How Much Uranium Does a Reactor Use? First Cores, Reloads and Annual Demand
60-second answer: A representative 1,000 MWe light-water reactor consumes roughly 20–25 tonnes of enriched uranium in reloads each year, which takes about 190–210 tonnes of natural uranium (around 0.5 million pounds of U₃O₈) to make (WNA, fuel cycle overview). But that's an answer with assumptions inside it: capacity factor, thermal efficiency, burnup and the enrichment tails assay all move the number, and the ~75-tonne first core a new reactor loads before it ever sells power is a separate purchase on top. This guide works the calculation end to end so you can reproduce it, and explains how to read shorthand like "170 tU per GWe-year" in our own demand model.
There is no single universal answer to "how much uranium does a reactor use," and any figure quoted without its assumptions is a screening estimate, not a specification. What there is instead is a short chain of physics you can follow from electricity back to mined uranium.
Start with the units
Three unit systems show up in every fuel conversation, and mixed-up units produce most of the wrong answers:
- MWe vs MWth: a reactor's electrical output (MWe) is roughly a third of its thermal output (MWth) for a light-water plant. Fuel burns against thermal energy; revenue is electrical. Never apply a per-GWe coefficient to a thermal rating.
- Enriched vs natural uranium: the fuel loaded into the core is enriched product; making one tonne of it consumes several tonnes of natural uranium feed, with the ratio set by the assays. The two must never be added together.
- tU vs lb U₃O₈: mined uranium trades in pounds of U₃O₈; one tonne of contained uranium is about 2,600 lb of U₃O₈. Our units converter handles the arithmetic.
The worked calculation, end to end
Take a 1,000 MWe reactor running at a 90% capacity factor with 33% thermal efficiency and a burnup of 45 GWd per tonne of enriched uranium. These are stated assumptions, not measurements of any particular plant.
Step 1: electricity to thermal energy. 1,000 MWe × 365 days × 90% = 7.88 TWh of electricity a year. At 33% efficiency that requires about 23.9 TWh of thermal energy, or 995,000 MWd (megawatt-days, the unit burnup is quoted in).
Step 2: thermal energy to enriched fuel. At 45 GWd/tU, the reactor consumes 995,000 ÷ 45,000 = 22.1 tonnes of enriched uranium per year. As one formula: annual tU of product = MWe × 365 × CF ÷ (1,000 × efficiency × burnup in GWd/tU).
Step 3: enriched fuel to natural uranium. At 4.5% product assay, 0.711% natural feed and a 0.22% tails assay, each kilogram of product takes (4.5 − 0.22) ÷ (0.711 − 0.22) = 8.72 kg of feed. So 22.1 tU of product needs about 193 tonnes of natural uranium, which is roughly 0.50 million pounds of U₃O₈ at ~2,600 lb per tU. Change the tails assay and the feed number moves; the trade-off is quantified in our SWU guide.
Cross-checking against the industry reference: WNA's material balance for a similar reactor (4.5% enrichment, 0.22% tails, 45 GWd/t burnup) runs 211 tonnes of natural uranium to 24.3 tonnes of enriched product a year, the same arithmetic at a slightly higher output assumption (WNA). Higher-burnup modern cores need less: WNA's 5%-enriched, 65 GWd/t example fabricates only 16.9 tU of product a year.
| Assumption | This example | What moving it does |
|---|---|---|
| Capacity factor | 90% | Fuel use scales almost linearly with it |
| Thermal efficiency | 33% | Higher efficiency, less fuel per kWh |
| Burnup | 45 GWd/tU | Higher burnup, fewer tonnes loaded (at higher enrichment) |
| Product assay | 4.5% | Set by fuel design, with burnup |
| Tails assay | 0.22% | Lower tails, less feed but more SWU |
A first core is not a reload
Before a reactor generates anything, its core must be filled. For a 1,000 MWe light-water reactor that first load is about 75 tonnes of enriched uranium (WNA), roughly three times an annual reload, purchased and fabricated before commercial operation. Thereafter only a third or so of the core is replaced at each refueling, every 12 to 24 months.
