The Reactor That Makes Electricity First

The most important company in radiopharma's supply chain may not be a radiopharma company at all. It might sell electricity for a living. And the developers building pipelines on top of it will need it far more than it needs them.

That sentence sounds like a provocation. It is closer to an accounting fact. To see why, you have to start with the machines the entire field is standing on, and how old they are.

The base everyone builds on

Three of the workhorse isotopes in nuclear medicine come out of the same place. Lutetium-177 powers Novartis's Pluvicto and Lutathera. Yttrium-90 sits inside the radioembolisation microspheres used against liver tumours. And molybdenum-99, the parent of technetium-99m, underpins roughly 40 million diagnostic procedures a year, somewhere near 80% of all nuclear medicine scans.

All three trace back to a small set of research reactors. Most of them are old.

HFR Petten in the Netherlands and BR2 in Belgium both reached criticality in 1961. SAFARI-1 in South Africa and HFIR at Oak Ridge both date to 1965. These are the machines that were never built for medical supply. They were physics-experiment reactors that picked up isotope production as a byproduct, and sixty years later the byproduct became the mission while the infrastructure stayed exactly where it was.

The constraint that follows is not subtle. A research reactor runs roughly 210 to 260 days a year. The rest of the time it is down for maintenance, refuelling, or unplanned repair. The supply chain has no buffer, because the products decay. Mo-99 has a 66-hour half-life. There is no stockpile, no strategic reserve, and no substitute for most of what Tc-99m does in a hospital.

In late 2024 the sector got a clean demonstration of what that means. A routine inspection at HFR Petten found a pipe deformation above the reactor vessel, forcing an immediate shutdown rather than a planned restart. The timing was unlucky. It overlapped with maintenance at other reactors, and the system had no slack to absorb it. Supply of Mo-99 fell by 50 to 100% across parts of Europe. A generator manufacturer formally notified the NHS that it could not supply for specific dates. The outage lasted around three weeks and touched millions of patients.

The episode was not an aberration. It was a preview. HFR Petten produces close to a quarter of globally available Mo-99 capacity, it is 65 years old, and its replacement, the PALLAS reactor, only broke ground at the Petten site in September 2025. PALLAS's own guidance places commercial operation at roughly 2030 to 2032. France's Jules Horowitz Reactor, originally planned for far earlier, is now projected for around 2032 as well.

That gap, between when the 1960s reactors become unreliable and when their replacements actually arrive, is the single largest systemic risk in the field. And it is the gap a different kind of reactor is now starting to fill.

Why half-life let these isotopes centralise in the first place

Before the power-reactor route makes sense, one piece of physics has to be on the table.

Lu-177 has a half-life of 6.7 days. Y-90 sits at 64 hours. Mo-99 at 66 hours. These are long enough to permit centralised production at a small number of facilities followed by overnight or multi-day freight to the clinic. That is precisely why Novartis could build Pluvicto's entire supply chain around a handful of sites and ship globally.

This matters because it means the reactor making these isotopes does not have to sit next to the patient. It can sit anywhere with the right neutron flux and a reliable operating schedule. The half-life tolerates distance. What it does not tolerate is downtime.

Which is the exact weakness of a 1960s research reactor, and the exact strength of a commercial power station.

What changed in Ontario

In May 2025 the Canadian Nuclear Safety Commission amended the operating licence of Ontario Power Generation's Darlington plant to authorise production of Lu-177 and Y-90. This was on top of the Mo-99 that Darlington already produces, having become, in 2024, the world's first commercial power reactor licensed to make it.

Darlington is not a research reactor. It is a baseload nuclear power station whose primary job is to put electricity onto the Ontario grid. Its CANDU heavy-water design has a property that the light-water research reactors do not: it can refuel and insert or retrieve isotope targets while running at full power. It does not stop to make medicine.

The practical consequence is an availability figure that research reactors cannot reach. A research reactor offline a third of the year cannot compete on uptime with a unit engineered to generate baseload power continuously for decades. The harvesting system Darlington uses, developed with BWXT Medical, was designed to pull isotopes in a fraction of the time of competing approaches.

