The fuel cycle · ten stages, one thread

From a hole in the ground to a reactor core, and back again.

Ten stages. One bottleneck. Each section below tells you what happens at that stage and where the material goes next - that handoff is the cycle.

Click any node in the diagram above to jump straight to its stage. The thread on the left tracks how long the journey takes - by the time you reach disposal, the elapsed time is geologic.

Stage 1 · front end

Mining

Uranium comes out of the ground three ways. Open-pit and underground mines blast and haul rock the way most people picture mining. The third method, in-situ recovery, pumps a mild acid or alkaline solution into a sandstone aquifer, dissolves the uranium in place, and pumps the loaded fluid back to the surface. Kazakhstan runs almost entirely on in-situ recovery, which is how a country with no historic mining tradition produces close to half the world's supply. Canada and Australia operate the highest-grade conventional mines, with Cigar Lake and Olympic Dam carrying ore grades above one percent uranium.

The mined output is rock or pregnant solution. Neither is something a reactor can use. The next stop is the mill, almost always built right next to the mine because moving rock is more expensive than moving concentrate.

→ ore (or loaded solution) goes to the on-site mill for first concentration.

Stage 2 · front end

Milling

Milling is what turns rock into something that ships. Crushed ore is dissolved in sulfuric acid or, for some deposits, an alkaline carbonate solution. The dissolved uranium is then separated from the spent rock with solvent extraction or ion exchange, precipitated out, and dried into yellowcake (U₃O₈), about 85 percent uranium oxide by weight. The name comes from the color of the early product; modern yellowcake is closer to brown or olive depending on calcination temperature.

Yellowcake leaves the mining country in steel drums. The uranium inside is still natural, with the same isotopic mix it had underground - about 0.7 percent fissile U-235, the rest U-238. Nothing has been enriched yet. The next plant in the chain is conversion.

→ drums of U₃O₈ ship to one of five commercial-export conversion plants.

Stage 3 · the choke point

Conversion // the narrowest pipe.

Conversion is the chemistry step that turns yellowcake into uranium hexafluoride, UF₆. Two reactions, both unglamorous: yellowcake is reduced with hydrogen to uranium dioxide, then fluorinated with hydrogen fluoride and elemental fluorine to produce UF₆. The product is a white crystalline solid at room temperature that sublimes directly to gas at 56°C. That low sublimation point is the entire reason this stage exists.

Centrifuge enrichment requires a feedstock that behaves as a gas at modest temperatures and that has only one element bonded to uranium. Fluorine has only one stable isotope, F-19, so any mass difference between two UF₆ molecules comes entirely from whether the uranium atom is U-235 or U-238. No other compound of uranium fits the requirement. Without conversion, enrichment does not happen. Without enrichment, light water reactors do not run.

Five plants run at commercial export scale: Cameco Port Hope (~11,200 tU/yr), Orano Tricastin (~11,000 ramping to 13,500), Solstice Metropolis (10,000+ committed for 2026), Westinghouse Springfields (restarting 2028 at 5,000 nameplate), and Rosatom Seversk (~12,500). Western 2025 capacity sits near 31,000 tU against ~44,000 tU of Western demand. Russia sells the difference. The 2024 Russian Uranium Imports Act lets US utilities keep buying through January 2028, after which the four Western plants have to cover the gap on their own.

→ UF₆ ships in steel cylinders to enrichment plants. Move the dials in the simulator to see what the wedge does under different assumptions.

Stage 4 · front end

Enrichment

Enrichment raises the share of fissile U-235 from natural 0.7 percent to what a reactor wants. Light water reactors run on low-enriched uranium (LEU), 3 to 5 percent. Advanced reactors and several SMR designs want high-assay LEU (HALEU), 5 to 20 percent. Centrifuges spin UF₆ gas at very high speed, the heavier U-238 drifts to the outer wall, the lighter U-235 concentrates near the center, and a cascade of thousands of machines in series gradually pulls the assay up. The work is measured in SWU; roughly 7 SWU produces 1 kg of 4.95 percent LEU.

Urenco (Germany, Netherlands, UK, US), Orano (Georges Besse II at Tricastin), Rosatom (Angarsk, Seversk, Novouralsk, Zelenogorsk), and CNNC are the major operators. Centrifuge technology is dual-use and tightly export-controlled, which is why HALEU supply is currently the second narrowest pipe in the cycle. Two streams come out: enriched UF₆ for the reactor side, and a much larger volume of depleted UF₆ tails.

→ enriched UF₆ goes to deconversion. Depleted UF₆ tails go to long-term storage (see Storage).

Stage 5 · front end

Deconversion

Conversion got the uranium into a gas the centrifuges could separate. Deconversion reverses that, turning the enriched UF₆ back into uranium dioxide (UO₂), the ceramic form a fuel pellet is made of. The gas is hydrolyzed to uranyl fluoride, then chemically reduced to UO₂ powder, with hydrogen fluoride recovered as a byproduct.

Most fuel fabricators do this step themselves on site, which is why deconversion is sometimes folded into fabrication on simpler diagrams. It is broken out here because the chemistry is distinct and because the same reverse step also handles depleted-UF₆ tails: Orano's W2 plant at Tricastin and the DOE-funded plants at Paducah and Portsmouth deconvert depleted UF₆ to U₃O₈ for stable long-term storage.

