Uranium gained 11.7% in 2025 to $81.55/lb, and the miners who dig it up rose 154.65% — silver miners jumped 199%. The price move is the easy part to see. The harder, more durable story is underneath it: hyperscale AI companies have started signing multi-decade power purchase agreements directly with nuclear operators, betting on a fuel supply chain — mining, conversion, enrichment — that cannot expand nearly as fast as a data center can be built.
Solar and wind are cheap and fast to build, but a data center running frontier AI training and inference around the clock needs power that does not depend on weather or time of day. That single operational fact has sent hyperscalers — Microsoft, Amazon, Google, and others — into direct, multi-decade power purchase agreements with nuclear plant operators, restarted retired reactors, and small modular reactor (SMR) developers, a corporate-nuclear-demand pattern that essentially did not exist five years ago at this scale. Discovery Alert's own 2026 analysis calls this a genuine "corporate nuclear renaissance" — not government policy pulling nuclear forward, but Big Tech balance sheets doing it directly.
The market has priced this in fast. 2025's uranium spot price gain of 11.7% to $81.55/lb undersells the real move — precious- and industrial-metals miners as a group rose 154.65% over the same period per Sprott's own tracking, with uranium specifically named among the ten defining commodity themes of 2026 alongside gold, silver, copper, and rare earths. The World Nuclear Association's own 2026 supply-demand outlook is explicit that this is a structural, multi-year tightening, not a speculative spike — global reactor fleets are growing (new builds, restarts, and life extensions all add demand simultaneously) at a moment when mine supply additions require years of permitting and capital lead time to answer.
A data center takes two to three years to build. A new uranium mine, or a new enrichment facility, takes closer to a decade. Big Tech just committed to the fast side of that mismatch at unprecedented scale.
Raw uranium reserves are not the binding constraint — the USGS's own 2026 assessment confirms globally identified resources are substantial. The issue is production concentration: Kazakhstan alone accounts for a large plurality of global mined uranium output, largely through state-linked producer Kazatomprom, with Canada's Cameco and a small number of other majors making up most of the rest. That concentration means geopolitical or operational disruption in a small number of jurisdictions can move global supply meaningfully — a structural fragility layered on top of otherwise adequate long-run reserves.
Mined uranium (U3O8) cannot go directly into a reactor — it must be converted to UF6 gas and then enriched, a capital-intensive, technically demanding process historically dominated by a small number of facilities. Russia's Rosatom has long held an outsized share of global enrichment capacity, a dependency the US and its allies have been actively trying to unwind since 2022 through new domestic enrichment investment — but Discovery Alert's own 2026 infrastructure analysis is blunt that replacing Russian conversion and enrichment capacity is a multi-year, multi-billion-dollar undertaking that is still in its early stages, not a solved problem.
Small modular reactors are the demand side's own attempt to shrink the multi-year construction timeline that makes matching new nuclear supply to data-center-speed demand so difficult. Multiple Big Tech-linked SMR programs are underway, and the nuclear fuel market's own 2026 outlook names SMR expansion as one of its principal growth drivers alongside energy-security and decarbonisation policy. The honest caveat: very few SMR designs have reached commercial operation at scale as of 2026 — this remains substantially a construction and licensing bet, not yet a proven, repeatable technology deployment.
This publication's own Energy Security trajectory documented $3.4 trillion in 2026 global energy investment rebuilding grids, nuclear, and LNG around a single lesson: never depend on a strait — or a supplier — you don't control. Uranium's fuel chain is the sharpest, most concrete example of exactly that lesson still being unresolved. The grids and reactors being built are the demand side of energy security; the mining, conversion, and enrichment capacity needed to actually fuel them reliably is the supply side that policy and capital have addressed far more slowly — precisely the gap a genuinely tight uranium market is now pricing.
The demand story is real and multi-decade; the supply response is genuinely years behind it. Big Tech's power purchase agreements run for decades, but new mines, conversion plants, and enrichment facilities take the better part of a decade just to permit and build — this mismatch, not any single company's announcement, is the actual investable trajectory.
Enrichment dependency on Russia is the single most underappreciated risk in the entire AI-energy narrative. Commentary on AI power demand overwhelmingly focuses on reactors and mines; the conversion-and-enrichment chokepoint gets far less attention despite being, per Discovery Alert's own 2026 analysis, the hardest link to rebuild quickly.
SMRs are a demand-side hope, not yet a supply-side certainty. Treat SMR-linked announcements as options on a technology still proving itself commercially, not as already-derisked new nuclear capacity — the historical base rate for large nuclear projects arriving on time and on budget is poor, and there is limited evidence yet that SMRs will be structurally different.
2025's price and miner-equity moves already reflect a lot of this thesis — the easy money is not obviously still on the table. Uranium's 11.7% gain and miners' 154.7% rise mean a meaningful re-rating has already happened; the remaining opportunity is more about which specific link in the fuel chain (mining vs. conversion vs. enrichment vs. SMR technology) still has the most genuine scarcity ahead of it, not a generic "buy uranium" call at this stage.
Part of an ongoing journal — observations recorded when something in the world economy is worth saying. No schedule. No noise. Not investment advice.