The Economics of US Power Generation

1. What a power plant costs to build

Overnight capital cost, reference plant, 2023 dollars

The starting number is overnight capital cost: what a developer would pay to build the plant if it could be built instantly, with no interest accruing during construction. It is quoted in dollars per kilowatt of net capacity, which makes plants of very different sizes comparable. Divide by 1,000 and you get the more intuitive cost per watt.

The spread is enormous. A gas combined-cycle plant costs about 87 cents per watt. A two-unit AP1000 nuclear station costs $7.86 per watt, nine times more for the same nameplate capacity. That single ratio explains most of what has been built in the United States over the past two decades.

Overnight capital excludes financing costs and regional multipliers. Lead time is the total from start of development to commercial operation. Heat rate of zero means the technology burns no fuel. The biomass case is the only biomass configuration EIA costed for AEO2025 and it includes 95% carbon capture, which is why it is far more expensive than the roughly $4,500 per kilowatt an uncontrolled biomass plant carried in earlier editions. Battery storage capital is also quoted as $436 per kilowatt-hour of storage.

2. The arithmetic that turns a plant into a price

Levelized cost of electricity, worked line by line

Capital cost alone cannot rank technologies, because a plant that runs constantly spreads its cost over far more electricity than one that runs occasionally. The standard way to compare is the levelized cost of electricity, or LCOE: total lifetime cost divided by total lifetime output, expressed per megawatt-hour. It is the break-even price. Sell above it and the project earns its cost of capital; sell below it and it does not.

Step 1: annualise the capital

A plant is paid for once and earns revenue for decades, so the capital cost has to be converted into an equivalent annual payment, exactly like a mortgage. That conversion is the capital recovery factor:

So each year the project must produce 8.059% of its capital cost just to service and retire the investment.

Step 2: work an actual plant

Take the reference gas combined-cycle plant: 1,227 MW at $868 per kilowatt, running at the 36.0% capacity factor the US gas fleet actually achieved in 2025.

That is 5.45 cents per kilowatt-hour, and the single largest line is not the plant. It is the gas.

3. The full comparison

Levelized cost by component, sorted cheapest to dearest

Running that same arithmetic across every technology gives the table below. Each row uses the capacity factor the US fleet actually achieved in 2025 where this site has the data, and a stated assumption where it does not.

Technologies without a measured 2025 fleet capacity factor on this site are modelled at 50%, except the H-class peaker at 10% and offshore wind at 45%. All figures exclude federal tax credits, which is a significant omission discussed in section 9.

4. The cost of capital decides more than the technology

Same plants, four different interest-rate environments

Because most of these technologies are mortgages with turbines attached, the interest rate is not a detail. It is frequently the largest single determinant of whether a project is viable. Holding everything else constant and moving only the cost of capital produces this:

Going from a 4% to a 13% cost of capital raises the cost of nuclear power by 84% and the cost of hydroelectricity by 89%, but raises gas combined-cycle power by only 43%. A capital-heavy, fuel-free plant is essentially a bond; a gas plant is essentially a fuel-conversion service with a modest amount of equipment attached.

5. Capacity factor is the other multiplier

The same capital spread over more or less electricity

Capital cost per kilowatt tells you what a plant costs. Capacity factor tells you how much electricity that purchase actually yields. The two together determine cost per unit far more than either alone.

Read the top row. At a 10% capacity factor, nuclear power would cost $913 per megawatt-hour and gas would cost $122. Read the bottom row: at 90%, nuclear is $111.70 and gas is $38.88. The technology did not change. Only the utilisation did.

This is why plant types specialise. A gas peaker uses the cheapest capital available, $836 per kilowatt, because it is expected to run only a few hundred hours a year and cannot spread a large investment over that. A nuclear plant does the opposite: it accepts a very large capital cost because it expects to run more than 90% of every hour of its life. Choosing the wrong technology for a duty cycle is more expensive than choosing the wrong technology.

6. Fuel exposure

What a gas price move does to a combined-cycle plant

Fuel is 52% of the lifetime cost of a gas combined-cycle plant at $4.00 per million Btu, so the gas price passes almost directly into the cost of the electricity.

A plant financed and built at a fixed cost still sees its output cost swing from $41.97 to $92.09 per megawatt-hour purely on fuel, a range of 119%. For comparison, nuclear fuel is $9.02 per megawatt-hour and coal is $21.16, while wind, solar, hydro and geothermal are zero. A doubling of the uranium price moves nuclear power by about 8%; a doubling of the gas price moves gas power by about 46%.

7. How long the money takes to come back

Power price required to repay capital within a given period

Developers and lenders do not usually think in levelized costs. They think about whether the debt gets repaid, and how quickly. The table below inverts the question: given the plant, what flat power price is needed to return the capital, with interest at 7%, within 10, 15, 20 or 30 years?

