# Cheeky Pint: When Satellites Got Cheap the Bottleneck Was Launch. Now Reactors Got Small and the Bottleneck Is Fuel Source: Realpha Blog (blog.getrealpha.com) Original article and charts: https://blog.getrealpha.com/en/blog/cheekypint-2026-11-03-the-company-building-america-s-uranium-supply-from/ > Notes on the Cheeky Pint episode published 3 November 2026, where Scott Nolan of General Matter explains why the United States has no uranium enrichment capacity of its own. He says about a quarter of American enriched uranium is still imported from Russia and the congressional ban takes effect on 1 January 2028, and that at today's market prices fuel can account for half the levelised cost of energy for some small reactor designs. Educational listening notes only — no investment advice, no price targets, and the judgements here may be wrong. Published: 2026-11-04 Locale: en Tags: uranium enrichment, podcast-notes, small modular reactors, supply chain bottlenecks, energy costs ![A steel-framed industrial building under construction at dusk, the first column in the foreground just bolted to its concrete pad, the bare slab receding into the distance and vanishing into unlit darkness](/covers/cheekypint-2026-11-03-the-company-building-america-s-uranium-supply-from-cover.png) > Men come in ten ranks, from king and duke down to the lowest attendant; let one rank be missing and the order of society will not stand. > > —— Song Yingxing, "The Works of Heaven and the Inception of Things", chapter "The Five Metals" (Ming dynasty, 1637; my own translation) The Cheeky Pint episode published on 3 November 2026 has the host talking with Scott Nolan, who leads General Matter, about uranium enrichment. He says about a quarter of American enriched uranium is still imported from Russia and that the import ban passed by Congress takes effect on 1 January 2028; at today's market prices, fuel can account for half the levelised cost of energy — the figure that spreads a reactor's upfront capital and operating costs across the power it sells — for some small reactor designs. My takeaway: once reactors get small, fuel becomes the first link to jam — and in America that link is currently empty. Their target is to be delivering from the plant in Paducah, Kentucky before the end of the decade. That target carries conditions: the licence application only went to the Nuclear Regulatory Commission in late October, the plant is still being built, and the enrichment technique they intend to use will not be disclosed until the first half of 2027. ## Why can't America make its own nuclear fuel? His answer is that it shut the capability down itself after the Cold War. Nolan says the United States built a first-generation enrichment technology to win the Second World War and stay dominant through the Cold War, and in that era it held over 90% of global supply — but that specific design could not hold up on cost against the routes Europe and Russia later took. After the Berlin Wall fell, America signed a trade agreement with Russia, imported Russian warheads, blended them down into fuel and ran its own reactors on it, under the nickname Megatons to Megawatts. With the Cold War over there was no reason to keep the more expensive plants running, so they closed, and for decades afterwards the country bought from Europe and Russia. ![A curve starts high at the upper left, slides all the way down to the floor, then runs flat along zero for decades, with a dashed line marking 2028 at the right and an empty dashed box beside it waiting to be filled.](/figures/us-enrichment-capacity-fell-to-zero-en.svg) The 94 reactors on the grid today supply 20% of American power, a quarter of their fuel still comes from Russia under a waiver that runs to 2028, and the rest comes from France and the European consortium Urenco. Most of the ore is dug in Kazakhstan and Canada, shipped across an ocean to be enriched, and shipped back as fuel. ![A wide band standing for an ocean separates the mines on the left from the US reactors on the right, and two arcs arch up over it to the enrichment plants and back again, while the enrichment box on the US side is left empty.](/figures/uranium-crosses-an-ocean-twice-en.svg) My favourite part of the episode is how he found this out. He had invested in several reactor companies at Founders Fund, so he asked them what the hardest part of the job was, and his own guess was licensing. They told him the hardest part was getting fuel — licensing was a process you could work through — and all the fuel had to be imported from Russia, which was difficult. He then asked why the American companies that make fuel for the existing fleet don't make what they need. The reply: what American companies? ## Why do small reactors need fuel enriched to just under 20%? Because a small core needs higher-octane fuel. Only 0.711% of natural uranium is the fissile U-235, the rest mostly U-238; the light water reactors on the grid run on low-enriched uranium at 3% to 5%, with a large core and a containment dome over it. He says the reactor itself is small; the dome over it is thick and big, sized for the most