Last July, President Biden signed into law the "Accelerating Deployment of Versatile, Advanced Nuclear for Clean Energy Act," a word salad that was constructed to create the acronym ADVANCE. What made the ADVANCE Act so notable was not the wordplay, but the 393-13 vote by which it passed the House and the 88-2 vote by which it passed the Senate. In our current political climate, it's hard to believe you could get a vote that lopsided on really anything. The merits of nuclear power appear to be the one thing both parties agree on.

With the sudden interest in nuclear power - supercharged by the possibility of demand from AI data centers - we thought it was a good time to provide some basic grounding from our seat as traders of uranium and uranium miners. We hope this will help you sort out the signal from the noise in the steady drumbeat of news.

Point 1: This will be slooooow in the developed world, way too slow to meaningfully align with urgent AI ambitions.

It takes ~8 years on average to build a nuclear plant, so nothing is going to sneak up on us. And there are 0 new full-size nuclear plants in development in the US. There will likely be some restarting of mothballed plants in the US, Europe, and Japan, but that will also be relatively slow and marginal in size. The buzz around SMRs is premature – there are incredibly challenging engineering and hardware problems that still need to be worked out. While we’d love to be wrong about this, we highly doubt any meaningful SMR capacity will come into the US this decade and think the merits and theoretical cost savings of them are probably overblown for most use cases. If massive AI data centers do ultimately get built in the near future, we believe they will be powered by natural gas.

Point 2: China has about as many reactors in active development (~30) as the rest of the world combined.

And as the old expression goes, practice makes perfect. A reactor in China costs about ~25% as much on average as one in the developed world, and is completed ~2 years faster.

Point 3: The world has plenty of uranium but nowhere close to enough uranium mines (or refining capacity).

Very roughly speaking, the world consumes 180mm pounds of uranium per year (penciling in 0 incremental AI demand) and pulls 140mm pounds out of the ground. The difference is mostly covered by drawing down inventories, plus a smattering of recycling. This imbalance has caused uranium term prices (which is the price at which utilities transact) to pop: +23% in 2021, +26% in 2022, +32% in 2023, and +22% in 2024. The largest miner by far is Kazatomprom in Kazakhstan, which is endlessly bogged down by operational issues and has also contractually committed more than half of its production to China. Cameco has the lowest-hanging-fruit in terms of bringing on additional production. But based on their public statements and our conversations with them and others, they seem very content...VERY content...to let prices continue to rise before flipping any switches and bringing on supply. When we map out short-term requirements for utilities we don't see much risk of a short-term squeeze, but we do think the fuse is lit and it is hard to know how long it is.

We hope you found this helpful. If you want to discuss this topic in more depth and/or how we think about investing in the space, let me know. - Josh