Should You Buy the Dip on This Soaring Energy Stock?
The uranium thesis for AI investors isn't about Energy Fuels specifically—it's about whether power constraints become a binding limitation on the hyperscaler capex cycle that's driving semiconductor demand. Energy Fuels' 400% run and subsequent 4% pullback is noise compared to the structural question: can data center operators secure enough reliable baseload power to justify the $200 billion-plus annual AI infrastructure spend that companies like Nvidia, Broadcom, and custom silicon efforts depend on?
The nuclear angle gained credibility when Microsoft signed a 20-year power purchase agreement with Constellation Energy to restart Three Mile Island, and Amazon dropped $650 million acquiring a data center campus directly connected to the Susquehanna nuclear plant. Google and Oracle have made similar moves. These aren't hedges—they're acknowledgments that renewable intermittency and grid constraints create genuine bottlenecks for facilities running 50-100 megawatt GPU clusters continuously. When Meta or Microsoft talks about building multi-gigawatt data center campuses, the power question becomes existential.
Energy Fuels produced roughly 745,000 pounds of uranium in 2023 and is ramping the White Mesa Mill in Utah. The company benefits from being the only conventional uranium mill operating in the U.S., giving it strategic value as utilities and government buyers prioritize domestic supply chains. But here's the reality check: Energy Fuels remains a small player in a market where Kazatomprom controls 43% of global production. The stock trades without meaningful earnings—it's a bet on uranium spot prices, which have moved from $50 per pound in early 2023 to around $90 recently, though still well below the $140 peak in 2007.
The 4% dip likely reflects profit-taking after an unsustainable run rather than thesis deterioration. Uranium equities are notoriously volatile and tend to overshoot both directions. Energy Fuels' market cap around $1.5 billion prices in substantial production growth and sustained elevated uranium prices. The risk isn't demand—utilities need to refuel existing reactors and new small modular reactors are entering development pipelines specifically for data center applications. The risk is supply response. Kazakhstan can increase output, and secondary supplies from decommissioned weapons remain available.
For AI infrastructure investors, the relevant insight is that power has become a competitive moat. Hyperscalers with existing nuclear agreements or access to stranded natural gas have advantages over competitors scrambling for grid connections with 5-7 year wait times. This creates asymmetry: companies like Microsoft and Amazon with locked-in power can deploy capital into AI compute more aggressively, while smaller cloud providers face genuine constraints.
The uranium trade itself is speculative with commodity price exposure that doesn't correlate cleanly with AI equity performance. Energy Fuels could double again or halve depending on spot prices that move on utility contracting cycles and geopolitical supply disruptions. The actionable takeaway isn't buying uranium miners—it's recognizing that power availability is becoming a material factor in data center site selection and capex deployment timelines. If nuclear restarts accelerate, that's bullish for the AI buildout continuing at pace. If they stall, expect hyperscalers to slow facility construction regardless of chip availability, which would matter significantly more for semiconductor valuations than a 4% move in a micro-cap uranium stock.