The 30-year column is the LCOE from section 3, which is the point: levelized cost is simply the break-even price over the full recovery period. The value of the table is the left-hand columns. A gas plant needs $71.51 per megawatt-hour to pay back in a decade, which is well within the range wholesale markets reach. A nuclear station needs $168.94, which is roughly double typical US wholesale power prices, and is the arithmetic reason new nuclear generally requires a regulated rate base, a long-term power purchase agreement, or direct government support to reach a final investment decision.

8. Staff, maintenance and the annual operating bill

What it costs to keep a plant running once it is built

Fixed operations and maintenance is everything the plant spends whether or not it generates: salaries, scheduled maintenance, contracted services, and general administration. It is the number that decides whether an older plant stays open. EIA publishes it as dollars per kilowatt of capacity per year.

For the reference coal plant, the engineering study breaks the bill out in full:

Labour alone is 48% of the fixed bill for a coal plant. Scaling that across the fleet explains a great deal about which technologies survive:

• Gas combined cycle, $12.12 per kilowatt-year. A 1,000 MW combined-cycle plant can be run by roughly 30 people. Highly automated, few moving parts relative to output, and no fuel handling beyond a pipe.
• Solar PV, $20.23 per kilowatt-year. No fuel, no rotating machinery, no combustion. The bill is inverter replacement, module washing, vegetation control and monitoring.
• Onshore wind, $33.06 per kilowatt-year. Gearboxes, blades and towers are mechanical assets in a hostile environment, so wind costs meaningfully more to maintain than solar.
• Coal, $61.60 per kilowatt-year. Fuel handling, ash disposal, emissions control chemistry and a large boiler crew. Roughly 150 staff at the reference plant.
• Nuclear, $156.20 per kilowatt-year. The Nuclear Energy Institute puts staffing at 500 to 800 people per plant, and the OECD Nuclear Energy Agency at about 600 for a single large unit. Regulatory staffing minimums, security, quality assurance and a permanent engineering organisation are not optional.
• Offshore wind, $154.00 per kilowatt-year. Nearly nuclear-level operating costs for an entirely different reason: every maintenance visit requires a vessel or a helicopter.
• Geothermal, $150.60 per kilowatt-year. High for a renewable, because the resource itself needs management: wells decline, scale and corrosion are constant, and make-up drilling is periodic.

9. Strengths and weaknesses, source by source

The financial case for and against each technology

10. What this analysis leaves out

Every levelized comparison hides things, and the honest way to present one is to say which.

• Tax credits. The capital costs used here exclude the investment and production tax credits. A 30% investment tax credit takes solar from $79.05 to roughly $56 per megawatt-hour and reorders the table. The credits are also politically unstable, which is itself a financing risk.
• Interest during construction. Overnight cost assumes instant construction. A nuclear plant taking seven years accrues very substantial interest before earning anything, which real project economics must carry but overnight cost does not show.
• Grid value is not equal across technologies. A megawatt-hour at 6pm on a hot evening is worth far more than one at 2am in spring. Dispatchable plants capture that; solar and wind mostly cannot. EIA models this separately as levelized avoided cost of electricity, and comparing LCOE alone understates dispatchable value.
• Transmission and integration. Remote wind and solar frequently need new transmission that is not in the plant cost. Nor is the cost of backup capacity for hours when they do not produce.
• Decommissioning and waste. Nuclear carries funded decommissioning obligations and spent-fuel liabilities that other technologies do not. Coal ash and turbine blade disposal are smaller but real.
• Carbon is priced at zero here. Any carbon price shifts the ranking sharply against coal first and gas second.
• These are new-build costs. Almost every plant on this site already exists and has sunk its capital. An existing paid-off plant competes on fuel and O&M alone, which is a completely different and much lower number.

11. Cross-checking against the market

An analysis that only agrees with itself is not worth much. Lazard publishes an independent levelized cost study built from market data rather than engineering estimates. Its June 2025 edition gives unsubsidised ranges of $50 to $110 per megawatt-hour for gas combined cycle, $37 to $86 for onshore wind, $38 to $78 for utility solar, $66 to $109 for geothermal, $70 to $170 for coal, $70 to $157 for offshore wind, and $140 to $220 for nuclear.

The model on this page lands inside those ranges for gas ($54.50), wind ($54.60), geothermal ($103.97), coal ($124.69) and offshore wind ($114.48). It sits slightly above Lazard for solar, because this page uses the 20.4% capacity factor the existing US fleet actually delivered rather than the higher figure a new tracking plant in a strong resource area would achieve. It sits below Lazard for nuclear, because Lazard anchors on the actual cost of Vogtle units 3 and 4, which came in at an estimated $169 and $228 per megawatt-hour after well-documented overruns, while this page uses the engineering estimate for a plant built to schedule. The gap between those two nuclear numbers, roughly $109 against $169 to $228, is the single most expensive lesson in recent US energy construction.