|
Part I
The Mechanism
|
Uranium.
The financial press is having a grand time with the AI-needs-nuclear story. Bloomberg runs pieces about hyperscaler power purchase agreements. CNBC has analysts explaining "the nuclear renaissance." They're not wrong, exactly. They're just staring at the demand dashboard while the engine is seizing three layers deeper.
Uranium doesn't go from the ground into a reactor. It passes through a fuel chain — mining, conversion, enrichment, fabrication — and right now, the chain is breaking at the link nobody on television can name: conversion. The step where yellowcake gets turned into uranium hexafluoride, the gas form that enrichment plants need as feedstock.
There are exactly three Western facilities in the world that do this. Three.
The spot price has been range-bound near $87. Analysts see that and call it consolidation. What they're not seeing is the long-term contract price at $94 — the highest since 2008 — which is where utilities actually buy fuel. That $6.45 gap between spot and term isn't noise. It's utilities telling you, quietly, that they can't secure physical fuel chain capacity. They're paying a premium not for pounds of yellowcake, but for guaranteed delivery through the bottleneck.
I've watched this pattern in other commodities. The term-spot spread is usually the last signal to move and the first one worth trusting. When buyers start paying tomorrow's price today, they know something the chart doesn't show.
|
Part II
The Diagram
|
The fuel chain, stripped to parts.
Step 1: Mining. Dig uranium ore, process it into yellowcake (U₃O₈). Global mine production: roughly 173 million pounds in 2025, covering about 85% of reactor requirements. The other 15% comes from inventory drawdowns and secondary supply — buffers that have been shrinking for a decade.
Step 2: Conversion. Turn yellowcake into uranium hexafluoride (UF₆). This is where the chain snapped.
Orano's Malvési plant in southern France — one of three Western converters — went down in January after severe flooding and didn't restart until late April. Three months of lost output. Orano's 2026 UF₆ production is now projected to fall 29%. That volume is not recoverable this year.
The other two: Cameco's Port Hope facility in Ontario, and ConverDyn's Metropolis Works plant in Illinois. ConverDyn was shut down entirely from 2017 to 2023 — six years dark — and is still ramping back to capacity. That's the entire Western conversion supply. Three facilities. One just lost nearly a third of its annual output. One spent six years offline. The third is running flat out.
Step 3: Enrichment. Convert UF₆ into low-enriched uranium (LEU). Russia controls 44% of global enrichment capacity — roughly 27.1 million separative work units. The US banned Russian LEU imports in 2024, hard cutoff January 1, 2028. That's sixteen months from today. Western enrichment runs about 25 million SWU against global demand of 50 to 60 million. The math doesn't work.
Step 4: Fabrication. Assemble fuel rods and load them into reactors. This step works fine. Nobody cares. It's steps 2 and 3 that are choking.
Reactor demand isn't slowing. Seventy-seven units under construction globally, thirty-seven of them in China alone. US data center electricity demand is projected to grow from 176 terawatt-hours to as much as 580 by 2028. Meta signed 7.7 gigawatts of nuclear PPAs. Microsoft signed over 800 megawatts. The demand signal is a freight train. The fuel chain has a two-lane bridge in the middle of it.
|
Part III
The Weak Link
|
The conversion bottleneck is bad. What makes it dangerous is what sits directly downstream: the enrichment cliff.
For the last two years, US utilities have been living on waivers. The 2024 Russian import ban included an exemption system — utilities could keep buying Russian LEU if the Department of Energy certified that alternatives weren't available. Which, in most cases, they weren't. That waiver window closes January 1, 2028. And Western enrichment capacity hasn't caught up.
Urenco's US facility in New Mexico has 4.3 million SWU of capacity. It's adding 700,000 SWU by 2027 and has plans for another 2.1 million — but that expansion doesn't come online until 2032. Centrus just finalized a $900 million HALEU contract with DOE at Piketon, Ohio. But its current output is roughly 900 kilograms per year. That's advanced reactor fuel, not utility-scale supply.
Here's where it gets uncomfortable. Kazatomprom — the world's largest uranium miner, roughly a fifth of global primary supply — just cut its 2026 production target by 10%. From 32,777 tonnes to 29,697. That's about 8 million pounds removed from a market already running a 10,000-tonne annual deficit between mine output and reactor demand. And the stated rationale for the cut? A deliberate "value over volume" market strategy — Kazatomprom withholding supply into a deficit. That's a producer exercising pricing power, not fighting a constraint.
Meanwhile, Sprott's Physical Uranium Trust keeps hoovering up spot pounds — 81.7 million and counting, valued at over $7 billion. Every pound locked in a vault in Ontario is a pound that won't pass through the conversion-enrichment chain. Financial buying is compounding the physical shortage at the exact moment the fuel chain can least afford it.
Last week, Sprott warned that European utilities will need to begin securing uranium supplies earlier than expected, given post-2030 requirements and continued reliance on Russian fuel chain services. The phrase "earlier than expected" is doing a lot of heavy lifting. I traded energy names during the 2022 gas crunch and learned that "earlier than expected" is the market's polite way of saying "yesterday." When the scramble starts, it won't be gradual.
|
Part IV
The Chain Reaction
|
The sequence is mechanical. If you've watched physical commodity squeezes — copper in 2021, nickel in 2022, cocoa in 2024 — the choreography is familiar even if the instrument is exotic. The difference here is that uranium has a longer fuse and a bigger blast radius, because reactor fuel cycles plan 18 to 24 months ahead. When the fuse runs out, there's no spot fix.
First: conversion stays bottlenecked. Orano's Malvési has been back online since late April, but 29% of 2026 output is already gone. That lost UF₆ doesn't get enriched, doesn't get fabricated into fuel rods, doesn't get loaded on schedule. Reactor operators don't run out of fuel overnight — they plan years ahead. But when planned deliveries start slipping, the phone calls get very short.
Second: utilities that have been nursing waiver-dependent Russian enrichment supply start locking in Western term contracts at whatever the ask is. Long-term price pushes from $94 toward $100-plus. Spot gets dragged up because the futures market finally prices in the fuel chain constraint, not just the mining constraint. Utilities placed 116 million pounds under long-term contracts in 2025 — still well above historical averages. That pace is about to accelerate.
Third: Kazatomprom's supply-disciplined output stays flat or declines while reactor demand grinds higher. The 10,000-tonne annual gap between mine production and consumption widens. The secondary supplies — a decade of inventory drawdowns that papered over the deficit — hit bottom. Sprott alone has removed nearly 82 million pounds from circulation. That's roughly half a year of global mine production sitting in a Canadian vault, doing nothing.
Where does capital go? Not into broad nuclear ETFs diluted with construction firms, consultancies, and companies that have never refined a gram of UF₆. The edge — if there is one — sits in companies that own fuel chain capacity. Cameco has mine-to-conversion integration through Port Hope. Centrus operates the only US facility licensed to produce HALEU. Urenco controls a third of current US enrichment demand. These aren't uranium miners. They're toll booths on the only road the fuel can travel.
The financial press will keep running AI-needs-nuclear headlines for another twelve months. They're looking at the destination. The fuel chain is the road. And the road has a hole in it.
