Economics of advanced nuclear

The last chapter of the options study is the one the consortium's finance committee will read first. Ingrid, Rohan and Elena meet with Hana to agree its structure. “Every option in this report is a capital project with long lead times,” Elena says. “The technology chapters tell them what is possible. This chapter tells them what it might cost, what drives the cost, and which risks they would be carrying.” Rohan adds one rule: every number must show its assumptions, especially the discount rate.

This lesson develops the economic tools used to compare nuclear and other energy options: overnight cost, interest during construction, capital recovery, levelized cost and the role of capacity factor. It examines why nuclear costs are dominated by capital and financing, what evidence exists on first-of-a-kind and repeat costs, how learning and standardization might lower them, and how the economics of advanced fission and fusion differ. It ends with the framework Hana uses to bring the options study together.

The structure of nuclear costs

The cost of electricity or heat from a nuclear plant has three parts: capital, operations and maintenance, and fuel. For water reactors, capital and its financing typically make up 60 to 80 percent of the total, operations and maintenance 15 to 30 percent, and fuel only 5 to 15 percent. That structure is the opposite of a gas-fired plant, where fuel dominates. It makes nuclear costs very sensitive to construction cost, construction time and the cost of money, and very insensitive to fuel prices.

Capital cost is usually quoted as an overnight cost: the cost per kilowatt of building the plant as if it were completed instantly, including engineering, equipment, construction labor, owner's costs and contingency, but excluding financing. Interest during construction then adds the cost of money spent years before the plant earns revenue. For a project that spends evenly over six years at a 6 percent cost of capital, interest during construction adds about 20 percent to the overnight cost; at 9 percent and eight years, about 40 percent.

Levelized cost

The levelized cost of electricity (LCOE) is the constant price per megawatt-hour that would recover all costs over a plant's life, including a return on capital at the chosen discount rate. Capital is converted into an annual payment with the capital recovery factor,

for discount rate r and economic life n years, and the LCOE is the sum of the annual capital charge and fixed operating cost, divided by the energy produced per year, plus variable operating and fuel costs per unit of energy. The same approach gives a levelized cost of heat or of hydrogen.

# Levelized cost of electricity for capital-intensive plants: overnight cost, interest during
# construction, discount rate and capacity factor, plus a learning curve for repeat units.
# All costs are illustrative, in US dollars of a common year, per kW of net electrical capacity.
def crf(r, n):
    return r * (1 + r) ** n / ((1 + r) ** n - 1)

def idc_factor(r, T):
    # spending spread evenly over T years, interest compounded to commercial operation
    return sum((1 / T) * (1 + r) ** (T - t - 0.5) for t in range(T))

def lcoe(overnight, T, r, cf, fom=150.0, fuel_vom=12.0, life=40):
    cap = overnight * idc_factor(r, T) * crf(r, life)    # dollars per kW per year
    return (cap + fom) / (8760 * cf) * 1000 + fuel_vom   # dollars per MWh

options = [  # name, overnight cost per kW, construction years
    ("Repeat SMR (nth unit)", 9000, 4), ("First SMR (first of a kind)", 16000, 6),
    ("Gen IV demonstration", 18000, 7), ("Fusion pilot (speculative)", 25000, 8)]
print(f"{'option':30s}{'overnight':>10s}" + "".join(f"{'r=' + str(int(r * 100)) + '%':>9s}" for r in (0.03, 0.06, 0.09)))
for name, oc, T in options:
    vals = "".join(f"{lcoe(oc, T, r, 0.90):9.0f}" for r in (0.03, 0.06, 0.09))
    print(f"{name:30s}{oc:10,d}{vals}")
print("(LCOE in dollars per MWh at 90% capacity factor)")
# Capacity factor matters as much as cost: the same plant at lower utilization
for cf in (0.95, 0.90, 0.75, 0.50):
    print(f"first SMR at r = 6%, capacity factor {cf:.0%}: {lcoe(16000, 6, 0.06, cf):5.0f} dollars/MWh")
# Learning curve: each doubling of units built cuts unit cost to LR times the previous
import math
LR = 0.90
b = math.log2(LR)
for n in (1, 2, 4, 8, 16):
    print(f"unit {n:2d}: overnight cost {16000 * n ** b:7,.0f} dollars/kW")

option                         overnight     r=3%     r=6%     r=9%
Repeat SMR (nth unit)              9,000       83      116      158
First SMR (first of a kind)       16,000      127      192      278
Gen IV demonstration              18,000      141      218      322
Fusion pilot (speculative)        25,000      186      299      455
(LCOE in dollars per MWh at 90% capacity factor)
first SMR at r = 6%, capacity factor 95%:   183 dollars/MWh
first SMR at r = 6%, capacity factor 90%:   192 dollars/MWh
first SMR at r = 6%, capacity factor 75%:   229 dollars/MWh
first SMR at r = 6%, capacity factor 50%:   337 dollars/MWh
unit  1: overnight cost  16,000 dollars/kW
unit  2: overnight cost  14,400 dollars/kW
unit  4: overnight cost  12,960 dollars/kW
unit  8: overnight cost  11,664 dollars/kW
unit 16: overnight cost  10,498 dollars/kW

Three lessons emerge from the table. First, the discount rate matters as much as the overnight cost: the same first small modular reactor costs about 130 dollars per MWh at 3 percent and nearly 280 at 9 percent. This is why government ownership, loan guarantees, regulated returns and low-cost public lending have been central to almost every nuclear project of the past two decades. Second, capacity factor matters, because a capital-heavy plant must run as much as possible: the same reactor at 50 percent utilization costs nearly twice as much per MWh as at 95 percent. Third, learning can help, but slowly: with a 90 percent learning rate, cost falls by a third only after about sixteen units.

