The Case for an American CANDU Nuclear Reactor

Introduction

It’s quite paradoxical that the most capable technologies in world-history, AI, nuclear energy, and quantum computing revived a three-centuries old policy response: mercantilism. Today, national self-interest has taken precedence over global cooperation. China and the US are in a technological arms race where narrow inputs are considered strategic assets.

The Commerce Department, for example, has proposed $1 billion in support for an IBM quantum-foundry subsidiary and took a 10% equity stake in Intel, while the administration has directed the Nuclear Regulatory Commission (NRC) to accelerate nuclear reactor licensing.

Today, the United States may have a chance to address several of these narrow inputs at once. One reactor design now entering NRC review would generate power for data centers while producing a byproduct that both quantum computing and nuclear fusion require.

Though its approval is far from certain, the CANDU nuclear reactor is an opportunity for America to secure strategic bottlenecks before they become catastrophic chokepoints. America should therefore test CANDU because it provides a second nuclear-fuel pathway while potentially converting an existing heavy-water stockpile into electricity and strategic isotope capacity.

What is CANDU

CANDU is a heavy-water nuclear reactor, using natural uranium as fuel and heavy water as a coolant and moderator. It is fundamentally different from every commercial reactor operating in the US which all use enriched uranium and light water. Thirty-four CANDU units have been built globally, accumulating nearly 1,000 reactor-years of experience, across Canada, Argentina, China, and South Korea. (AtkinsRéalis; Department of Energy)

Since CANDU is a heavy water reactor, as it generates electricity, it also produces a strategic byproduct: tritium. Neutron absorption creates tritium in the reactor's heavy water. Tritium is the principal fuel pursued by deuterium-tritium fusion developers.

Stored tritium also decays into helium-3. Helium-3 has no known substitute in dilution refrigeration, the cooling technology used by superconducting quantum computers. (Ontario Power Generation; Department of Energy; NIST)

An American CANDU program paired with extraction infrastructure could therefore produce electricity for AI and a civilian supply of tightly controlled, supply-constrained strategic isotopes that would mitigate a potential supply catastrophe.

CANDU’s Third Arrival

Global data center electricity demand grew 17% in 2025, with AI-focused facilities driving most of that demand. (IEA) Nuclear power emerged as a mechanism to meet this surge in demand.

The US federal government catalyzed that emergence. Executive Order 14300 directed the NRC to establish an eighteen-month deadline for final decisions on new-reactor applications. The ADVANCE Act reduced licensing costs and directs the agency to make reviews more efficient, while Part 53 establishes a technology-inclusive framework for new designs. (White House; Nuclear Regulatory Commission; Federal Register)

On July 22nd, the Department of Energy requested industry input for the reuse of up to 530,000 gallons of heavy water currently stored at the Savannah River Site in South Carolina. CANDU would draw on that stockpile directly, requiring roughly 470 metric tonnes (about 112,000 U.S. gallons). Once loaded, the heavy water is continuously recycled rather than consumed as fuel. Despite some operating losses and replacement supplies, the Savannah River inventory could theoretically support the initial reactor load for the foreseeable future. (AtkinsRéalis, DOE)

On August 4th, Aternium announced its site for the first US commercial heavy water and electrolytic hydrogen production facility. The announcement represents the beginnings of a potential domestic supply chain, although construction and production have not yet begun. (Aternium)

The strategic value assigned to nuclear energy is the strongest in years. The regulatory environment is more efficient. America is starting to domestically produce heavy water. The infrastructure for CANDU is being built.

Conclusion

AtkinsRéalis advertises a total project time of 69 months. That estimate does not include the multi-year American licensing process, which is only beginning. Additionally, no authoritative U.S. cost estimate exists, and first-of-a-kind licensing, supply-chain development, and financing would likely make the initial reactor significantly more expensive than an established design.

Even if CANDU is approved, the first reactor is unlikely to operate before the mid-2030s. It cannot single-handedly solve AI’s immediate demand for power, nor should it become the first step toward replacing the American light-water fleet.

America should instead test one CANDU because it would provide a second nuclear-fuel pathway, put an existing heavy-water inventory to strategic use, add firm power during the 2030s, and create an option for domestic tritium and helium-3 production.

CANDU is a 2030s hedge against today’s bottlenecks becoming tomorrow’s supply crisis.