Stated full-scale targets for the site are large. Company guidance has pointed to a potential capacity measured in the millions of Lu-177 doses a year, which would make it among the largest single Lu-177 production sites anywhere. Those are targets, not current output. Commercial Y-90 is expected around 2026 and commercial Lu-177 around 2027. The figures should be read as the trajectory, not today's reality.

The model is also not confined to Canada. China's Qinshan plant, another heavy-water reactor, launched commercial reactor isotope production in late 2024 and produced its first Y-90 glass microspheres in 2025. Bruce Power, also in Ontario, is installing a second isotope production system specifically to hold Lu-177 supply steady through a major component replacement outage on another unit later this decade. That last detail matters: it shows operators already engineering redundancy across units, the way a utility plans, rather than hoping a single ageing reactor holds.

Why this is structurally different, not just more capacity

It would be easy to file this under "more supply is coming," alongside every other facility announcement. That misreads it.

The research-reactor model is a system in managed decline. Government-funded machines built for science, repurposed for medicine, operated part-time, retiring faster than they are replaced. The power-reactor model is something else: a utility-scale industrial commodity produced inside an asset designed for continuous, decades-long operation, not subject to the erratic maintenance cycles of a sixty-year-old research facility, and not competing with physics experiments for beam time.

The report this analysis draws from put it bluntly. A commercial power reactor producing the world's leading therapeutic beta emitter and its leading radioembolisation isotope, alongside the diagnostic backbone, is not a research facility with a medical sideline. It is a medical isotope platform that also happens to produce electricity. That inversion of priority is the structural shift.

But the same inversion that delivers the reliability is exactly what should make a developer cautious. Because it changes who holds the power in the relationship.

Who actually holds the leverage

A research reactor needs its isotope customers. The contracts help justify the funding, the staff, and the continued existence of the facility. The customer has standing.

A power utility does not need radiopharma. The numbers make this obvious. A large baseload nuclear station generates revenue from selling electricity measured in the billions. Isotope production, however valuable to the medical world, is a rounding error against that P&L. It is a high-margin and reputationally useful line of business, not a load-bearing one.

That asymmetry rewrites the negotiation. The utility runs to a grid schedule and to a nuclear regulator, not to anyone's clinical timeline. Its maintenance and refurbishment calendar is set by the economics of power generation and the demands of nuclear safety oversight, both of which dwarf medical isotope revenue. When priorities conflict, the megawatts win. And the utility, not the developer, decides whose offtake gets served first.

None of this means the established research-reactor supply disappears. It does not. The high-flux capacity at facilities serving the existing market remains essential and will for years. The point is narrower and sharper: the most reliable new capacity in the system is being built inside an asset whose owner's incentives are not aligned with the people who depend on it.

So a developer used to managing vendors finds itself in an unfamiliar seat. It becomes the junior counterparty to a counterparty that can walk away from the entire product line without noticing the dent.

The hire nobody is mapping

There is a second-order consequence that sits squarely in the operational reality of making this work, and it is the part the market has not priced.

Producing GMP-grade pharmaceutical material inside a nuclear power utility forces two cultures into the same building that were never designed to coexist.

A baseload nuclear plant runs on a heavy-industrial, safety-first, grid-compliant operating culture, governed by a nuclear regulator. Pharmaceutical production runs on a quality-assurance culture built around batch release, tech transfer, inspection readiness, and a clinical clock, governed by a medicines regulator. The first answers to the people who keep a reactor safe and the lights on. The second answers to the people who release a product for injection into a patient.

Marrying them needs a leader who can hold both at once. Someone fluent in reactor operations and in GMP quality, credible to a nuclear safety inspector and to a medicines inspector in the same week.

That crossover barely exists. It is not produced by either training pipeline, because the two pipelines have never had a reason to meet. The radiopharma talent shortage is already well documented: the radiochemists, the Qualified Persons who can release a decaying dose, the Authorized Users a clinical site cannot operate without. This is a rarer animal still. And whether a power-reactor isotope programme actually delivers releasable product, rather than simply irradiated material, will depend more on finding that person than on the reactor itself.

