Natural vs. Enriched: Canada's Fuel Cycle Reversal

For the entirety of Canada's nuclear power generation, Canada has opted for natural uranium rather than enriched uranium. Of the 25 power reactors Canada has ever built domestically, every one has run on natural, unenriched fuel. The United States, by contrast, has zero operational natural-uranium power reactors, with all of its roughly 96 operating units running on enriched uranium. Worldwide, about 90% of reactors run on enriched fuel. That is now changing, and the standard explanation for why Canada chose natural uranium in the first place has the causation backwards.

The reason usually given is that Canada is the world's second-largest uranium producer, exporting roughly 85% of what it mines, producing roughly 24% of global mined uranium in 2024, but has no domestic enrichment facilities. So the story goes: natural-uranium reactors need more uranium per unit of energy, but since Canada had uranium to spare and no enrichment plant of its own, it used more uranium instead of paying the United States to enrich it.

That explanation is backwards. Natural-uranium use was never the objective. It was the consequence of a reactor architecture that came out of US nuclear containment policy in the 1940s, and that architecture happened to buy Canada a capability that was never properly priced. In December 2021, Ontario Power Generation selected the BWRX-300 for the new Darlington project, an enriched-uranium design; construction was licensed in April 2025 and the province approved CAD 20.9bn for the project a month later. Other announced Canadian reactors are likely to follow. That decision is what is now being partially exited.

1The choice, 1942–1960s

Canada's choice to build for natural rather than enriched uranium is rooted in Second World War-era nuclear containment policy. In December 1942, the US restricted what it would share with Britain on heavy water, uranium separation, and plutonium chemistry, and the 1943 Quebec Agreement, signed by Churchill and Roosevelt, formalized US–UK–Canada nuclear cooperation on those terms: the United States refused to share enrichment and plutonium-extraction research with its closest allies, but did approve heavy-water reactor work, since it was thought a bomb couldn't be engineered from that design.

The Montreal Laboratory, the Anglo-Canadian-French wing of the Allied bomb project, was set up in 1942 under Canada's National Research Council and was already working with heavy water, an initial stock of which had been smuggled out of France ahead of the German advance in 1940, originally produced in Norway. A small heavy-water plant was built at Trail, British Columbia in 1943 under US sponsorship (the Manhattan Project's P-9 Project) and remained in operation until 1956. Its output was a potential input to the atomic bomb but was never used for one. It did, however, begin Canadian heavy-water production.

Canada was also the site of the first operating reactor outside the United States. ZEEP went critical on 5 September 1945 at Chalk River, designed for natural uranium specifically to bypass enrichment. NRX followed, then NRU. NRX became the basis for the US Savannah River production reactors. AECL's own history states that Canada chose the heavy-water power reactor after studying alternatives, because it made use of existing heavy-water research experience and enabled use of Canadian uranium without enrichment in foreign facilities, all of which at the time had been built primarily for weapons programs.

This is what led to the CANDU, Canada's indigenous reactor design, developed by the crown corporation Atomic Energy of Canada Limited, spun out of the wartime Montreal Lab and the Chalk River work from the 1950s onward. The CANDU is a product of Canadian nuclear political economy in the 1950s. It is most practical for a nation with uranium, heavy water and hydroelectricity but no enrichment plant or access to one — and for a nation without a heavy forging industry.

That forging capacity constraint matters more than it might first appear. A light water reactor generally needs a single large steel pressure vessel, forged in one piece. Canada did not have the facilities to make one, and there was no reliable global seller. CANDU bypasses the pressure vessel entirely and distributes the pressure boundary across hundreds of small pressure tubes, each individually manufacturable within the Canadian industrial base of the time. Heavy-water moderation and pressure-tube construction reinforce each other, and both were workarounds to constraints of the era. Neither constraint exists any more, but the architecture built around them remains.

CANDU is also the design Canada proliferated abroad. At the time this was considered relatively safe, since no state had yet built a weapon without enriched uranium. That did not hold. In 1974 India demonstrated a weapon in the Smiling Buddha test, built using plutonium separated from CIRUS, a Canadian-supplied heavy-water natural-uranium research reactor — not a CANDU power reactor, though the underlying vulnerability is the same one; Douglas Point was also the design basis for India's first CANDU-derived power reactors. Unenriched fuel carries a higher proportion of U-238, and U-238 is what breeds into plutonium. CANDU's on-power refuelling adds to this: it allows fuel to be discharged at low burnup, which is precisely the condition that produces weapons-grade plutonium.

2Why the United States went the other way

The United States shifted in the opposite direction, opting entirely for enriched uranium. It had already built enrichment facilities for the Manhattan Project, and while those facilities were extraordinarily energy-intensive, the enriched uranium could be supplied to utilities at reasonable prices. The US Navy also invested heavily in developing light water reactors for its submarine program. These reactors have obvious advantages for that use case: they are compact, and high enrichment enables long intervals between refuelling — initially every two to three years in the 1950s, and every thirty to forty years in modern submarines. The Navy paid for the initial development, and the commercial sector inherited it.

