As electricity systems accumulate ever higher shares of variable renewables, the structural need for dispatchable capacity becomes more pronounced. The difficulty, argues Nicolas Breyton, CEO of reactor company Stellaria, is that the fixed cost of maintaining a dispatchable fleet large enough to cover prolonged periods without sun or wind now dominates system economics. “The fixed cost is 80% of the price of the power,” he tells NEi, noting that utilities must size firm capacity for worst-case conditions, regardless of how often those conditions occur. Industrial users increasingly find themselves unable to secure around-the-clock power at predictable prices, pushing some operators toward self-sufficient energy solutions. For Stellaria these conditions frame specific technical requirements: a source of on-demand power that is local, high-temperature, dispatchable, and capable of long-term autonomous operation with minimal fuel logistics. The answer, Breyton concludes, is a fast spectrum reactor. He explains: “The only way to do that is renewable fission,” he say, noting that by ‘renewable’ he means a closed fuel cycle in which fertile isotopes – primarily uranium 238 – are continuously converted to fissile plutonium 239 within the molten-salt core of a fast reactor. Breyton explains: “Molten salt reactors are very interesting because they can close the cycle and can also burn minor actinides which are the most difficult long life high activity waste that one has.”
While the idea is not new – fast reactors have been studied and built for decades – Breyton points out that past programmes were hindered not by physics but by engineering economics. “We were lacking consistency,” he said of historical fast reactor efforts. “The main reason is that commercially it’s not viable.” Sodium- and lead-cooled reactors rely on solid pins within their fuel assemblies, but the inherent heterogeneity of solid fuel leads to problems in fast-spectrum conditions. “They accumulate plutonium outside the solid bar and they destroy fissile atoms inside,” Breyton explains. After roughly three years, assemblies must be removed and reprocessed. “Reprocessing takes 15 years, and it’s quite expensive, above the price of the reactor,” he says. While fast reactors offer an appealing waste-management capability – “they can reprocess almost indefinitely the spent fuel from existing reactors” – the combination of high core costs and long, complex reprocessing chains “has not been commercially viable so far.”
Breyton’s argument is that the limiting factor is not fast-spectrum physics but the choice of fuel form. A liquid-fuel system, he says, allows homogeneous fuel composition, continuous in-core breeding, and long residence times without the degradation constraints of solid pins.
Designing a fast-spectrum molten salt reactor
The core of Stellaria’s design is a core containing a mixture of fertile and fissile chlorides – primarily UCl₃ and PuCl₃ – dissolved in a purified sodium-chloride-based carrier salt.
The design requires strict control of isotopic composition. For example, natural NaCl contains both Cl 35 and Cl 37 and in a high-flux environment, Cl 35 captures neutrons to produce Cl 36, a long-lived radiological hazard. “Our reactor could work with chlorine 35 but not well,” Breyton says. To avoid the generation of Cl 36, the fuel salt would use “only chlorine 37,” necessitating isotopic separation at industrial scale. This requirement is non-trivial, as Cl 37 separation is rare in current industrial supply chains.
The purified NaCl must also be free of moisture and oxygen to prevent unwanted chemical reactions and to improve corrosion performance. The fuel mixture includes further unspecified additives which Breyton describes as “our secret sauce”.
Breyton emphasises the parallel development needed between reactor hardware and fuel-salt manufacturing. “We are a disruptive innovation; we need to have both the track and the road on time,” he says, adding: “The track is the reactor, and the road is the fuel.” Coordination with Orano, which Breyton says possesses the only commercial-scale facility capable of producing chloride fuel precursors, is viewed as essential.
As with the classic fast-spectrum breeder configuration as the plutonium decays one neutron induces fission in another plutonium nucleus, maintaining the chain reaction. The second neutron is absorbed by uranium 238 in the salt, converting it to plutonium 239. “Because the fuel is fully mixed”, says Breyton, “plutonium 239 is replenished in situ and in a spatially uniform manner”.
That homogeneity is the distinguishing feature and enables the fissile inventory to remain nearly constant across decades of operation. In this system, operators add small quantities of U 238 over time to maintain the fertile-to-fissile ratio as fission products accumulate. This results in what Brayton describes as ‘isoreactivity’. “We always have the same amount of fissile atoms from the beginning to the end,” he said. Because the fissile population is held nearly constant by on-line breeding, there is no initial excess reactivity and no depletion reactivity swing. In principle, the reactor could maintain a stable power output for decades without the extensive fuel assembly management typical of light-water reactors. The design thus allows for long-term operation without fuel replacement until fission-product poisoning limits reactivity, potentially as long as 40 or 50 years.
Passive safety design features
Within the Stellaria design the heat-transport system relies on natural convection within the core vessel. As the salt heats up, its density falls and it rises through the fluid column. It then flows through integral primary heat exchangers, is cooled and then descends along the outer annulus of the fission chamber before re-entering the central region. This establishes a buoyancy-driven circulation loop without pumps, reducing mechanical complexity.
