Notes on Uranium

There are a handful of investment positions I hold that I expect to argue about for years before I'm vindicated or proved wrong. Uranium is the one I most enjoy arguing about, because the people who think it's a stupid trade typically don't know the supply-demand arithmetic, and the people who think it's the next great trade typically haven't stress-tested the upside.

I sit between the two. The position is sized for sleep. The thesis is real.

The demand side is no longer hypothetical

For a decade, the uranium thesis was structural but slow. Civilian nuclear had been quietly building under the radar (Asia adding capacity, Europe extending plant life) but the rate of change wasn't investable.

Three things have happened in the last eighteen months that change the picture. The first is policy reversal in Europe. France committed to fourteen new EPR2 reactors. Germany has formally extended the operating life of its remaining plants. The Netherlands, Belgium, Sweden, the UK, all moving in the same direction.

The second is the AI compute build-out. Hyperscaler power demand is rising faster than grid capacity in most data-centre geographies. Microsoft signed the Three Mile Island deal. Amazon, Google, Meta, all sourcing nuclear power for new compute campuses. The data-centre power demand growth alone is forecast to add 200 TWh of consumption by 2030. Nuclear is the only baseload that can scale into that timeline without unacceptable emissions.

The third is China's continued buildout. Forty-plus reactors operating, twenty under construction, fifty in the pipeline. Chinese demand alone exceeds current global uranium production.

The demand curve isn't theoretical. It's contracted.

The supply side is structurally constrained

Kazakhstan is roughly 40% of global production, dominated by Kazatomprom. Canada, via Cameco, is the other major Western source. Australia has the reserves but limited willingness to develop them at scale.

Restarting mothballed mines is slow. Five to seven years from greenfield to first production for a new project. The industry spent a decade in capex contraction after Fukushima. There's no quick supply response available even if prices spike.

The supply-demand gap, as currently modelled by the industry trade bodies, has been deepening since 2023. The market is in structural deficit, with utility inventories drawing down.

The pricing story

The U3O8 spot price has roughly tripled since 2020, from around $25/lb to $80–100/lb depending on the week. This is meaningful but small relative to the historical peak (over $130/lb in 2007). At marginal cost of production for new mines being roughly $60–80/lb, the price sits around incentive level rather than scarcity level.

The interesting question is what happens to the price when the demand acceleration genuinely materialises against the constrained supply. Most utilities have inventory through 2027–28. After that, the contract market has to renegotiate against producers who are sold out. This is where the price discovery becomes asymmetric.

How I'm positioned

A small position in URA (the uranium miners ETF), sized as approximately 3% of the equity portfolio. The vehicle is diversified enough that single-name risk is manageable. The position is held without a price target, because the upside is highly path-dependent on the demand acceleration.

The downside case is well-defined. A major reactor accident, a political reversal on civilian nuclear, or a global economic slowdown severe enough to delay the data-centre buildout. These are real risks. None of them are zero. None of them are likely on the current trajectory.

What makes this trade hard

The trade is hard not because the thesis is complex but because the time horizon is long and the price action is whippy. Uranium has fake-outs. The 2024 rally faded. The 2025 consolidation was painful. The 2026 reacceleration has just begun.

The Girardian observation that applies here. This is a non-mimetic trade by construction. The position isn't crowded with retail. The financial press treats it as a curiosity. The macro tourists haven't arrived. That makes the trade structurally lonely (you won't get social reinforcement for holding the position) and structurally durable in the same breath.

The lonelier a position, the more credible the thesis must be. The cybersecurity position is similarly lonely, for the same reason, and I hold both for the same structural argument.

The deeper point

Most investable theses with high asymmetry are uncomfortable to articulate at dinner. The ones that are comfortable to articulate are usually already priced. Uranium is uncomfortable. It's also one of the cleanest examples I can find of a multi-year supply-demand mismatch where the marginal buyer (Chinese SOE, U.S. hyperscaler, European utility) is price-insensitive and the marginal seller (Kazatomprom, Cameco) is supply-constrained.

The math is the math. The path is uncertain. I'm sized for the path.