What levelized cost of energy means
LCOE answers one question: what constant price per kilowatt-hour, received every year for the life of the asset, would exactly pay back everything the asset costs at your required rate of return? It is a break-even tariff, not an accounting cost. That framing is what makes it comparable across technologies with completely different cost shapes — a solar array that is almost all capital and almost no fuel, against a gas turbine that is the reverse.
The definition used here is the standard one published by NREL and used in the IEA and IRENA cost reviews: present value of total costs divided by present value of total energy. Both sums are discounted at the same rate. That symmetry is the whole trick, and it is also the part practitioners most often get wrong.
LCOE deliberately excludes some things. It says nothing about when the energy arrives, so it cannot distinguish a kilowatt-hour delivered at the evening peak from one delivered at noon in the spring. It says nothing about how the energy is valued — for a rooftop system that depends entirely on the net metering or net billing tariff. And it says nothing about risk beyond whatever is baked into the discount rate. Treat it as a cost benchmark, then value the output separately.
Why energy is discounted too
The numerator is uncontroversial: capital cost at year zero, plus each year's O&M and any replacement, each divided by (1 + r)ᵗ. The denominator is where people balk. Discounting kilowatt-hours looks like a category error — a kilowatt-hour in 2050 is physically identical to one today.
It is not an error, and here is why. Suppose the asset sells energy at a constant price p. Revenue in year t is p · Eₜ, and the present value of all revenue is p · Σ Eₜ/(1+r)ᵗ. Set that equal to the present value of costs and solve for p: you get exactly the LCOE formula. The energy discounting is not a claim about physics; it is the algebraic consequence of demanding a single flat price. Energy delivered late buys less present value than energy delivered early because the money it earns arrives later.
Degradation enters through Eₜ = E₁(1−d)^(t−1), so year one is undegraded and each subsequent year loses a compounding fraction. Over 25 years at 0.5% per year, the final year produces 0.995²⁴ = 88.7% of year one, and the undiscounted lifetime total is about 6% below what a naive 25 × E₁ would suggest. Discounting compounds the penalty because the lost energy is concentrated in the late, heavily discounted years. The degradation calculator handles that decay in isolation.
Escalation on O&M works the same way in reverse: Mₜ = M₁(1+e)^(t−1). Escalating costs belong with a nominal discount rate. If you prefer to work in today's dollars, set escalation to zero and use a real discount rate obtained from the Fisher relation shown above. Mixing them — flat costs with a nominal rate, or escalating costs with a real rate — is the most common modelling error in this calculation, and it moves the answer by more than most of the input uncertainty.
Worked example: a three-year asset you can check on paper
Long horizons hide arithmetic, so work a short one first. Take a $6,000 system producing 5,000 kWh in year one, degrading 1% a year, costing $100 a year to maintain with no escalation, discounted at 5% over three years.
Costs. Capital is $6,000 at year zero, undiscounted. Then:
- Year 1: $100 ÷ 1.05 = $95.238
- Year 2: $100 ÷ 1.05² = $100 ÷ 1.1025 = $90.703
- Year 3: $100 ÷ 1.05³ = $100 ÷ 1.157625 = $86.384
Discounted cost = 6,000 + 95.238 + 90.703 + 86.384 = $6,272.325.
Energy. Production is 5,000, then 5,000 × 0.99 = 4,950, then 5,000 × 0.99² = 4,900.5 kWh.
- Year 1: 5,000 ÷ 1.05 = 4,761.90 kWh
- Year 2: 4,950 ÷ 1.1025 = 4,489.80 kWh
- Year 3: 4,900.5 ÷ 1.157625 = 4,233.24 kWh
Discounted energy = 13,484.94 kWh. Undiscounted it would be 14,850.5 kWh.
LCOE = $6,272.325 ÷ 13,484.94 kWh = $0.4651/kWh, or 46.51 cents. Had you divided by the undiscounted energy instead you would have got 42.24 cents — a 9% understatement from that single mistake, on a three-year asset. Over 25 years the same error is far larger.
Now the realistic case, with the calculator's defaults: $22,000 installed, 14,000 kWh in year one, 0.5% degradation, $150/yr O&M escalating at 2.5%, a $2,500 inverter replacement in year 13, 6% discount rate, 25 years. Discounted cost is $25,607 against 171,111 discounted kilowatt-hours, giving 14.97 cents per kWh. Against a 17-cent tariff that is a margin of about 2 cents.
How to read the number
An LCOE has no meaning without its discount rate. Quoting "11 cents" is like quoting a payment without a term. The same array can be 7.6 cents at a 0% discount rate and 17.7 cents at 10% — the reference table below shows exactly that spread on one fixed set of physical assumptions. When you compare two published LCOEs, the first thing to reconcile is the discount rate, the second is whether they are real or nominal, and the third is the analysis period.
