Module replacements and production capacity across parallel assembly lines
Abstract
Copenhagen Atomics aims to mass-manufacture its thorium molten salt reactors at a rate of at least one reactor per day. Because reactor modules are expected to last only three to five years in the first commercial generation, and because their replacement is included in the annual fee, every replacement has to be produced by the same factories that build new reactors. In this analysis, the overall target of 365 reactors per year is held fixed, while the output of an individual line is treated as an unknown, so that the same target may be reached by one line or by several lines working in parallel. The analysis shows that the number of lines leaves the timing of saturation unchanged, since the production system still runs out of capacity for new customers after four years in the baseline, but that it raises the attainable fleet, because parallel lines can share the capacity cushion needed to protect replacement deadlines. The analysis is hypothetical and concerns the long run; its parameters can be modified in the panel below.
1. Background
The stated goal of Copenhagen Atomics is to deliver energy that is cheaper than coal, at a scale large enough to matter for global demand. Such a goal can only be reached if reactors are produced in very large numbers, which makes production capacity as decisive as the technology itself. To this end, the company is developing a thorium molten salt reactor that fits within a 40-foot container and, rather than building each reactor on site as is customary for conventional nuclear plants, plans to mass-manufacture it on assembly lines. Customers pay an upfront price per reactor together with an annual fee that covers both fuel and the replacement of reactor modules [1][2].
As a result, production capacity serves two purposes at once: building reactors for new customers and building replacement modules for the reactors already in operation, with the latter growing in proportion to the installed fleet.
The analysis rests on two figures taken from the public roadmap [1]:
- 1 reactor per day. The production target, which the roadmap equates to 15 GW of capacity per year, that is, 365 reactors annually.
- 3–5 years. The expected life of a reactor module in the first commercial generation, limited mainly by the core and the pumps [3], after which the module is replaced in its entirety.
The roadmap expresses the target per assembly line, but how the daily output will actually be organised is not known. One reactor per day could come from a single line with a cycle of one day, or from several lines working in parallel, each producing a fraction of a reactor per day. This analysis therefore treats the output of a single line, r, as an unknown, while holding the overall target of 365 reactors per year fixed, and examines how the number of parallel lines affects the results.
2. Replacement demand per reactor
A plant is designed to operate for 50 years or more [3]. Over this period, each reactor installed at a customer site requires a number of modules given by
where M is the number of modules per reactor, T the plant life in years and L the module life.
It follows that the sale of a reactor does not commit the factory to a single unit, but to a stream of production extending over several decades. This result does not depend on how production is organised, and is therefore identical to that of the single-line case.
3. Production output over time
4. The fleet ceiling
With the total output fixed at C = 365 reactors per year, the number of lines working in parallel follows directly from the output of a single line:
where r is the output of one line in reactors per day. In steady state, as in the single-line case, the replacements falling due each year absorb the entire planned output, which yields the maximum fleet the production system can sustain:
where L is the module life, u the planned utilization and k the work for a replacement relative to a new unit. Since C is fixed, the number of lines does not enter equation (3) directly; it does so only through u, as explained below.
4.1 Planned utilization and risk pooling
The planned utilization u is the share of annual output scheduled for customers, encompassing both new reactors and replacements, while the remainder forms a capacity cushion that is deliberately left unscheduled. The cushion protects replacement deadlines: a delay in delivering a new reactor postpones a sale, whereas a delay in delivering a replacement brings a running reactor to a halt, removing 42 MWe from the grid at the expense of Copenhagen Atomics, since replacements are covered by the fee. Planned utilization should not be confused with OEE, whose losses are already reflected in the actual output of 365 units per year.
When output is split across several lines, the cushion no longer needs to be sized for each line in isolation. If the modules are interchangeable and disruptions on different lines are independent, a stoppage on one line can be absorbed by the spare capacity of the others. This is the principle of risk pooling, and the square-root law commonly used for safety stocks gives a first approximation of its effect: the cushion required for the same level of protection decreases with the square root of the number of lines,
where c1 is the cushion that a single line would need on its own (10% in the baseline). With four lines, the cushion falls to 5% and the planned utilization rises to 95%. The square-root law is an approximation: the actual benefit would be smaller if disruptions were correlated, for instance through a shared supplier, or if the lines were located in different countries, as the roadmap suggests for mass manufacturing [1].
4.2 Work for a replacement vs. a new unit (k)
The parameter k expresses the factory work required for a replacement module relative to that required for a brand-new reactor. Since public sources indicate that the whole reactor unit is replaced, vessel included [1][3], the baseline assumes that a replacement is equivalent to a complete new unit (k = 1). Should part of the retired unit prove reusable, a replacement would demand less work, thereby freeing capacity for new customers and raising the fleet ceiling accordingly.
4.3 Line efficiency and cycle time
5. Lines needed only for replacements
6. Storage of used modules at the customer's site
7. Comparison with the single-line case
Treating the output of a single line as an unknown changes some results and leaves others untouched. The timing of saturation is unaffected: whatever the number of lines, the production system runs out of capacity for new customers after L years, because that timing depends only on module life. Replacement demand per reactor and the storage required at customer sites are likewise independent of how production is organised. What changes is the fleet ceiling, which rises with the number of lines because parallel lines can share the capacity cushion, and the design of each line, whose cycle time lengthens in proportion to the number of lines. In the baseline, moving from one line to four raises the ceiling from 1,314 to about 1,387 reactors, an increase of roughly 6% obtained without any additional output.
8. Limitations
The model is intentionally simple, since its purpose is to illustrate a mechanism and its order of magnitude rather than to produce a forecast. It assumes that modules do not fail before the end of their expected life, that production reaches full output immediately, and that installation and regulatory approval introduce no delays; it further assumes a plant life of 50 years and an output of 42 MWe per reactor. The pooling effect relies on interchangeable modules and independent disruptions across lines, which may not hold in practice. The values adopted for c1 and k are the author's assumptions and do not reflect company data.
References
- Copenhagen Atomics (2026). Company roadmap, version 1.8, February 2026. PDF
- Copenhagen Atomics. About Copenhagen Atomics (accessed October 2026). copenhagenatomics.com
- Glahn, J. (2025). A thorium molten salt reactor that refuels with fertile rather than fissile fuel. Modern Power Systems, May 2025. Link
The interpretations presented here are the author's own, and any corrections are welcome.