
The fuel before the fission
The West's nuclear value chain needs to be rearchitected
Written by Madelyn O'Farrell, Santosh Sankar, and Jon Bradford, 2026-10-12
In mid-September, more than a thousand nuclear leaders, financiers, and policymakers gathered in London for the 51st World Nuclear Symposium. The mood in the room was telling. At the opening Finance Summit, bankers admitted they feel like they are missing out if they have not yet done a nuclear deal.
We think the enthusiasm is well placed, for a reason most of the room is not focused on. Everyone wants to talk about reactors: new designs, small modular reactors (SMRs), life extensions, and the gigawatts coming back online. This is for good reason: the IEA's Electricity 2026 report forecasts nuclear generation growing at 2.8% annually through 2030, more than double the 1.3% rate of 2021–2025. This is driven in part by the data center demand shock, with electricity needs projected to grow from 460 TWh in 2024 to over 1,000 TWh by 2030 and 1,300 TWh by 2035. As a reference, 1,000 TWh is roughly equivalent to the entire electricity needs of Japan.
Reactors are the visible part of the story, but what actually determines whether the industry can deliver sits upstream, in the value chain functions that deliver feedstock to plants. This piece walks through the full chain, from mining ore, processing, and enrichment, to spent fuel, and expounds on where we believe there is the compelling opportunity for investment.
The Nuclear Fuel Value Chain
Each step of the nuclear fuel value chain has its own economics, geography, and timeline. What follows is an assessment of that sequence.


*China’s most recently available conversion capacity data dates 2021, and so is stale but provides a baseline estimate
*Russia’s enrichment demand data is an estimate based on total uranium demand
Mining
Of the six stages, mining is the one where the underlying resource is not actually scarce for the West. Kazakhstan supplies roughly 40% of world output, but Canada, Australia, Namibia, and the US all hold sizable identified resources, and idle capacity is sitting on standby:
- Five US in-situ recovery mines, where uranium is dissolved underground and pumped to the surface rather than dug out, sat idle at the end of last year. This represents roughly 4,400 tons of annual capacity, mothballed because domestic operations cost more to run than Kazakhstan's more favorable geology and low cost of labor
- Cameco, the world's largest publicly traded uranium company, says it could bring back roughly 12,000 tons of capacity globally if prices justified it
Constraints are economic and time-based, not physical or technical, which sets mining apart from conversion or enrichment. Utilities ordinarily buy fuel years in advance through long-term contracts, not on the spot market, so a producer needs a bankable offtake agreement before it restarts a mine.
That demand curve is worth taking into consideration. Global reactor uranium requirements sit at roughly 69,000 tonnes today, and the World Nuclear Association's reference case has that climbing past 150,000 tonnes by 2040, above 200,000 tonnes in its upper scenario.
In the short term, restarting idle mines is a meaningful solution to solving the shortage, as mines can return quickly. Cameco’s filings put a McArthur River restart at a minimum of nine months, and Uranium Energy Corp took under two years to bring its mothballed Christensen Ranch operation in Wyoming back to production. Shelved projects with permits already in hand are slower to restart than idle mines, but they can still come online in years rather than the decades a new project requires. That said, they are small next to the growth needed by 2040. Most of that supply must come from new mines, which take 15 to 20 years from discovery to first production, so the decisions need to be made now.
Conversion
Conversion is the stage we think deserves the most attention. After uranium is milled into yellowcake, it has to be converted into UF6 gas before enrichment can happen, and only five plants on the planet do this:
- Philippe Coste (France): Operated by Orano, this plant produced over 12,000 tons in 2025 against a nameplate capacity of roughly 16,000 tons, close to 29% of world capacity and the single largest share
- Metropolis Works (Illinois, US): owned by Solstice Advanced Materials, spun off from Honeywell; producing just over 10,000 tons this year against roughly 14,000 tons of nameplate capacity, close to a quarter of world capacity
- Seversk (Russia): Rosatom's Siberian Chemical Combine, capacity of roughly 15,000 tons (~ 26% of world capacity), losing Western access due to sanctions and the enriched-uranium import ban which came into effect August 2024
- Port Hope (Canada): Operated by Cameco, nameplate capacity of roughly 10,500 tons, about 19% of world capacity
- Jiuquan and Hengyang (China): Operated by CNNC, run as a closed domestic system rather than a competitor for international customers; estimated combined capacity of 12,000 to 18,000 tons, comparable in scale to Rosatom's or Orano's, but built solely to supply China's own reactor fleet
Metropolis Works is sold out through 2028, with a backlog reportedly worth north of $2B. Utilities are locking in capacity years ahead of a potential shortage.