Compare LCOE to the right benchmark. For a behind-the-meter residential system, the benchmark is the retail tariff you avoid — but only for the fraction of generation you actually consume on site. For exported energy the benchmark is the export credit, which is usually far lower, so a system whose LCOE beats the retail rate may still not beat its blended value. For a utility-scale project, the benchmark is the power purchase agreement price or the wholesale market's time-weighted value.
Watch the sensitivity, not just the point estimate. The chart under the calculator sweeps the discount rate from 0% to 12%. Where a large share of lifetime cost sits at year zero, as with solar, that curve is steep, and the LCOE is more sensitive to your cost of capital than to almost any physical parameter. Where costs are mostly recurring, the curve flattens. Knowing which regime you are in tells you where to spend your due diligence.
Finally, remember that LCOE is pre-tax and pre-financing in the form given here. Depreciation, tax credits taken over time and debt structure all change the answer for a taxable owner. Subtract only incentives you receive as cash at the start; anything realised later belongs in a full after-tax cash-flow model, not in this formula.
How the discount rate moves the LCOE
| Discount rate | Discounted cost ($) | Discounted energy (kWh) | LCOE (¢/kWh) |
|---|---|---|---|
| 0% | 25,124 | 329,783 | 7.62 |
| 3% | 23,436 | 231,459 | 10.13 |
| 5% | 22,715 | 188,231 | 12.07 |
| 7% | 22,195 | 156,324 | 14.20 |
| 10% | 21,658 | 122,477 | 17.68 |
Cells are produced by the same routine the calculator runs. Note that the discounted cost falls as the rate rises — the recurring O&M shrinks — while the discounted energy falls much faster, because the $20,000 of capital is never discounted at all. That asymmetry is why LCOE rises with the discount rate for capital-heavy assets.
Assumptions, limits and common errors
- Mixing real and nominal. Escalating O&M at 2.5% while discounting at a 6% real rate double-counts inflation. Pick a convention and hold it across every input.
- Dividing by undiscounted energy. The single most common arithmetic error. It always produces a lower LCOE than the correct calculation whenever the discount rate is above zero, and the gap grows with both the rate and the horizon.
- Using DC nameplate times sun-hours as production. Year-one energy must be AC output after inverter, soiling, wiring, shading and temperature losses. Take it from a modelled yield, not a rule of thumb.
- Ignoring the inverter. String inverters generally do not last 25 years. Leaving the replacement out understates LCOE; putting it in at year-one prices rather than the money of the replacement year understates it too, though by less.
- Treating LCOE as project value. It is a cost, not a return. A project can have a low LCOE and still be uneconomic if the energy it makes is worth less than that, which is exactly the situation a low export credit creates.
- Comparing across different analysis periods. A 20-year and a 30-year LCOE for the same hardware are different numbers. Normalise the horizon before comparing, or compare net present values instead.
- Forgetting that the discount rate encodes risk. A project with contracted revenue and one selling into a spot market do not deserve the same rate, and using one rate for both hides the difference the analysis was meant to reveal.
The same formula works for wind, hydro and storage
Nothing in the arithmetic is solar-specific. Replace year-one energy with a wind project's annual energy production, set degradation to zero or to a small availability decline, and add fuel to the recurring cost line for a thermal plant. The one adjustment worth making for storage is that a battery consumes energy as well as producing it, so the denominator should be net discharged energy and the round-trip loss belongs in the cost side as purchased electricity.
LCOE against payback, NPV and IRR
LCOE, simple payback, net present value and internal rate of return answer four different questions, and choosing the wrong one is how good projects get rejected.
Simple payback — years until cumulative savings equal cost — ignores the time value of money and everything after the payback date. It is a screening heuristic, useful for a homeowner comparing two quotes and misleading for anything with a long tail. The payback calculator covers it properly, including rate escalation.
Net present value is the right tool when you know what the energy is worth. It multiplies production by an actual value stream and discounts the difference, so it captures export credits, time-of-use structure and rate escalation that LCOE cannot see. IRR answers the same question as a rate rather than a dollar figure, which makes it comparable to your cost of capital but ill-behaved when cash flows change sign more than once — an inverter replacement can do that.
LCOE is for when you want to compare sources rather than evaluate a deal. It is the standard basis in NREL's Annual Technology Baseline and the IEA and IRENA cost surveys, which is why it is the number quoted when people say solar is cheaper than coal. Used that way it is excellent. Used as a proxy for whether one particular roof pays for itself, it quietly omits the two things that decide the answer: the tariff and the incentives. Run it alongside the energy payback time if you also care about the physical rather than financial return.