Both the Canadian and American positions are path-dependent on the situation immediately after the Manhattan Project. Early decisions were made on the basis of technology already built and left over, heavily influenced by what the United States needed militarily to remain the dominant nuclear power. Long-run efficiency comparisons between the two approaches were much less of a factor — understandably, since both plant types far exceeded the available alternatives for power generation anyway.

Neither choice was a clean technology judgement. Each country made a rational decision conditioned on an asset it already happened to own.

3The physics, and what it actually buys

Comparing natural and enriched reactors requires understanding neutron economy: the balance between neutron production from fission and neutron losses through absorption or leakage. A high-efficiency design has "leftover" neutrons available to sustain the chain reaction, breed new fuel, or burn up material that would otherwise be wasted.

Light water (H₂O) absorbs a non-trivial fraction of neutrons. To sustain the chain reaction you compensate by enriching the fuel to roughly 3–5% U-235 — enrichment exists as an industry because of this property. Heavy water (D₂O) absorbs very few neutrons, so natural uranium at 0.711% U-235 works without modification. "Natural uranium" is a consequence of high neutron economy, not the objective. That is the crux of the argument that follows.

4The fuelling cost advantage is real but smaller than the rhetoric

The consequence of high neutron economy is that a natural-uranium reactor avoids the expensive enrichment process entirely. Enrichment also consumes uranium, leaving a meaningful quantity of U-235 discarded in the depleted stream. The net result is that a CANDU requires roughly 25–40% less mined uranium per unit of energy than a light water reactor — estimates vary by source, but land consistently in that range. Fabrication costs are also low, with a simple bundle and high reliability.

What offsets this is heavy water. Creating and holding the initial heavy-water inventory is a steep upfront capital cost — commonly cited at roughly 20% of a CANDU station's capital cost, with total inventory running 460–592 tonnes. That front-loads a cost that would otherwise be spread across decades of fuel purchases in a light water reactor.

The mechanism working against the natural-uranium advantage is tails assay. Enrichment faces a trade-off between natural uranium feed and separative work (SWU), and enrichment plants choose an operating point depending on the price of uranium. If uranium is cheap relative to SWU, plants run a higher tails assay — throwing away more U-235 in the depleted stream but buying less separative work and consuming more feed. If uranium is expensive, they strip harder to conserve it. Enrichers can re-optimize with the same equipment as prices move. The upshot: the uranium savings from natural-uranium fuelling are worth less when uranium is cheap and more when it is expensive, and that value is not fixed over time.

On the back end, CANDU fuel discharges at roughly 7–8 MWd/kg against 35–50 for an LWR, so a CANDU produces something like five to seven times the spent-fuel mass per unit of electricity. The disposal penalty is smaller than that ratio suggests, though, because repository cost tracks heat load, and low-burnup fuel runs far cooler; the small, simple bundle also keeps the emplacement system cheap. It nets out as roughly neutral, not a clear loss or a clear win.

Then there is the question of how much any of this matters. Fuel is roughly 20% of the levelized cost of new nuclear electricity, with the uranium concentrate itself no more than about 5%. Conversion, enrichment and fabrication together make up roughly half of total fuel cost, which puts enrichment somewhere near 5–8% of LCOE. Not the sole deciding factor, but not trivial either.

The conclusion is straightforward: the case for natural uranium does not depend on avoided enrichment costs.

5What neutron economy actually buys

The same neutron economy that lets a CANDU run on unenriched fuel also lets it run on almost anything else.

Recovered uranium from reprocessed LWR fuel sits at roughly 0.9% U-235, above natural, and a CANDU can use it without any further processing, reaching burnups of 13 MWd/kg or better and extracting twice the energy an LWR gets from the same material — cutting front-end fuel costs by roughly a third against natural uranium. The TANDEM cycle, using unseparated uranium and plutonium at around 1.5% fissile material — an atomic substance that can sustain a self-driven chain reaction when struck by a low-energy neutron — reaches roughly 25 MWd/kg, about four times the energy of simply recycling plutonium back into an LWR. In general, a CANDU can extract up to twice the thermal energy from a given fissile inventory that an LWR can, depending on what that inventory is.

The key takeaway is that the two reactor types are complements rather than substitutes. A CANDU fleet is a natural back end for a light water fleet: what an LWR discards is still more energy-dense than what comes out of the ground.

Canada does not do this, largely because of a Carter-era 1977 ban on commercial reprocessing in the United States. Reagan lifted it in 1981, but the reprocessing capacity was never built. Canadian officials have discussed the possibility, but nothing has moved publicly since 2019. The prospect exists but is effectively dormant.