One element of this design with particular significance for passive safety is the thermal expansion coefficient of the molten salt carrier. Breyton stresses the importance of the salt’s finely balanced physical properties as a mechanism in ensuring both passive operations and safety. “Dilation of the salt is very interesting and important” he says, adding: “If you add 20°C, the density will be slightly lower and it’s enough to stop the fission.” Acting as an intrinsic negative reactivity coefficient, Breyton describes the response time as occurring “at the speed of sound,” as the density changes propagate within the fluid. A gas-filled zone on top of this liquid salt allows it to easily dilate if needed.
Breyton contrasts this with the behaviour of solid-fuel reactors, which can experience positive reactivity transients in certain fast-spectrum regimes. In a homogeneous liquid, no internal power peaking arises from cracked or relocated fuel, nor is there a risk of cladding-driven reactivity changes. “The core is already melted,” Breyton says, adding: “so you cannot have a meltdown”.
Nonetheless, the design does incorporate rapid salt-drain capability to distribute the fuel and salt mix into subcritical storage tanks. “In case of an accident, it’s
very easy to remove the fuel from the fission chamber and put it in fifteen reservoirs that are, of course, subcritical,” Breyton says.
The drain down tanks provide a subcritical configuration with large surface area for passive decay-heat rejection and in high-threat scenarios, Breyton claims operators could drain the core within fifteen minutes.
In addition, mechanically actuated shutdown rods are retained in the design, although not for fine reactivity control. “You don’t need control rods,” he says, describing them instead as “start and stop rods”. Their purpose is to secure a safe criticality approach during initial startup or to hold the system in shutdown while maintaining fuel in the vessel. If operators require a more conservative safety posture, for example, in the case of extreme external events, the drain-down function provides a final passive shutdown state.
The reactor is designed around multiple negative feedback processes, including the natural convection within the primary circuit, the vessel radiation, and the cooling of lead reflectors by a network of water inner pipes from passive pools. To support the sustained breeding of Pu 239 “we have 50-70 cm of lead around the reactor reflecting neutrons,” Breyton says.
The reactor building is also half-buried with 15 metres sited below grade behind three concrete barriers and 15 metres above grade. The above-ground portion houses a crane capable of lifting the vessel for periodic replacement while the half-buried section contributes to mechanical stability under seismic loading and aircraft-impact.
According to Breyton, the various safety features allow for a plant design requiring no evacuation beyond the fence line even in the event of severe accidents including earthquakes, tsunamis and aircraft impact. Extensive simulations of seismic and impact loads, he says, support the safety case. As he says: “When these events happen, you can still do agriculture beside the reactor”.
Reactor life and re-processing
Within a Stellarium – the name of the commercial reactor design as envisaged – the effective operating life of a core load is limited not by fissile depletion but by accumulation of fission products. During operations a wide variety of isotopes accumulate within the salt, some of which have high neutron-absorption cross-sections. “After 40 to 50 years, 1% or 2% of these atoms start to pollute the neutronics,” says Breyton. When neutron economy becomes insufficient to sustain criticality at the rated power, additional U 238 can be added to the salt directly as fertile feedstock. “It’s not radioactive,” Breyton says, “at least in terms of decay heat and handling risk. As a result, small periodic additions of U 238 chloride are operationally straightforward, avoiding the enrichment and fabrication infrastructure used for today’s reactors”.
The design also includes two mechanisms that allow for removal of a significant fraction of the fission products that accumulate. Firstly, a substantial proportion of the non-soluble fission products adhere to the vessel walls. “One-third of the fission products… will stick on the walls of the vessel,” Breyton says, noting that this mechanism effectively reduces the specific concentration of the problematic decay products. Secondly, volatile species also accumulate in the cover gas. “The second third [of the decay products]… will be in the gas above the liquid, so you remove two-thirds of the fission products without changing the salt,” he says. Even so, every 10 years, the inner vessel and primary heat exchangers would be replaced. This schedule, Breyton argues, aligns with current regulatory inspection practices and also goes some way to mitigating the corrosion risks inherent to chloride salts. To enable replacement, the reactor incorporates a double-walled vessel arrangement on the primary side. Both walls are replaceable.
A complete chemical purification step is also required at the end of the operating life. This process will rely on a pyrochemical separation process, which Breyton indicated still needs to be developed. However, given the long lead times between installing the first Stellarium and the need to commission an optimised process, this is not seen as a significant obstacle to development: “We have 40 years ahead of us,” he says.
Following purification, the salt is intended to be reused in the same reactor type. “We multi-recycle the fuel,” Breyton says. “We create the salt, it stays for at least 20 years, and then the salt is cleansed for reuse.”
