The company building America’s uranium supply from scratch
Topics
AI Data Center Power Bottlenecks
Venture Capital Bets Big on Hard Tech
Hard Tech R&D Discipline and Commercialization
Pioneers
Scott Nolan
Summary
- General Matter’s fuel-supply thesis rests on existing reactors as well as advanced nuclear. Nolan cites 94 US reactors supplying 20% of electricity, about a quarter of enrichment sourced from Russia, and import waivers ending January 1st, 2028. Utilities account for most contracting volume, cushioning the business from near-term AI power choices. Advanced-reactor founders’ bottleneck is more immediate: “It’s actually getting the fuel.”
- Higher enrichment enables compact reactors but can make fuel a dominant cost. HALEU at 19.75% supports smaller cores and potentially longer refueling cycles than today’s 3–5% fuel. At current market prices, fuel including enrichment can represent 50% of some SMRs’ levelized energy cost, depending on design: “the fuel cost matters proportionately more.” Availability is the first problem; cheaper enrichment is the second.
- General Matter promises enriched-uranium deliveries by the end of the decade. With just over 200 people, a recently submitted NRC application and a new Paducah facility planned, Nolan calls this the public commitment: “This is one we’ll hit.” A $900 million capacity-building award reduces the first facility’s financing burden, while licensing, construction and manufacturing remain critical execution work.
- The proposed cost advantage comes from modernization, with the enrichment process still undisclosed. Challenged on competing without starting 10x or 100x larger than incumbents, Nolan points to “decades of technology improvement that no one’s captured.” He promises to disclose the technique in the first half of next year; the ambition is competitiveness against Europe, Russia and China.
- Nolan’s nuclear bull case depends on faster construction and cheaper power. He says 20% of Vogtle’s cost overruns were interest-payment overruns, illustrating how long build times affect project economics: “You have to be the cheapest.” Collison presses the competing solar-plus-batteries cost curve; Nolan remains unconvinced by battery economics for highly reliable data centers, while allowing that orbital solar “could possibly win.”
- Nolan’s operating lesson is that incentives and execution speed can overpower incumbent scale. Boeing’s cost-plus structure contrasted with a SpaceX team he believed could win despite being at most 1/1,000th the size. General Matter is bringing construction capabilities in-house, while using aggressive internal schedules to expose bottlenecks without sacrificing architecture. “Bet on the small guy” is conditional on vision, differentiated insight and the ability to execute.
Deep dive
1. The fuel shortage opens a market that did not exist a decade ago
- Nolan’s discovery came while investing at Founders Fund in mid-2023: reactor founders said their hardest problem was fuel, not licensing. Asked why domestic suppliers could not make it, they replied, “What US companies?” The existing fleet alone comprises 94 reactors producing 20% of US power; enrichment suppliers are Russian or European.
- Fuel takes five steps: mining, conversion into gas, enrichment, reconversion into solid, then fuel-pellet fabrication. Nolan says most US uranium comes from Kazakhstan and Canada; enrichment comes from Russia, France and Urenco, a European consortium.
- The US once supplied over 90% globally, Nolan says, but first-generation enrichment lost competitiveness against Europe and Russia. After the Cold War, “Megatons to Megawatts” downblended Russian warheads into US reactor fuel. “Why are we running these more expensive plants” when trade supplies fuel? Later American technology efforts also struggled to become competitive.
- Russia supplies about a quarter of US enrichment and, Nolan says, is today’s only commercial HALEU supplier. Congress passed an import ban in mid-2024; waivers end January 1st, 2028. A decade earlier, fully contracted utilities and limited advanced-reactor demand left little room for an entrant. “Today, that window is open.”
2. Higher enrichment changes reactor size and the fuel bill
- Today’s light-water reactors typically use 3–5% enriched uranium, with a 1-gigawatt reactor refueled every 18 or 24 months. HALEU at 19.75% is “higher octane,” keeping smaller cores critical longer and enabling factory construction; remote designs may seek 5- or 10-year cycles. The 20% threshold changes the regulatory regime. Submarines run north of 90% to minimize refueling, which requires cutting them open.
- When Collison links enrichment to passive safety, Nolan corrects him: “The enrichment level itself does not lead to safety advantages, but reactor formats do.” His concrete example is molten salt: thermal properties self-regulate, and an overheating-triggered melt plug can drain the reservoir so “criticality stops.”
- The customer spectrum runs from Antares’ military/space applications and Radiant’s containerized 1-megawatt unit to Oklo, X-energy and TerraPower’s larger designs. Some choose LEU or LEU+ at 6–10% because it is available; AP1000s cannot run HALEU. For some SMRs, fuel including enrichment can represent 50% of levelized energy cost at today’s market prices, depending on design. That makes cheaper enrichment Nolan’s second lever, beyond availability.
3. Nuclear has to win on construction time and power cost
- The AI power rush has not changed near-term demand: “most of the contracting volume that we do is with the existing fleet.” Data centers’ initial reliance on gas would not change that. In five years, Nolan expects pressure toward nuclear; the gigawatt-versus-SMR split depends on deployment speed and cost per megawatt.