What it means, by seat

The shift reads differently depending on where you sit.

For drug developers, the supply agreement stops being a procurement footnote and becomes part of the asset. A clinical programme without secured, reliable isotope supply carries an execution risk that no amount of clinical data resolves. The developers who lock multi-year offtake from a continuously operating power reactor, two to three years ahead of their commercial timelines, will carry a reliability advantage that competitors leaning on part-time research reactors cannot match.

For investors, this is a diligence variable that linear models miss. Two otherwise identical Lu-177 assets are not equal if one has anchored supply on utility-grade capacity and the other has not. The supply chain belongs in the valuation, not in the appendix.

For utilities and the operators standing up these programmes, the binding constraint is not the reactor or the licence. It is the leadership and quality function capable of running regulated pharmaceutical production inside a power plant, and that function has to be built before the first commercial batch, not after.

For the wider CDMO and infrastructure layer, it confirms the broader pattern the sector has been slow to accept. The molecule is licensable. The isotope is, increasingly, available. The reliable, continuous means of production is the part that cannot be acquired on a 24-month timetable, and it is migrating toward an industry most radiopharma boards have never had to negotiate with.

The four-year read

Pulling the threads forward, here is the structural call, with the dates that make it falsifiable.

Through 2026 and 2027, the first commercial Y-90 and Lu-177 come off Darlington, moving the power-reactor model from proof of concept to industrial fact. Over the same window, the serious multi-year offtake agreements for these isotopes begin shifting toward power utilities rather than research reactors. The first developer to anchor its supply on a CANDU agreement will carry lower supply risk than peers still tied to Petten-class reactors, and over time will be priced for it.

Other CANDU operators follow, in South Korea, Romania, Argentina, India, and China, and the question of whether other reactor designs can bolt on isotope production starts to get asked seriously. Meanwhile the old base ages out. The PALLAS handover from HFR Petten remains the most consequential single event in the supply chain, and every month the 1960s reactors run past their planned retirement raises the odds of another 2024-style outage, potentially longer and with less recovery capacity behind it.

The end state is an inversion most supply models have not absorbed. The reliability layer of radiopharma, for the reactor-made isotopes at least, ends up owned by companies that think of themselves as electricity generators.

What this does not cover, and where it could be wrong

A few honest boundaries, because the thesis is strong only where it is true.

This applies to reactor-produced isotopes with half-lives long enough to ship: Lu-177, Y-90, Mo-99 and their relatives. It does not apply to the short-lived alpha emitters and point-of-care isotopes, whose physics forces an entirely different, decentralised supply architecture. Those follow their own rules and are not part of this story.

Power reactors are not invulnerable either. Commercial isotope production at this scale, inside a power station, is still early. Units do undergo long refurbishment outages, which is exactly why operators are already building cross-unit redundancy. And the regulatory pathway for routine pharmaceutical-grade production inside a power utility is still maturing rather than settled. The research-reactor base, for its part, is not going anywhere soon, and the established high-flux suppliers remain central to the market today.

None of that breaks the argument. It sharpens it. The direction of travel is clear even if the timeline slips: reliability for the workhorse isotopes is moving toward an asset class that radiopharma neither owns nor controls.

The takeaway

In a supply chain where every isotope decays on a fixed schedule, the scarce input was never capital. It was reliable time, and the people who can produce it under two sets of rules at once.

The operators who come through this will be the ones who locked supply early, who treated the agreement as part of the asset rather than a logistics line, and who started looking for the leader who can run the thing long before they needed one. The rest will keep planning around machines older than the moon landing, and calling it a strategy.

If you are stress-testing your isotope supply position, or working out who could actually run a programme like this, that is the work we do. The full facility-level picture sits in our Isotope Production Map.

ProGen Search runs retained executive search and market intelligence across radiopharma, CDMO, CRO and advanced modalities. This article is provided for information only and is not investment, legal or regulatory advice. Facility timelines, capacities and regulatory milestones reflect the state of play as of Q2 2026 and may have changed since publication.