNM/URA for diversification. Size positions carefully (max 8% total uranium exposure). This is a multi-year trade — the tail risk of a nuclear accident demands disciplined sizing.
Disclaimer: This analysis is for educational and informational purposes only. It does not constitute financial advice, investment recommendations, or a solicitation to buy or sell any security. All trade setups are hypothetical and based on technical and fundamental analysis as of February 2026. Past performance is not indicative of future results. Always conduct your own due diligence and consult with a licensed financial advisor before making investment decisions. Market Watch and its authors may hold positions in securities discussed. Nuclear energy investments carry unique risks including regulatory, political, and safety-related risks that may cause extreme volatility.