First-of-a-kind costs and the evidence

Recent experience shows how much first-of-a-kind projects cost. Vogtle Units 3 and 4 in Georgia, the first new US reactors in a generation, entered commercial operation in 2023 and 2024, years late and at a total project cost widely reported to exceed 30 billion dollars for about 2,200 MWe. In Canada, Ontario's first BWRX-300 small modular reactor at Darlington is estimated at about 6.1 billion Canadian dollars for the first 300 MWe unit, including shared infrastructure, and about 20.9 billion for four units. TVA has applied to build a BWRX-300 at Clinch River, near Knoxville, and Hana notes that a repeat of the same design in the United States would be a test of how quickly first-of-a-kind costs fall.

Analyses of past programs show both sides. France's fleet of the 1970s and 1980s and South Korea's program of standardized units kept costs fairly stable by building the same design repeatedly with the same teams. The United States in the 1970s and 1980s, and several Western projects since 2005, saw costs rise sharply as designs changed during construction, regulations changed, supply chains and skilled trades atrophied, and projects started before designs were complete. The single strongest predictor of cost overrun is starting construction with an incomplete design.

The economics of Generation IV

Advanced fission reactors add both cost risks and sources of value. Their first units carry the cost of new fuel types, new coolant technologies and new regulatory reviews, and their supply chains are thin. On the other hand, higher outlet temperatures raise efficiency and open heat markets; low-pressure coolants and passive safety could reduce the scale of safety-grade structures; and features such as Natrium's molten salt storage let a plant sell more of its output when prices are high. For a client like the consortium, the value of steam that replaces natural gas, plus avoided carbon costs, can be compared directly with the levelized cost of heat, which for a given reactor is lower than its levelized cost of electricity per unit of energy because no conversion losses are incurred.

The economics of fusion

Fusion economics are far less certain because no fusion plant has been built, and published cost estimates range widely. A few structural features are nevertheless clear. A fusion plant would be at least as capital-intensive as a fission plant, with expensive magnets or drivers, a large vacuum vessel, extensive remote handling equipment and a tritium plant. Its fuel cost would be negligible. Its recirculating power, for heating, current drive, cryogenics and pumping, would reduce net output, so a plant with an engineering gain of 4 would need to build 33 percent more gross capacity than it sells. Its availability would depend on how often first walls, blankets and divertors must be replaced, and how long each replacement takes; a plant that must stop for six months every four years cannot exceed about 85 percent availability even if nothing else goes wrong. And its licensing, under a lighter framework than fission, might cost less and take less time.

Studies that try to estimate future fusion costs generally find that fusion could compete only if overnight costs fall to levels similar to those hoped for advanced fission, availability exceeds about 75 percent and the physics allows compact, high-power-density designs. Fusion companies argue that high-field magnets and factory-built components will deliver those conditions; skeptics point out that no confinement scheme has yet operated with a breeding blanket, and that the first plants will carry first-of-a-kind costs just as fission plants do. Both sides agree that fusion's first contribution to supply will come later than that of advanced fission.

Policy, risk and financing

Because the cost of capital dominates, the way a project is financed and its risks shared can matter more than its engineering. Federal measures in the United States include clean electricity and hydrogen tax credits, Department of Energy loan programs and cost sharing for demonstrations under the Advanced Reactor Demonstration Program, and large technology companies have begun signing long-term power purchase agreements with advanced fission and fusion developers. For an industrial consortium, the options include buying heat or power under long-term contracts from a utility-owned reactor, joint ownership with a utility, or a build-own-operate arrangement with a vendor, each with different allocations of construction, performance and regulatory risk.

Bringing the options study together

Hana's final chapter for the consortium ranks the options not by a single cost figure but by readiness, cost range, fit to the site's heat demand, and the milestones that would change the ranking. Water-cooled small modular reactors are being built now and could supply steam below 300 °C and electricity in the 2030s, at first-of-a-kind costs that should fall if fleets are built. High-temperature gas reactors and other Generation IV designs fit the site's higher-temperature heat and hydrogen needs better, with first commercial units expected around 2030 and costs still uncertain. Fusion is a strategic option to monitor for the 2040s, to be reassessed as pilot plants demonstrate net electricity, tritium breeding and availability.

Ingrid reads the last page and adds a sentence of her own before the report goes out: the most valuable thing the consortium can do now is to keep its site, grid connection, steam network and permitting options open, so that whichever technology matures first can be used.