113% upside: NXE + the selloff that missed the point

$1.3 trillion in semiconductor market value gone in a week. Intel down 21%, Micron down 22%. If you've been waiting for the AI trade to get cheaper, it just did. The question is whether you're buying the right part of it.

While semiconductor stocks shed $1.3 trillion this week on fears that AI capex can't justify its valuations, NexGen Energy (NYSE/TSX: NXE) is sitting at roughly $9.13 — near its 52-week low, down with everything else, and almost entirely unrelated to whether Nvidia's forward P/E is 21x or 50x. That's the point.

NXE isn't a chip stock. It's a uranium developer. And the reason the AI selloff doesn't touch its thesis is that data centres need power long before they need the chips to fill them. The 20-analyst consensus price target sits at $15.81, with some targets as high as $21.20, implying 113% upside. The market just handed you a cheaper entry into a thesis that got structurally stronger this week, not weaker.

The catalyst that matters came in March. On 5 March 2026, Canada's Nuclear Safety Commission did something it hadn't done in over two decades: it issued NexGen a Licence to Prepare Site and Construct for the Rook I Project, clearing the final regulatory hurdle before groundbreaking on what is, by most measures, the highest-grade undeveloped uranium deposit on the planet.

The Arrow deposit holds 357 million pounds of U3O8 at a grade of 3.10%, and the feasibility study projects production of up to 14 million kilograms annually for 24 years. Enough to cover roughly a quarter of current global primary supply from a single mine. In its first six years, Rook I could produce almost 30 million pounds per year.

The stock didn't move on the licence news. Which, in a week where the market is re-rating AI infrastructure names on valuation fears, makes even more sense as a contrarian entry. The DeepSeek chip headlines that spooked semiconductors don't change electricity demand by a single kilowatt. Microsoft and other hyperscalers have reportedly entered preliminary talks with NexGen about directly financing the project — because the power problem exists whether or not the chip wars resolve in Nvidia's favour.

The real risk is the clock. The first 12 months of construction will cost around CAD$300 million, and first production is roughly four years out. Four years is a long time in a pre-revenue name. But uranium doesn't wait for comfortable entry points. The last time this asset class moved, it didn't telegraph the turn.

Mining uranium is only half the trade. The fuel has to be enriched before it goes near a reactor, and for the next generation of advanced reactors, it needs to be enriched to a much higher specification — High-Assay Low-Enriched Uranium, or HALEU. There is exactly one company in the Western world licensed to produce it: Centrus Energy (NYSE: LEU), which holds a regulatory monopoly on HALEU production in the United States.

That monopoly is starting to look valuable in real cash terms. In Q1 2026, Centrus reported $76.7 million in revenue, raised its full-year forecast to $450-500 million, and disclosed a contracted backlog of roughly $3.8 billion extending through 2040. Tucked inside that backlog: a $900 million HALEU task order from the U.S. Department of Energy, which amounts to a government guarantee on a meaningful chunk of future revenue. It also holds $1.87 billion in cash, which is a lot of runway for a company with a $450 million revenue base.

Centrus isn't the mine. It won't 10x on a uranium spot price spike the way NXE might. But if you want exposure to the nuclear supply chain without betting on a single development-stage project delivering on time, LEU is the cleaner vehicle. It already has revenue, contracted cash flows, and a structural position in a market segment with no competition. That combination is rare.

That's how much power Goldman Sachs projects US data centres will draw in 2026 alone, up from 31GW in 2025, rising to 66GW by 2027. To put that in terms that matter for the uranium thesis: data centres accounted for roughly 50% of all US electricity demand growth in 2025, according to the IEA, and the agency expects that share to hold through 2030. Nuclear is the only dispatchable, carbon-free baseload source that can deliver power at that scale, around the clock, without depending on weather. Every gigawatt of AI infrastructure built is, indirectly, a vote for uranium.

Every time uranium becomes a narrative, investors default to URA. It's the path of least resistance. Broad exposure, liquid, moves when the sector moves. It also owns Cameco at roughly 23% of the fund, which means you're mostly buying a large-cap producer with decent margins and a 78-times forward P/E, with a dozen smaller names stapled on for texture.

Cameco is a fine company. It's not where the asymmetry lives. The uranium spot price peaked above $101 per pound in January 2026 before pulling back, and long-term contract prices remain firm as utilities secure future supply. That's a spot-price pullback into structural tightness, which is historically when developers — not producers — generate the real returns. A development-stage asset like NXE prices off future production at future prices. Cameco prices off current production at current prices. Those are very different bets.

The risk with NXE is real: four years to first production is a long time, and capital cost inflation could eat into those margin projections. But you're not being compensated for that risk at $9.13 a share. That's the part the URA crowd misses.

NexGen Rook I gets its construction licence The full World Nuclear News breakdown of what the March 5 CNSC approval actually authorises — and what comes next in the 48-month build timeline. World Nuclear News

Centrus Q1 2026 earnings and backlog detail How $76.7M in quarterly revenue, a $3.8B contracted backlog, and a $900M DOE task order stack up for the only US HALEU enricher. Carbon Credits

Goldman Sachs on US data centre power demand The research team's projection that data centre power demand hits 66GW by 2027 — and what that implies for grid investment and nuclear demand. Goldman Sachs

Fortune on data centres and US electricity Why data centres drove half of all new US electricity demand growth in 2025 and are drawing political friction from bipartisan lawmakers. Fortune

Uranium Q2 2026 market review The INN deep read on spot price consolidation, term contract strength, and what utility contracting activity signals about the next leg up. Investing News Network

NXE analyst price target consensus All 20 analyst ratings on NexGen Energy in one place, including the Scotiabank target raise to C$22 and the May UBS buy reiteration. ChartMill

Re-reading Daniel Yergin's The New Map this week. Written in 2020. Every prediction about energy security feeling dangerously accurate right now. The parts about uranium are the ones he got most wrong.