Two consequences for demand analysis:
- First cores are one-time inventory demand, and they land before grid dates. A wave of new reactors moves uranium procurement years ahead of the electricity statistics.
- Do not model a first core as "three years of reloads." It is a disclosed design quantity at specific loading assays; the one-third-per-refueling rhythm afterwards is what the annual figures describe. Multiplying an annual coefficient by three substitutes a guess for a number the developer has actually disclosed.
When utilities buy vs when reactors consume
Physical consumption is smooth; procurement is lumpy. Utilities contract years ahead, take deliveries on contract schedules, hold inventory, and send material through conversion, enrichment and fabrication with lead times at each stage. The US survey data makes the distinction visible: EIA's Uranium Marketing Annual reports deliveries, enrichment-feed deliveries and forward coverage separately, and its 2025 survey shows deliveries and anticipated requirements year by year through 2035 (EIA UMA). Our coverage tracker charts that forward curve; the difference between contracted deliveries and anticipated requirements is procurement, not reactor physics.
A long refueling interval, as some advanced designs target, changes the timing of purchases (bigger, rarer batches, more up-front inventory) more than it changes cumulative fuel per unit of electricity, which stays governed by thermal efficiency and burnup.
Why SMRs, HALEU and recycling change the calculation
Everything above assumes a conventional light-water reactor. Three things break the template:
- Design-specific fuel: small modular and advanced designs differ in efficiency, burnup, enrichment and fuel form. A per-GWe coefficient calibrated on the light-water fleet is only a screening number for them; a named design needs its own specification. Our reactor tracker follows the pipeline.
- HALEU assays: designs using 5–20% enrichment consume much more separative work and feed per kilogram of product; supply is its own story, covered in our HALEU supply guide.
- Heavy-water reactors and recycling: CANDU-type reactors run on natural uranium with no enrichment step at all (WNA), and recycled MOX or reprocessed uranium displaces some fresh feed where it is used, mostly in Europe. See secondary supply.
How to read Yellowcake's demand model
Our site-wide screening coefficient is 170 tU per GWe-year of natural uranium, converted at 0.0026 Mlb U₃O₈ per tU (about 0.44 Mlb per GWe-year). That constant already embeds a fleet-average capacity factor of roughly 85%, so it is applied to nameplate capacity directly. Two rules for using it honestly:
- Never multiply it by a capacity factor again. The utilization is inside the coefficient; applying CF twice understates demand.
- Treat it as a screening assumption, not a forecast. It sits deliberately below the worked example above (which used a 90% CF and specific assays) and is labeled MODELED wherever it appears, for example in the nuclear-for-AI deal book. For any named project, the design's own fuel specification beats the coefficient.
The methodology page documents the model; when the assumptions change, the figures across the site change with them.
Frequently asked questions
How much uranium does a nuclear reactor use per year? A representative 1,000 MWe light-water reactor uses about 20–25 tonnes of enriched uranium in reloads annually, requiring roughly 190–210 tonnes of natural uranium (about 0.5 Mlb of U₃O₈) depending on capacity factor, burnup, enrichment level and tails assay. Always check the assumptions behind any quoted figure.
How much uranium is in a first core? About 75 tonnes of enriched uranium for a 1,000 MWe light-water reactor, per WNA's reference figures. It is a one-time inventory purchase made before commercial operation and should not be modeled as a multiple of annual reloads.
How much uranium per gigawatt? Yellowcake's screening model uses 170 tU of natural uranium per GWe-year (about 0.44 Mlb U₃O₈), with fleet-average utilization already embedded. Physics-based estimates at specific assumptions run higher or lower; the coefficient is for screening announced capacity, not forecasting a named plant.
Do reactors buy uranium the year they burn it? No. Utilities contract years ahead, hold inventories, and push material through conversion, enrichment and fabrication with long lead times. EIA's annual survey separates deliveries from requirements precisely because procurement and consumption are different time series.
Do SMRs use more or less uranium than large reactors? Per unit of electricity it depends entirely on the design's efficiency, burnup and fuel form; there is no general answer. Light-water SMRs resemble the existing fleet, while HALEU-fueled designs consume more natural uranium and separative work per kilogram of product but load fuel less often.
This article is for informational purposes only, not investment advice.