extreme accident scenarios, and it is part of what makes a plant a multi-billion-dollar construction project. Take enrichment up to 19.75% and the core can be small enough to mass-produce modularly in a factory and truck to the site in containers, and that is HALEU, high-assay low-enriched uranium. Twenty percent is the regulatory cutoff internationally and domestically; above it you are getting too close to weapons grade. Submarines run north of 90%, because refuelling one means cutting the hull open, so the fewer times the better — and no civilian company may touch that band. ![At one shared scale, the left side shows a thick, huge half-dome with a surprisingly small reactor inside it, while the right side shows a block the size of a shipping container, so the difference in volume is obvious at a glance.](/figures/dome-versus-container-same-scale-en.svg) Why the small core and the high enrichment go together, he explains cleanly: as the fuel reacts its enrichment level drops, and at some point the reactor fizzles out. A gigawatt-scale reactor has a big core that stays critical for a long time, so refuelling every 18 to 24 months is acceptable; with a small core the slide from acceptable power output to not even being critical happens much faster, and some of these units sit somewhere remote and want five- or ten-year refuelling cycles. ![A scale runs from zero to one hundred percent with the submarine level marked near the top, and a short stretch at the left end is blown up into a second scale where natural uranium, grid-sized reactors and HALEU sit in order, the last one pressed right against a dashed boundary.](/figures/enrichment-scale-and-the-20-percent-line-en.svg) He also corrected an assumption of mine: safety comes from the reactor format, and the enrichment level has no say in it. In a molten salt reactor the plug at the bottom melts if things overheat, everything drains into a reservoir and criticality stops; in a helium-cooled design the helium cannot be activated, so a coolant release puts out no radioactive material. ![Three lines slope down to the right toward a dashed floor where the reaction dies, with the big-core line falling gently, the small core on low-enriched fuel hitting the floor fast, and the small core on high-enriched fuel starting higher and lasting longest.](/figures/burnup-slope-decides-refuelling-interval-en.svg) ## Did nuclear stall because of the accidents? He puts the main cause on cost. Safety fears may have paused it, but he says that gets blamed as the only thing. Vogtle ran over budget and over schedule, and in an era of historically low rates 20% of its overruns were interest overruns — from which he draws a line worth copying down: when a plant takes ten years to build, the overnight cost matters less than interest rates and how long you actually take. Oil prices shocked in the 1970s, rates peaked in 1981 with mortgages at 18%, and you cannot finance a huge infrastructure project at a cost of capital well north of ten percent. The other half is that nobody was pushing prices down. Utilities are cost-plus organisations, paid their costs plus a layer, and they outsource construction. He says the dynamic resembles the launch industry: once the economic pressure to cut prices is gone, it stops being a market and becomes a place where everyone extracts rent in their own corner, probably without anyone intending it. The question he throws back is hard to argue with. In the 1970s and 1980s global warming was not yet a topic; you tell an ordinary person there is a new technology, the safety arguments are confusing enough that they cannot tell which side to believe, and their electricity bill will be higher — why would they say yes? He says nuclear is the cleanest and safest baseload there is, with lower downstream mortality per terawatt-hour than wind and lower downstream CO2 than wind or solar, and the one thing it has not won on is cost. ## How does a new company beat Russia and Europe on cost? The host sets the challenge directly: cost advantages usually come from scale or simplicity of design, and your market is the size it is while everyone knows the technical routes. Nolan says there is a third one missing — decades of technology improvement that nobody has captured. The incumbents are Russia, China and two European firms using the same underlying technology, and to their knowledge the last clean-sheet redesign of enrichment was decades ago, with no public domain technology in the field since the 1950s or 1960s. Which approach they use, he says, gets discussed in the first half of 2027. What happens inside the plant is simpler than I assumed: no nuclear reactions, no chemical reactions, just the separation of a gas. Uranium comes out of the ground as a solid, gets milled into U3O8, is converted into a gas by a separate chemical facility, and only then arrives for isotope separation. The team is just over 200 people, few of them nuclear scientists, and those mostly do safety analysis — making sure those reactions do not happen. The site search has the most human detail in the episode. They spent over a year, went to