LCOE and LCOS
Levelized cost of energy (LCOE) is the price per unit of energy at which a project's lifetime net present value comes out to zero — the minimum price at which energy must be sold to break even. The simple version of the formula that most people quote excludes financing, discounting, future replacement and degradation, and its own documentation says so.
Levelized cost of storage (LCOS) applies the same net-present-value logic to a battery but divides by energy discharged rather than generated, which drags round-trip efficiency, degradation and cycle life into the denominator.
LCOE has published methodologies from national laboratories and intergovernmental agencies; LCOS, as of mid-2026, has no common definition, and peer-reviewed work finds studies vary in what they include. For a merchant battery earning on price spread rather than volume, a single cost per MWh is the wrong shape of answer.
Reviewed August 2026 by Sergey Syrvachev
New to BESS? Start free with the 7-email fundamentals course — no cost, no account.
What it is (precise)
LCOE is an economic assessment of an energy-generating system covering every cost over its lifetime — initial investment, operations and maintenance, fuel, cost of capital — with a net present value calculation solved so that at the chosen LCOE the project's NPV becomes zero. That makes it the minimum price at which energy must be sold for the project to break even.
The US national laboratory that publishes the Simple LCOE Calculator (the National Laboratory of the Rockies, on DOE contract DE-AC36-08GO28308 — the same laboratory whose 2024 Annual Technology Baseline pages still carry NREL attribution) states that LCOEs are typically calculated over 20 to 40 year lifetimes and quoted in currency per kilowatt-hour or per megawatt-hour. IRENA says the same thing in ratio form: discounted lifetime costs divided by discounted lifetime generation.
Two formulas matter. The simple one is sLCOE = (overnight capital cost × capital recovery factor + fixed O&M) / (8760 × capacity factor) + (fuel cost × heat rate) + variable O&M, with capital cost in $/kW, fixed O&M in $/kW-yr, variable O&M in $/kWh, fuel in $/MMBtu and heat rate in Btu/kWh.
Its documentation is blunt about what is missing: no financing, no discounting, no future replacement, no degradation. The Annual Technology Baseline publishes the full form, LCOE = (FCR × CAPEX + FOM) / (CF × 8,760) + VOM + FUEL − PTC, where the fixed charge rate is the capital recovery factor times a project financial factor.
LCOS is total lifetime investment cost divided by cumulative delivered electricity — the internal average price at which discharged electricity must be sold for the investment's net present value to be zero. Schmidt, Melchior, Hawkes and Staffell write it as investment cost plus discounted O&M plus discounted charging cost plus end-of-life cost, over discounted discharged electricity; note their end-of-life term is discounted at (1+r) to the power N+1, a year beyond the horizon of everything else.
They state that LCOS and LCOE can be compared directly, but that different concepts of providing electricity, and the resulting differences in cost calculation methodology, are exactly why a different name is used.
Why it matters in a real grid-scale project
The laboratory that publishes the Annual Technology Baseline warns on its own definitions page that LCOE does not necessarily show which technology would be the lowest-cost option for the grid in a given place and time, because it does not capture the economic value of a generation type to the system and therefore may not serve as an appropriate basis for comparing technologies.
The US Energy Information Administration goes further: direct comparisons of LCOE and LCOS across technologies are "problematic and misleading" as a method to assess economic competitiveness, and EIA recommends pairing them with the value metric it calls LACE, expressed as a value-cost ratio.
EIA's worked example is the one to remember. In AEO2022 the solar LCOE sits below combined-cycle LCOE, yet the model builds more combined-cycle capacity, because LCOE does not price relative system value. EIA also states the structural difference for storage plainly: a storage technology contributes to generation when discharging and consumes electricity from the grid when charging, and LCOS is calculated differently depending on whether it supplies energy or provides capacity reliability.
In its NEMS model EIA computes battery LCOS for energy-arbitrage applications only, not for generation capacity reliability — so a published LCOS already carries an assumption about what the battery is for.
For a merchant battery the deeper problem is that a cost per MWh discharged answers the wrong question. Value comes from the gap between the price at which you charge and the price at which you discharge, not from how many MWh you push.
