Levelized Cost of Energy: What It Measures and What It Hides
Almost every claim that one energy source is cheaper than another rests on LCOE. It is a genuinely useful number and a narrow one, and the gap between those two facts is where most energy arguments go wrong. Understanding what it includes takes five minutes; understanding what it excludes changes how the whole debate reads.
How the Number Is Built
LCOE asks a single question: what price per megawatt-hour would a plant need to receive, constantly across its entire life, to exactly repay everything it cost? Sum every expenditure - capital, fuel, operation, maintenance, finance, decommissioning - and divide by every megawatt-hour produced.
The complication is that money and electricity both arrive across decades, and a euro in 2050 is not a euro today. Both streams are therefore discounted to present value before the division, which is what turns a simple ratio into a calculation whose result depends heavily on an assumption.
Four inputs dominate. Capital cost is the overnight construction expense. The capacity factor determines how many megawatt-hours the denominator contains - a plant running 90 percent of the time spreads its capital across three times the output of one running 30 percent. Fuel cost is zero for wind and solar and substantial for gas. And the discount rate converts future money into present money.
That last input is where the arguments live, because it is a judgement rather than a measurement. It represents the cost of capital: what the money could have earned elsewhere, plus a premium for the risk that the plant underperforms or the market changes.
Why the Discount Rate Decides So Much
Technologies differ in when they spend. A nuclear plant or an offshore wind farm spends almost everything upfront and then runs for decades on low operating costs. A gas plant is cheap to build and expensive to run, with the fuel bill spread across its life.
Discounting penalises the first pattern and forgives the second. At a 4 percent discount rate, capital-heavy technologies look excellent, because their large early spending is weighed against decades of nearly free output. At 8 or 10 percent, the same spending is discounted far more harshly and the same plant looks expensive, while the gas plant barely moves because its costs arrive later and are discounted too.
The effect is large enough to reverse rankings. Studies have repeatedly shown nuclear and offshore wind changing from cheapest to among the most expensive purely by moving the discount rate across the range that different institutions consider reasonable. No physical property of the plant changed.
This is why comparing LCOE figures from different sources is often meaningless. Unless the discount rate, the assumed lifetime and the capacity factor are the same, the numbers are not measuring the same thing - and those assumptions are frequently in a footnote rather than the headline.
The Four Things It Leaves Out
Timing is the first and largest omission. LCOE treats every megawatt-hour as identical, but a grid does not. Electricity delivered during a winter evening peak is worth several times electricity delivered at midday in June when solar output already exceeds demand. As solar penetration rises, additional solar arrives increasingly at moments when the price is already low or negative - a phenomenon known as value deflation, and one that LCOE cannot see by construction.
System costs are the second. A source that varies requires the rest of the grid to compensate: storage, transmission, or dispatchable plant held in reserve. Those costs are real and are borne by the system rather than the project, so they appear nowhere in a project's LCOE. Estimates of their size vary widely, which is itself part of the problem.
Externalities are the third. Health damage from air pollution, climate damage from emissions and land use consequences are all excluded unless a carbon price internalises part of them. Coal's LCOE looks competitive precisely because the costs it imposes on other people are not in the calculation.
Grid services are the fourth and least visible. A conventional generator supplies inertia, voltage support and black-start capability as a by-product of spinning. An inverter-connected source does not, and the equipment that replaces those services costs money that appears in nobody's LCOE.
What to Use Instead
Value-adjusted LCOE weights each megawatt-hour by what the market actually pays at that moment, capturing the timing that plain LCOE ignores. It requires a model of the market it is being applied to, which makes it far less portable between studies and far less quotable in a headline.
System LCOE adds the balancing, network and backup costs a source imposes. It is the more complete figure and depends entirely on the grid in question - a solar farm in a system that is already 60 percent solar carries a very different system cost from the first one built.
The practical rule is that LCOE compares well within a category and badly across categories. Two wind farms, or two nuclear designs, can be ranked usefully. Comparing a wind farm with a gas plant using LCOE alone compares a variable source with a dispatchable one as though they were the same product, which they are not.
The honest formulation is that LCOE answers what a megawatt-hour costs to produce, and grids buy something more specific: a megawatt-hour at a particular time, with particular reliability, delivered to a particular place. Those questions have different answers, and the gap between them is where most of the disagreement in energy policy actually sits.
Frequently asked questions
What does LCOE actually mean?
The average price per megawatt-hour a plant would need to receive across its whole life to exactly cover every cost - construction, fuel, operation, finance and decommissioning - with both costs and output discounted to present value. It is a lifetime break-even price, not a market price.
Why do different sources give such different LCOE figures?
Because the result depends heavily on assumptions that are rarely in the headline: the discount rate, the assumed plant lifetime and the capacity factor. Moving the discount rate from 4 to 8 percent roughly doubles the LCOE of capital-heavy technologies while barely affecting gas. Unless the assumptions match, the numbers are not comparable.
Why does LCOE ignore when electricity is produced?
Because it is defined as total cost divided by total output, with no time weighting. A megawatt-hour at 3 a.m. in June counts the same as one during a winter evening peak, though the second may be worth ten times more. This is the single largest limitation and the reason value-adjusted LCOE exists.
What are system costs?
The costs a generator imposes on the rest of the grid rather than on its own project: transmission to reach it, storage or backup capacity to cover when it is not producing, and equipment to replace grid services that conventional plant supplies for free. They are real, they vary with how much of that source is already on the system, and they appear in no project's LCOE.
Is LCOE useless then?
No. It compares well within a category - two wind farms, two reactor designs - and it captures the genuine collapse in the cost of building solar and wind. It compares badly across categories, because a variable source and a dispatchable one are not the same product. The error is not using LCOE; it is using it alone.
