Building a domestic nuclear fuel supply chain

Start with the pipeline, because the shape of it explains the vulnerability. Uranium is mined and milled into yellowcake, converted into a gas by stripping off oxygen and adding fluorine, enriched, deconverted back to a solid, and fabricated into fuel. The US has some capability at every one of those steps and commercial-scale capability at all of them except the middle one. Geography compounds it: most of the uranium the US consumes comes from Kazakhstan and Canada, conversion runs through a single long-lived US plant in Illinois plus facilities in Canada and Europe, and enrichment has no US-owned commercial producer at all, though a European firm does operate one plant on US soil. The gap is about ownership as much as geography. One commercial detail makes the rest of the episode legible. Utilities buy the uranium and keep title to it; every step after mining is a service performed on somebody else’s material, priced per unit of separation rather than per pound of metal. Enrichment is a tolling business, which is why an enricher can be genuinely indifferent to the uranium price, and why the customer for enrichment is the plant owner rather than a fuel vendor.

How the gap opened is a story about technology losing a race. After the Berlin Wall fell the US entered the Megatons to Megawatts program, importing Russian warheads, downblending the material and burning it in American reactors, then sending depleted uranium back to Russia to be enriched. Russia built out gas centrifuges on the back of that work while the US was still running gaseous diffusion, a first-generation technology. Against European and Russian centrifuges the US could not enrich profitably, so it privatized the business and then progressively shut it down over about twenty years. The residue is a supply mix of roughly three-quarters European producers and one-quarter Russia. Kann frames that dependence as something Washington is unwilling to touch because it cannot afford to, and Nolan corrects him: Congress banned Russian uranium imports in 2024, and what keeps the material flowing is a waiver the Energy Secretary can grant when a utility has no alternative. Those waivers stop on January 1, 2028, and utilities have already started drifting their Russian share down ahead of that cliff.

The harder problem sits one layer further out, and it is where the episode’s real tension lives. Existing reactors run on low-enriched uranium at three to five percent; advanced reactors mostly want high-assay low-enriched uranium at 19.75%, just under the 20% line. The reason is geometric rather than ideological. A gigawatt-scale reactor has a large core and can reach criticality and burn for a long time on weakly enriched fuel, while a smaller factory-built core needs richer fuel to get criticality, good burnup and a refueling cycle that works. Making the richer fuel is the same separation process repeated; what changes is criticality control and the licensing that reflects it, so vessels holding uranium have to be smaller and the deconversion and fabrication steps need more care. Commercially, though, the two fuels are opposite problems. Low-enriched demand is known, stable and large, so a new producer is displacing an incumbent. High-assay demand is close to zero, exists only because the Department of Energy released enough to carry advanced reactors through first demonstrations and first deployments, and depends entirely on how fast those reactors scale. Kann puts the objection plainly: the error bars on 2030 to 2032 demand are enormous. Nolan does not dispute the uncertainty. He reframes it as a deadlock that only breaks if a producer moves first, since no reactor developer can raise money without a fuel path and no enricher wants to build for a market that does not exist, and says his company is taking the long side of the bet because someone has to. His confidence rests on load growth, specifically data centers for AI compute arriving on a grid that has not grown in decades and wanting clean baseload. That is a conviction, not evidence, and he presents it as one.

What is actually under construction is a hundred leased acres at the south end of the Department of Energy site in Paducah, Kentucky, the last place the US enriched uranium commercially before the gaseous diffusion plant there shut in 2013. The company searched for about a year across more than ten states and something like a thousand parcels and chose Paducah for the community’s comfort with enrichment, the residual workforce and the power already on a former Manhattan Project site. The plan is to be running before the end of the decade with enough high-assay capacity to serve US demand into the middle of the next decade and possibly to 2040, plus enough low-enriched capacity to displace adversarial imports. Every one of those figures is a target set by a company that has not yet produced a kilogram of anything, and should be read as such. And fixing enrichment moves the constraint rather than removing it: past roughly a doubling of US enrichment on US soil the conversion step needs expanding, and US-mined uranium still gets shipped out of the country to be converted and shipped back, which is why Nolan’s magic-wand answer is not enrichment at all but permitting reform for domestic mining.