Belderbos and colleagues at KU Leuven make that formal with the required average price spread: RAPS is the required average discharge price minus the average charging cost, and it reduces to capital costs, operation and maintenance costs, and the cost of round-trip efficiency losses, divided by discounted discharged MWh. A battery clears its hurdle when realised spreads beat RAPS. Throughput on its own tells you nothing.
- LCOE in one line
- The minimum price per unit of energy at which lifetime NPV = 0
- LCOS in one line
- Discounted lifetime cost / discounted lifetime energy discharged
- What simple LCOE omits
- Financing, discounting, future replacement and degradation costs
- LCOS charging term
- Discounted charging cost / discharged MWh = electricity price / round-trip efficiency
- LCOS denominator inputs
- Annual cycles, DoD, nominal capacity, RTE, self-discharge, cycle + time degradation
- LCOS lifetime
- The lesser of shelf life and cycle life measured against annual cycles
- Lazard v11.0 (2026), 100 MW / 4-hour
- $210-292/MWh unsubsidised; $148-209/MWh with the ITC
- Lazard v10.0 (June 2025), same case
- $115-254/MWh unsubsidised; $83-192/MWh with the ITC — costs rose year on year
- Standardisation status
- As of mid-2026 no common LCOS definition; studies vary on what they include
Typical values and standards
Published levelized costs are only as good as the financing assumptions behind them. IRENA's Renewable Power Generation Costs in 2018 reports LCOE in real 2018 USD at a fixed real cost of capital of 7.5% in OECD countries and China and 10% in the rest of the world, and states that all its LCOE calculations exclude the impact of any financial support.
IRENA is also explicit that more detailed discounted-cash-flow approaches — taking taxation, subsidies and other incentives into account — are what developers use to assess the profitability of real-world projects, and that those were beyond the scope of the report.
No standards body publishes an LCOS, so the market quotes an investment bank. Lazard's Levelized Cost of Storage Version 11.0 (2026) puts utility-scale standalone 100 MW / 4-hour storage at $210-292/MWh unsubsidised and $148-209/MWh with the investment tax credit; the 100 MW / 2-hour case runs $215-414/MWh unsubsidised and $152-282/MWh with the ITC.
The direction of travel reversed: Version 10.0 in June 2025 had the same 4-hour case at $115-254/MWh unsubsidised and $83-192/MWh with the ITC, and the 2-hour case at $129-277/MWh. Lazard states plainly that the analysis is not a forecasting tool and should not be used as such.
The Version 11.0 component breakdown shows where the money sits for the utility-scale 4-hour case: $123/MWh capital plus $28 fixed O&M plus $60 charging at the low end, published as $210/MWh, and $174 plus $51 plus $67 at the high end, published as $292/MWh — so charging energy is roughly 23-29% of LCOS. Read the methodology as well.
Lazard solves for the $/MWh value that produces a levered IRR equal to the assumed cost of equity, not a simple ratio, and its June 2025 utility-scale case assumes 91% efficiency, one cycle per day at 90% depth of discharge, 350 operating days a year and a 20-year economic life, on a 20% debt at 8% and 80% equity at 12% structure that it says is explicitly different from its LCOE work.
How it shows up in specs, studies and contracts
When a model hands you an LCOS, interrogate the denominator. In the Schmidt formulation, discharged electricity is annual cycles multiplied by depth of discharge, nominal energy capacity, round-trip efficiency and (1 − self-discharge), then multiplied year by year by (1 − cycle degradation) raised to the cumulative cycle count and (1 − time degradation) raised to the year, and discounted with construction time added to the exponent.
The lifetime is the lesser of shelf life and cycle life measured against annual cycles. Every warranty parameter you negotiate — depth of discharge, cycles per year, the degradation table, the end-of-life floor — lands directly inside that expression.
The charging term collapses to something you can check in your head. Discounted charging cost divided by discounted discharged electricity equals the electricity price divided by round-trip efficiency. At the 91% efficiency Lazard assumes, $50/MWh of charging energy costs about $55/MWh of energy discharged; six points lower and the same power costs about $59/MWh.
That is the formal mechanism by which efficiency losses raise levelized cost, and it is why a round-trip efficiency guarantee written at the wrong measurement boundary quietly moves LCOS by dollars per MWh. Make the efficiency in the financial model the same net AC number the EPC contract actually guarantees.