more than ten states and looked at more than a thousand plots of land, and the answer was the obvious one: go back to the community that most recently did this. They felt the community mattered most, because the plant has to run for decades, and Paducah was familiar with enrichment, saw it as good for the area, and had the talent. The Department of Energy site had been heavily characterised environmentally, so the environmental work went faster than on an uncharacterised plot, and the 100 acres were just enough for a first facility. On money they took the non-dilutive route: Congress authorised a capacity-building programme alongside the ban, they competed for it, were down-selected under Biden and awarded $900 million under Trump this year, which takes a large slice of the first plant's cost off the table — hardware companies most often die burning through cash before revenue arrives. They do not outsource construction either; by his own account they copied the Tesla playbook and built in-house engineering, procurement and construction teams. He says people assume nuclear is about core technology or licensing, while for them the thing that decides it is manufacturing — and few American companies have proven they can do that at scale. ## How do I use this bottleneck-hunting idea myself? Write your guess down first, then go ask the people downstream what the hardest step is. The hole I have fallen into goes like this: a theme heats up, I follow it, and what I end up buying is the theme itself — which is where the most crowded money is. This episode offers a comparison. In the early 2000s everyone was talking about SmallSats and CubeSats, and then there was no cheap way to get them up. When a satellite costs a billion dollars, he says, a hundred million for the launch is not so bad; once satellites get cheap, the launch cost sticks out. Small reactors have the same shape: you no longer have a multi-billion-dollar construction project, and then the fuel jumps out, with its share of total cost rising until fuel plus enrichment can be half the levelised cost of energy on some designs. ![Two before-and-after stacked bars show that once the tall segment is cut down, the small segment underneath stays exactly as tall as before yet grows from a thin sliver to half the total, and the satellite-and-launch pair looks identical to the reactor-and-fuel pair.](/figures/cost-moves-to-the-next-link-en.svg) So I now take that question to my own holdings: once the cost of one link is removed, whose share jumps up. He also says he does not know where the other opportunities are, that it takes a lot of digging to see them, and that they usually sit ahead of a trend rather than inside one. The bar he sets for himself is worth borrowing: extremely important, nobody doing it, and genuinely matched to your own skills. The intersection of those three circles is small enough that the activation energy is high, which is why both Anduril and his own company began with a long search for someone already doing it — if there had been, it would not be worth starting a new company. His SpaceX internship stuck with me too: the company was under fifty people, and he had previously worked on a cost-plus military programme at Boeing, watching more and more people get added to the project without it going faster. By the end of his first month he was confident SpaceX would eventually win, when SpaceX was at most a thousandth of its rival's size. His reason was speed of execution — not one word of it was about technology. ## Worth a look - The Cheeky Pint episode of 3 November 2026, the source for this piece - General Matter's website and their fuel facility licence application to the Nuclear Regulatory Commission - The 2024 ban on Russian uranium imports, and the Department of Energy's HALEU availability programme alongside it - Public reporting on the cost and schedule of Vogtle units 3 and 4 - Stripe Press's "Built to Grow" and "Nomad Letters", the two new books mentioned in the episode ## The one thing to take with you **When one link gets cheap, the cost moves to the next link in the chain; to find out where things are stuck, listen to what the people downstream are complaining about.** Nolan's whole company started in that gap: he guessed licensing, the answer was fuel. He had no plan to start a company at the time, and the gap carried him here. It only works if you actually ask, and if you ask the person who gets stuck. ![Two columns each hold three ranked boxes, with lines joining the matching items, and one of those lines runs steeply from the bottom of the left column to the top of the right column, crossing the other two parallel lines.](/figures/the-gap-between-your-guess-and-the-answer-en.svg) Something I have tried that holds up well outside investing: before you ask someone what the hardest step in their job is, write your own guess on a piece of paper first — a colleague, your partner, a kid learning the piano — then put the two answers side by side. If you guessed right, your model of that person is accurate. If you guessed wrong, that gap is the thing you have been failing to see.