This is also where we see the most interesting near-term build-out, because conversion is comparatively cheap to expand relative to enrichment. Examples of recent builds include:
- Raven-Flint Nuclear: selected by the DOE for a pilot plant, targeting 500 tons a year, using a process that eliminates elemental fluorine from the chemistry
- FluxPoint Energy: building the first entirely new US conversion facility in >70 years, with an initial 2,500 tons, scaling to 10,000 tons in production in 2030
- Uranium Energy Corp: pursuing a conversion facility with an initial 10,000 tons a year, which would make it the largest conversion plant in the US
- Westinghouse: evaluating reviving conversion at its existing Lancashire site, targeting ~5,000 tons a year of natural uranium conversion plus another 2,000 tons dedicated to reprocessed uranium
Enrichment
Enrichment’s capacity is geographically concentrated, and even before the war reshaped supply chains, the West's capacity was not enough to meet its demand.

Closing this gap is genuinely hard for reasons that compound on each other. That’s not to say it is off limits for venture investment in the current cycle, but worth considering.
- Capital: these are multi-billion-dollar projects. For example: Orano's Project IKE in Tennessee is estimated at $5B for roughly 7.4M SWU a year, and Urenco's New Mexico expansion adds 2.1M SWU, nearly 50% on top of the site's current capacity, and takes Urenco's cumulative investment there from more than $5B to more than $8B.
- Time: even funded projects are not near-term fixes. Urenco's expansion begins construction in 2029 with first production only in 2032; General Matter, the Founders Fund-backed enrichment startup, is building a new facility that is not expected online until 2034. Every project we can point to lands between 2032 and 2035.
- Capability: manufacturing of centrifuges - the machines that perform the enrichment - is a capability that each enrichment player has to build in-house because they are tightly export- and treaty-controlled. The skilled workforce needed to run enrichment equipment safely is also thin after the US let its expertise atrophy following the Cold War build-down. Licensing adds more time on top, since every new cascade needs NRC approval and the NRC's own staff capacity limits how many applications it can move at once.
There are major efforts being pursued to build up Western enrichment capacity. The Department of Energy has committed up to $2.7B over ten years across six companies – Centrus, General Matter, Global Laser Enrichment, Urenco's Louisiana Energy Services, Laser Isotope Separation Technologies, and Orano Federal Services – aimed at rebuilding domestic enrichment. These efforts have a particular focus on the fuel nearly every next-generation small modular reactor design is built around (HALEU), and essentially the only commercial-scale production of it today is Russian. A domestic alternative will need to be seeded before SMR developers hit a wall, and that can’t happen without conversion.
Because enrichment takes so much longer and requires so much more capital to build (and more regulation to navigate), the rational sequence is bringing conversion capacity online first, and building a buffer of converted material that sits ready the moment new enrichment equipment switches on in the early 2030s. A cushion covering a year or two of incremental enrichment demand would run 15,000 to 25,000 tonnes of UF6, the material worth roughly $4B to $7B at current pricing.
Fuel Fabrication
The final stage before a reactor is fuel fabrication, where – and please bear with us on the technical terminology here – enriched UF6 is turned back into solid uranium dioxide, pressed into ceramic pellets, and loaded into zirconium alloy rods (aka fuel rods).
Unlike conversion and enrichment, this stage is not geographically concentrated, because fuel assemblies are engineered specifically for each reactor design and most nuclear countries run their own plants. Westinghouse and Framatome, Orano's former reactor fuel arm, run the major US plants; globally, Framatome operates in France, Rosatom's TVEL supplies Russian-design reactors, and Korea's KEPCO and China's CNNC serve their own fleets.
The risk here is lock-in, not capacity. A plant built on Russian VVER (water-water energetic reactor) technology, common across Eastern Europe, can only run on TVEL’s fuel unless an alternative design is requalified for that exact reactor, itself a multi-year engineering effort. It is a live issue in Europe today, and a good illustration of how dependence in this industry can come down to design and switching costs as much as raw capacity.