The metric itself is not standardised, and that is a contractual problem rather than an academic one. The peer-reviewed work behind the widely cited LCOS projections says findings differ because of varying input data and methodologies: replacement and end-of-life costs are often neglected, and cycle life, capacity degradation and self-discharge are not always considered.
Some studies report a net internal cost of storing electricity that excludes the electricity price and storage efficiency entirely; others report cost per unit discharged, which includes both. Those are different numbers wearing the same name. Write the definition you mean into the model assumptions, and follow the authors' advice that LCOS comparisons should always be application-specific.
Common pitfalls
The most common conceptual error is to say that charging cost is a term LCOS has and LCOE does not. LCOE already carries an input-energy term: Fuel in the IEA/NEA formulation quoted by the KU Leuven authors, and fuel cost × heat rate in the simple version. Charging cost is the storage analogue of that fuel term.
What is structurally different is that it is tied directly to round-trip efficiency and that, as the KU Leuven authors put it, the "fuel" — charged electricity — and the "generated electricity" — discharged electricity — are one and the same commodity. A one-for-one translation simply turns "fuel cost" into "charging cost"; treating LCOS as LCOE plus a term is wrong.
The second trap is treating a levelized cost as sufficient. The KU Leuven working paper is blunt: an LCOS defined as the fictitious average discharge price needed to break even is "incomplete", "insufficiently precise" and "might therefore lead to poor investment decisions".
The reason is concrete — energy capacity is not explicitly accounted for in these cost metrics, so the effect of a short duration on the number of discharging hours, and on the average price at which you can charge, is invisible in the result. Their recommendation is to use the levelized metric together with an analysis of entire historical price profiles for the market the asset will bid into.
Third, arbitrage-only cost metrics are pessimistic about revenue. The KU Leuven analysis accounts only for arbitrage revenues and says so, noting the results give a pessimistic outlook because technically suitable storage units are expected to provide ancillary services as well.
Lazard makes the same admission structurally: alongside the headline LCOS it publishes location-specific Storage Value Snapshots built on hourly wholesale prices and stacked revenue streams, and states that levelized costs in those snapshots differ in certain cases from the headline LCOS. Both of Lazard's analyses also exclude congestion, curtailment and other integration-related costs and permitting and development costs, so the headline figure is not the whole cost either.
- Simple Levelized Cost of Energy (LCOE) Calculator Documentation — National Laboratory of the Rockies (DOE)
- Annual Technology Baseline 2024 — Equations & Variables — National Laboratory of the Rockies (DOE)
- Annual Technology Baseline 2024 — Definitions (LCOE limitations) — National Laboratory of the Rockies (DOE)
- IRENA — Renewable Power Generation Costs in 2018 (LCOE methodology)
- Schmidt, Melchior, Hawkes & Staffell — Projecting the future levelized cost of electricity storage technologies (accepted manuscript, Imperial College London repository)
- Belderbos, Delarue, Kessels & D'haeseleer — The Levelized Cost of Storage critically analyzed, KU Leuven TME Working Paper WP EN2017-08
- Lazard's LCOE+ (2026) — Levelized Cost of Storage Analysis Version 11.0, Introduction, p.20
- Lazard's LCOE+ (June 2025) — Levelized Cost of Storage Version 10.0 (p.19), LCOS methodology (p.42), Storage Value Snapshots (p.24), and LCOE Version 18.0 exclusions (p.7)
- U.S. EIA — Levelized Costs of New Generation Resources in the Annual Energy Outlook 2022 (pp.3-4)
LCOS is just LCOE for batteries, so you can line a battery's LCOS up against a solar or gas LCOE and pick the cheaper one.
In reality: EIA states that direct comparisons of LCOE and LCOS across technologies are "problematic and misleading" as a method to assess economic competitiveness, and pairs them with a value metric (LACE) as a value-cost ratio; the Annual Technology Baseline makes the same point about LCOE alone, which does not capture economic value to the system. Worse, LCOS has no common definition — peer-reviewed work finds studies variously omit replacement, end-of-life, cycle life, capacity degradation and self-discharge, and some exclude the electricity price and storage efficiency altogether. Two LCOS figures may not be measuring the same quantity.
LCOE and LCOS, in context.
The Grid-Scale BESS course covers lcoe and lcos — and the rest of the system — from the ground up, the way it actually gets deployed.