→ enriched UO₂ powder feeds straight into the pellet press at fabrication.

Stage 6 · front end

Fuel fabrication

Fabrication turns UO₂ powder into something a reactor can swallow. The powder is pressed into small ceramic pellets, sintered in a furnace, and ground to spec. Pellets are stacked into long zircaloy tubes, sealed, bundled into assemblies, and shipped to a reactor site. Each assembly holds several hundred kilograms of uranium and is designed to sit in a core for four to six years before it has to come out.

Westinghouse, Framatome, GNF, TVEL, and KEPCO Nuclear Fuel are the major fabricators. Assemblies are reactor-specific - a Westinghouse PWR bundle will not fit a CANDU. Switching fabricators is technically possible but slow because each new design has to be qualified by the regulator. Ukraine's switch from TVEL to Westinghouse for its VVER fleet took roughly a decade.

→ finished fuel assemblies ship by truck or rail to the reactor site.

Stage 7 · front end ends here

Power generation

Inside the reactor, a neutron strikes a U-235 nucleus, the nucleus splits into two roughly equal daughter products, two or three more neutrons come out, and about 200 MeV of energy are released as heat. That heat boils water, the steam spins a turbine, and the turbine spins a generator. The 438 commercial reactors operating worldwide as of April 2026 need roughly 69,000 tU of natural uranium equivalent per year to keep going.

Most of the fissioning that produces the power is U-235, but a meaningful share comes from plutonium-239 bred in place when U-238 absorbs a neutron. By the time an assembly is pulled out, several percent of its mass has been transmuted, the U-235 is mostly gone, and a complicated mix of fission products and transuranics has built up. The assembly is now spent fuel.

→ spent assemblies move into the on-site spent-fuel pool to cool.

Stage 8 · back end

Storage

Two material streams sit in storage. Spent fuel from reactors comes out thermally and radiotoxically hot, sits five years or more in a water-filled spent-fuel pool to bleed off decay heat, then loads into dry casks of steel and concrete on a pad at the reactor site. The IAEA estimates roughly 400,000 tonnes of heavy metal in spent fuel is stored worldwide, the large majority of it on reactor sites. No operating geological repository for commercial spent fuel exists anywhere in the world; Finland's Onkalo expects first emplacement in the late 2020s.

The other stream is depleted UF₆ tails from enrichment, deconverted to stable U₃O₈ and stored in drums on a pad. The US has roughly 700,000 tonnes of depleted-uranium inventory in long-term storage at Paducah and Portsmouth. Whether either pile is treated as waste or as future feedstock is a policy choice, not a technical one.

→ spent fuel either sits indefinitely or moves to reprocessing; high-level waste eventually goes to disposal.

Stage 9 · back end

Reprocessing

Reprocessing chemically separates the uranium and plutonium in spent fuel from the fission products and from each other. The standard process is PUREX (plutonium uranium reduction extraction), developed in the 1940s and refined since. The recovered plutonium can be blended with depleted uranium to make MOX (mixed oxide) fuel, which is burned in suitably licensed light water reactors. Recovered uranium can be re-enriched and reused.

France operates the largest commercial site at La Hague (1,700 tHM/yr nominal). Sellafield has wound down commercial reprocessing. Russia runs a smaller line at Mayak. Japan's Rokkasho plant is still in commissioning. The economics depend on the price of natural uranium, the cost of waste disposal, and political tolerance for separated plutonium. Reprocessing is the only stage that meaningfully closes the loop, which is why it keeps coming back into discussion even when the economics are quiet.

→ three streams come out: PuO₂ loops back to fabrication as MOX, recovered uranium loops back to conversion, and fission products go to disposal.

Stage 10 · back end

Disposal

Disposal is where the cycle finally stops. The high-level waste stream is the fission-product fraction left after reprocessing, plus any spent fuel a country has chosen to treat as waste rather than feedstock. The standard end state is vitrification: liquid waste is mixed with molten borosilicate glass, poured into stainless steel canisters, cooled, and parked in interim storage until a geological repository is ready. The glass locks the radionuclides into a chemically stable matrix that resists leaching for the timescales that matter.

No operating deep geological repository for commercial high-level waste exists yet. Finland's Onkalo at Olkiluoto is closest, with first emplacement expected in the late 2020s. Sweden's Forsmark site is licensed. The US Yucca Mountain project remains stalled. Until a repository opens, every country in the cycle is running an open-ended interim-storage program, which is the unresolved policy problem under everything on this page.

→ the loop closes, or it parks. The simulator at /simulator lets you change the assumptions yourself.

Ten stages and one bottleneck. Mining and milling have spare capacity, multiple suppliers, and reasonable substitutability. Reactors and fabrication are diverse enough that no single failure shuts the system down. Conversion is the one stage where five plants carry the world, where one of the five sits inside a country the West is sanctioning, and where a single regulatory deadline in January 2028 forces the rest of the system to absorb the gap. The simulator at /simulator lets you move the dials yourself. The full data audit and source list live at /methodology. If a number on this page does not match what you find there, the audit wins.

Where every stage sits.