Generation
It is worth grounding the fuel chain argument in the demand, which is being driven by both government policy and the data center build-out, with hyperscalers leading the charge. Across the four hyperscalers shown below, there is 16.7 GW of nuclear demand alone. These buyers are willing to put dollars not just behind traditional nuclear infrastructure, but up-and-coming offerings like SMRs (ex: Valar Atomic’s recent 9.6 GW deal with a data center project in Utah). These buyers act as a catalyst for bringing these technologies to scale - particularly because they are price-insensitive- which gives SMRs a boost in becoming a meaningful contributor to the energy mix.

Among the planned capacity are restarts of currently dormant nuclear plants. Three decommissioned US reactors are mid-restart for the first time in the industry's history.
- Holtec's Palisades plant in Michigan, shut since 2022, moved from decommissioning to operational last year
- Constellation's Three Mile Island Unit 1 is restarting under a 20-year power purchase agreement with Microsoft, targeting as early as 2027
- NextEra's Duane Arnold in Iowa, backed by a $1.9B DOE loan, is targeting early 2029
None of this was commercially conceivable five years ago, when the industry's direction of travel was retirements, not restarts.
SMRs are also moving from slide decks to signed contracts. Google has ordered 500 megawatts from Kairos Power, Amazon has backed X-energy, and Meta signed agreements with Vistra, TerraPower, and Oklo in January covering up to 6.6 GW of nuclear capacity through 2035, largely to power its data centers. SMR designs in particular are being built around a fuel type (HALEU) that is almost entirely dependent on Russian supply today.
The common thread across restarts and SMRs alike is the same data center demand shock we alluded to. Every one of those megawatts, restarted or newly built, needs fuel that has to be mined, converted, enriched, and fabricated. In late 2025, one of the large power generation and backup power system providers explained to us the unsustainable fervor for their products and just how large backlogs are as a result of this shock. We’d add that this doesn’t account for the slow-to-unfold manufacturing revival across the country - this would add another 24 GW of peak energy demand over five years, according to S&P Global estimates.
Waste
Spent fuel has had little development for decades. More than 95,000 metric tons of spent fuel sit in storage at roughly 80 sites across 39 states, growing by about 2,000 tons a year, with no permanent US repository since the Yucca Mountain project stalled over a decade ago. However, in 2026 the DOE approved an initial, early-stage decision to build a federal consolidated interim storage facility, sized to start at roughly 15,000 tons with room to expand.
Beyond storage, a handful of companies are trying to turn spent fuel from a liability into feedstock:
- Oklo, already public, is building a $1.68B facility in Oak Ridge to turn used fuel into feedstock for its own fast reactors
- Curio is pursuing a different chemistry, extracting UF6, noble gases, and platinum-group metals from spent fuel, and filed its own NRC pre-application letter in May 2026
The case for this space is real: converging reactor technology, tens of billions of unmonetized material already sitting on utility and government balance sheets, and the first federal siting decision in a decade. The case against it is just as real: the longest, most politically fraught history in the entire fuel chain, and where the customer for any recycled product is a fast-reactor fleet that does not yet exist at scale. We are watching this closely, but the latter point in particular leads us to believe it might be too early to make a bet here.
Where We See Opportunity
The fuel chain behind nuclear power is an excellent example of the opportunities we seek to find: critical physical infrastructure that the market has underappreciated because it sits upstream of the more visible reactor story.
Every stage in this value chain matters, and mining, fabrication, and the eventual back end all deserve their own scrutiny (for more on mining, see our prior piece here). But if we had to point at one place where capital, timeline, and scarcity intersect most sharply right now, it is conversion. Secondary supply, the drawn-down stockpiles that quietly covered part of world demand for two decades, is thinning out, and a further tightening of Russian access to Western conversion markets would remove one of the few remaining safety valves that have been used in the face of heightened demand. While we favor conversion at present, we are monitoring developments in the enrichment space. We note that the regulatory controls make it hard to build centrifuge manufacturing capabilities; at some point, this bottleneck will come to bear and bring with it investment opportunities. Apart from that, both mirror the capex dynamics and labor stresses of a typical industrial facility. Overall, we feel investing around the value chain allows us to capitalize on the nuclear wave without betting on underlying generation tech or timing, as it remains relevant regardless, as a result of the mismatch of supply and demand.