China has demonstrated it. Recovered uranium from its pressurized light water fleet has been used as CANDU fuel at Qinshan, blended with depleted uranium to natural-uranium equivalence, starting with the first NUE fuel bundles loaded in March 2010. It is the only demonstration of the LWR-to-PHWR synergy anywhere, and it has not been done at scale.

There is one more thing neutron economy buys that has nothing to do with fuel cycles. On-power refuelling and surplus neutrons let CANDU units produce medical and industrial isotopes as a side product of normal operation. Canadian CANDUs have supplied cobalt-60 since 1986, used to sterilize roughly 30% of the world's single-use medical devices, and Bruce Power began producing lutetium-177 for cancer therapy in a commercial power reactor for the first time anywhere. An enriched-fuel LWR fleet does not replace that capability — it is a genuinely separate national asset that happens to sit on the same physics.

6The reversal started inside the programme

It would be a simplification to describe the shift to enriched fuel as a series of procurement decisions made unit by unit, without any strategic reconsideration inside the program itself. That is not accurate. AECL abandoned natural uranium well before the pivot to enrichment became a customer-facing decision. By the late 1990s, CANDU 9 sales were in decline, and AECL moved to the Advanced CANDU Reactor program, commonly dated to around 1998. The resulting ACR-700 dropped natural uranium for 1–2% low-enriched fuel and dropped heavy-water coolant for light water, keeping heavy water only as moderator, with the stated goal of a smaller, cheaper core and a smaller heavy-water inventory for better economics against LWR competitors. The ACR program was wound down after AECL's reactor division was sold to SNC-Lavalin in 2011. The natural-uranium constraint was judged worth trading away roughly twenty-five years before the Darlington SMR was licensed. The switch to enriched fuel is not a case of procurement drifting away from a natural-only doctrine — the doctrine had already been abandoned inside AECL.

7The current situation

Canada's plan to supply the Darlington SMRs starts at Cameco's uranium mines in Saskatchewan. The uranium is converted at Port Hope, Ontario, then shipped to Urenco USA at Eunice, New Mexico for enrichment, then through US fabrication of assemblies, then back to Canada for use at Darlington.

Meanwhile the supply chain for the 17 operating CANDUs remains almost entirely domestic. Canada is the only country capable of supplying its whole reactor fleet from its own production. That is the thing being partially scaled back.

How much of the sovereignty aspect is lost depends on the size of the enriched fleet. A 2023 NRCan draft, obtained through Access to Information, listed 18 reactors requiring enriched fuel expected in Canada between 2028 and 2040. The technology mix of those 18 is not public, and can swing outcomes enormously. For example, eighteen 300 MW SMRs is a very different demand figure from a mix of reactors including AP1000s.

For scale, a useful anchor is that a 1,300 MW station consumes roughly 120,000 SWU/yr, and a 1M SWU/yr plant fuels about eight large stations. Four BWRX-300s (1,200 MW combined) come to roughly 110,000–140,000 SWU/yr; the BWRX-300 runs higher enrichment than the anchor case, so the top of that range is the right read. Eunice produces about 4.3M SWU/yr across 64 cascades, coming to approximately 67,000 SWU per cascade, with newer cascades nearer 87,000. So the four Darlington units together amount to something like one and a half to two cascades, or about 3% of a single American plant's output.

Important to note is that the entire first tranche of Canada's enriched fleet is a rounding error inside one facility in New Mexico.

8What is actually being foreclosed

Canada is now doing three things at once.

It is building an enriched-fuel fleet. It is committed to a deep geological repository at Ignace, selected in November 2024, a $26bn project meant to hold Canada's used fuel for the next 160 years. And for the time being it still owns the only commercial reactor architecture in the world that can burn what light water reactors discard, without re-enrichment and without separating plutonium.

These three commitments are in tension. The configuration with the CANDU fleet functioning as a back end for the enriched fleet, and in which the repository takes materially less material because a meaningful fraction of it has been through a CANDU first, is not being chosen against. It is simply not being sufficiently considered. Each individual procurement is defensible. However, the aggregate quietly forecloses an option that no other country has.

9Limitations

1. The overall 20%/5%/50% fuel-cost splits are WNA figures. The 5–8%-of-LCOE estimate for enrichment specifically is my own derivation from that split, not a disclosed number — treat it as an estimate.
2. The 18-reactor NRCan figure comes from a 2023 draft obtained under Access to Information, and CBC's later reporting is the only public account of it. As many as 10 of those 18 reactor vendors have reportedly since run into trouble, so the figure is likely a ceiling rather than a current plan, and the technology mix behind it was never public in the first place.
3. Recovered-uranium and TANDEM-cycle synergies are documented in AECL/IAEA analysis and lattice-cell calculation, and demonstrated at Qinshan in the recovered-uranium case, but not at commercial scale in a Canadian context.
4. The proliferation critique is acknowledged and sketched here but not fully developed.