- Collison challenges him with solar plus batteries’ falling costs. Nolan believes solar’s curve, but has not seen equally compelling battery data; five-nines data-center availability needs substantial storage. He calls nuclear the cleanest, safest baseload, safer even than wind per terawatt-hour. Solar powering data centers in sun-synchronous orbit “could possibly win”; on the ground, he thinks nuclear wins.
- Build faster and build more, Nolan argues, to stabilize supply chains and workforces. At Vogtle, he says 20% of cost overruns were interest-payment overruns despite historically low rates. Ten-year construction amplifies financing costs; mortgages peaked, he thinks, at 18% in 1981. How do you finance huge infrastructure with capital costs “well north of 10%”?
- Collison challenges the accident-and-overregulation explanation for nuclear’s decline. Nolan allows safety concerns “maybe” started a pause, but calls the cycle “probably inevitable anyway when your product is not cost-competitive.” Cost-plus utilities outsourcing construction lacked strong price pressure. “It’s the cleanest base load, the safest base load, not the cheapest base load. You have to be the cheapest.”
4. The proposed advantage is decades of uncaptured technology
- Enrichment separates uranium “from itself”: natural uranium contains 0.711% U-235, with the rest primarily U-238. General Matter separates isotopes in a gas, with no nuclear or chemical reactions in its facility. Nuclear scientists and chemists are still needed, largely for safety analysis and preventing unwanted reactions.
- Nolan lists gaseous diffusion, gas centrifuges, aerodynamic, aqueous and laser approaches, but withholds General Matter’s technique until “first half of next year.” Collison mistakes Paducah’s former process for the new one; Nolan corrects him. The old diffusion plant closed in 2013. General Matter is building a fresh facility on the site.
- Collison pushes back: General Matter will not start 10x or 100x larger than incumbents, and the techniques are familiar, so where is the cost edge? Nolan adds “decades of technology improvement that no one’s captured” to scale and simplicity. To his knowledge, incumbents last did clean-sheet redesigns decades ago; public-domain technology dates to the ’50s or ’60s. The goal is competitiveness against Europe, Russia and China.
5. Paducah combines federal support with a local operating base
- General Matter began at the end of 2023 and now has just over 200 people, with engineering and headquarters in LA. Nolan says the NRC application was recently submitted. Paducah deliveries are targeted by the end of the decade, around advanced reactors’ first commercial deployments in ’28 and ’29; utility inventories and other contracts can bridge some of the import gap.
- After more than a year, over 10 states and over 1,000 plots, Paducah won for community acceptance and relevant talent. The 100-acre DOE site offers shared services, security and extensive environmental characterization, shortening preliminary work. “We’re building a facility that’s going to last decades,” Nolan says.
- Collison notes the NRC’s bad reputation; Nolan’s experience is “Great.” Commercial operations primarily require NRC licensing around safety, security and environmental impact; government sales or research can have DOE pathways. DOE land does not change this plant’s commercial regulatory route. Nolan also sees bipartisan support, citing Biden’s goal to triple US nuclear production by 2050 and Trump’s goal to quadruple it.
- General Matter won a $900 million congressional capacity-building award: down-selected under Biden, awarded under Trump earlier this year. Alongside equity and debt, that non-dilutive funding “takes a lot of that cost off the table,” helping finance the first facility. It addresses Collison’s concern that hardware companies can exhaust capital before proving themselves.
6. The schedule depends on manufacturing and disciplined incentives
- Collison challenges the deadline as a potentially overoptimistic SpaceX-style promise. “This is one we’ll hit,” Nolan answers. Internally, “green lights to Malibu” strips buffers to expose bottlenecks. But it can encourage waste, wrong architectural choices and difficult performance judgments; leaders must understand engineering well enough to give time back when a shortcut harms the long-term design.
- General Matter follows Tesla’s playbook with an in-house engineering, procurement and construction team, including its own general contractor. Licensing, construction and equipment manufacturing must converge on commissioning. Hardware has to meet the target economics: “Manufacturing is the thing that relatively few companies in the US have really proven that they can do at scale.”
- Nolan contrasts Boeing’s cost-plus program, where more people did not seem to mean more speed, with SpaceX, where he worked from 2003 to 2007. Within his first week or month there, he felt pretty certain SpaceX would eventually win despite being at most 1/1,000th the size. “It was just because of the speed of execution.”
- Nolan grants cost-plus a purpose when government urgently needs something unprecedented with highly uncertain costs. “Once you get to equilibrium, I don’t think that structure makes sense.” Large companies can retain urgency, he argues, by undertaking new projects that are “borderline existential,” rather than settling into collecting economic rents.
7. Founders Fund’s filter starts with a neglected bottleneck
- The SpaceX lesson was to find the missing bottleneck. “When your satellite costs a billion dollars, 100 million for a launch isn’t so bad.” But SmallSats and CubeSats exposed launch costs; smaller reactors similarly expose fuel costs.
- Founders Fund does not programmatically start companies, Nolan stresses. Leaving an attractive investing career required three conditions: an extremely important problem, nobody else solving it, and skills that could improve the odds. He searched for existing suppliers first. That narrow Venn diagram produced “extreme conviction” before launch; he thinks most remaining risk is execution.
- His resulting rule is “Avoid competition”: start with a small market that needs a new solution, make customers love the product, then expand. Asked where equally large opportunities are, Nolan’s honest answer is, “I don’t know what they are.” They become visible by going deep into a field, often ahead of a trend.