Development Path for ALVIN, MEGALVIN and First-of-a-Kind
Stellaria’s early development has centred on integrated multi-physics modelling and Breyton claims some 2,500 simulations have already been performed using nuclear codes derived from historical French fast-reactor studies.
With simulation and modelling underway, a three-stage technical demonstration roadmap has been devised prior to commercial deployment. The first experimental facility, dubbed ALVIN, is a small-scale system designed to demonstrate first criticality and characterise intrinsic safety behaviour, especially the thermal-expansion shutdown mechanism. “We will insert a lot of reactivity inside the reactor, expect the dilation of the fuel, and show how fast it is to stop the fission,” he says. Data from ALVIN will be used to support regulatory licensing and refine core models.
The second stage in the development process is MEGALVIN, a roughly 10 MWth prototype with sufficient neutron flux to test structural materials, pumps, and heat exchangers under representative irradiation conditions. MEGALVIN is intended to validate integrated systems performance and materials ageing. Brayton’s timeline places ALVIN operational at the end of 2030, MEGALVIN at the end of 2032, and the first commercial 250 MWe reactor at the end of 2035.
To this end, Stellaria positions itself as a reactor-technology developer rather than a plant constructor, but is working closely with multiple industrial partners to develop the physical plants. Overall, the approach reflects a modular supply chain. Engineering, procurement and construction (EPC) would be undertaken by Technip Energies, while Schneider Electric would provide electrical systems and automation. The French CEA is a simulation partner, and Orano is responsible for fuel-cycle tasks.
“We will not build the reactor ourselves,” Breyton says. “We provide the core system.” He expects projects to vary by site, depending on whether customers require electricity, hydrogen, steam, or mixed outputs.
In November 2025, Stellaria reached a first pre-order agreement for 500 MW with Equinix for its AI-Ready data centres using its Stellarium design. Commenting on the agreement Régis Castagne, Managing Director of Equinix France said in a statement: “We chose Stellaria because it is one of the few companies in the world capable of making our high-performance AI data centres energy resilient, while combining high security and flexibility.“
Heat and power
Stellaria’s commercial nuclear power unit is specified at 250 MWe per reactor, but the units will be built in pairs to share common infrastructure such as cranes and auxiliary systems. Operating salt outlet temperature is planned at approximately 605°C. At these temperatures, standard superheated steam cycles can be employed. Breyton estimates an electrical conversion efficiency of “at least 45%, maybe 47%,” significantly above the ~33% typical of large light-water reactors. He attributes this higher efficiency to the better thermodynamic qualities of the fast-spectrum molten-salt coolant.
Furthermore, because the steam conditions are compatible with those used in coal-fired power plants, he emphasises the retrofit option. “The steam turbines are the same as coal power plants,” he says. Existing power blocks can therefore theoretically be converted by replacing the coal-fired boiler with the reactor modules placed in an excavated pit matching the original boiler footprint. Brayton suggests that as many as 250 coal-fired plants in European alone could be candidates for decarbonisation using this approach.
A high-temperature heat output also opens the door to industrial applications for the design. Breyton describes potential deployments in hydrogen production or high-temperature process heat, siting reactor modules within industrial clusters to serve multiple users. “Very often those industries are together in the same area,” he observes.
Industrial operators, Breyton says, increasingly seek long-term power price predictability for capital-intensive facilities. “You know exactly the initial cost and can depreciate it easily,” he argues. The objective is a fully defined fuel cost trajectory for 20–40 years. He suggests that such conditions could even reverse industrial decline in regions with constrained grids: “If you have access to predictable-price power, you can secure a business plan and build factories.”
Challenges Ahead
While scaling from a 10 MWth prototype to a commercial 600°C, 250 MWe fast MSR in five years clearly represents a significant step and major R&D tasks, Breyton argues that the main challenge is actually financial. “We need to have enough financial support to cross the desert,” he says, referring to the pre-revenue period before commercial reactors are built. He says both private and public funding will be necessary to complete both ALVIN and MEGALVIN.
Regulatory licensing for a first-of-a-kind fast MSR will also require unprecedented depth of materials and safety data, although Breyton notes that licensing submissions are already underway with the French regulator ASN.
The Stellaria reactor combines established fast-spectrum physics with numerous passive safety features with inherent negative feedback loops and passive drain-down. There are also multiple claimed specific advantages of the liquid fuel design including a homogeneous composition, long residence times and fuel breeding. While substantial engineering challenges remain, Breyton argues that given the physics is sound, industrial energy requirements are now sufficiently intense to justify pursuing a fast-spectrum molten-salt architecture. The next decade, beginning with the ALVIN experiment, will determine whether the engineering path he outlines can be realised. Breyton, though, believes that Stellaria reactors will do for 21st-century industry what hydroelectric dams did for 20th-century electrification.
