Agreements

Capacity maintenance agreement CMA

A capacity maintenance agreement is the obligation to keep a battery plant's deliverable energy at or above a contracted floor for the term, by adding capacity or by another agreed remedy. That is a different promise from an energy-retention capacity warranty, which permits the fleet to fade along a guaranteed curve and pays damages if it fades faster — a promise with money attached.

The distinction is narrower than it first looks: a maintenance or augmentation warranty, the other structure the capacity-warranty entry documents, already carries a work obligation, and where it does the two instruments differ mainly in which contract the obligation is written into.

A capacity maintenance agreement fixes a floor and owes the work that holds it: racks, containers, a crew and an outage. It is the occasional third contract alongside the battery energy supply agreement and the long-term service agreement, and it is just as often a clause inside one of them, so the name follows the packaging rather than the obligation.

Nothing about it is statutory. The floor, the triggers, the term and the remedies mean what the specific agreement says they mean.

Reviewed August 2026 by Sergey Syrvachev

New to BESS? Start free with the 7-email fundamentals course — no cost, no account.

A promise with money versus a promise with work

Under an energy-retention warranty the guaranteed quantity declines. A year-by-year table sets the minimum retained capacity, and the supplier owes a remedy only when a measured result falls below it. The owner carries the shape of that curve: it funds the day-one overbuild that keeps a faded plant above its contracted obligation, and it keeps whatever shortfall survives a capped, sole-remedy damages payment.

Under a capacity maintenance agreement the guaranteed quantity is flat, or steps on an agreed schedule, and the supplier owes performance rather than payment — when measured energy approaches the floor, capacity gets added.

The two are layers more than alternatives. A CMA normally sits on top of a warranty curve: the warranty defines the fade the supplier is entitled to, the CMA defines the number the plant must hold regardless, and both are conditioned on the same operating envelope of cycles, depth of discharge, temperature, resting state of charge and C-rate.

Operate outside it and the CMA is as voidable as the warranty. The cap question moves too. Instead of asking what percentage of the fee damages are capped at, ask how many megawatt-hours of augmentation the supplier owes across the term, at whose price, and what happens once that quantity is exhausted.

What it actually obliges

Four mechanics carry the contract. A measurement: a periodic capacity test at a named protocol, boundary and condition set, because the floor is only as well defined as the test that proves it — and it has to be the same quantity, at the same meter, in the same year as the offtake's contracted energy.

A trigger: written either against a measured result below the floor plus a margin, or against a forecast crossing derived from trended state-of-health data. Forecast triggers are the ones that work, because racks have lead times and an installation window inside an energised plant does not get scheduled in a week.

A remedy with a cure period: additional racks or containers, occasionally a widened SOC window or a replacement, within a stated number of days from the failed test. And an outage allowance, because installing augmentation takes the affected blocks down — unless the availability guarantee excludes or caps those hours, the supplier discharges one obligation by breaching another. Both documents are usually drafted by different people, and both usually assume the other one handled it.

The physical preconditions are fixed years before the obligation bites, at COD. Reserved land and foundations, spare feeder positions and conduit, a DC bus voltage window the future racks can land in, PCS and transformer headroom, and enclosure separation that still complies when new containers arrive. A CMA on a site with no reserved plot is a promise that year-eight engineering will be free.

A capacity maintenance agreement owes WORK that holds a floor; an energy-retention warranty permits fade and pays damages — the two differ in what the remedy is made of.
reserved land and foundationsfeeder positionsa compatible DC bus voltage windowPCS and transformer headroomseparation that still compliesunder the NFPA 855 edition in force AT THETIME of the additionaugmentation can actually landthe obligation has to arrive somewherephysicalThe work is priced, not gifted — and the promise is worth the promisor's credit in year twelve.

If those preconditions were not reserved at COD, the supplier's promise collides with the as-built plant and becomes a construction project inside an energised site. Installing augmentation also takes blocks offline: unless the availability guarantee excludes or caps those hours, one obligation is discharged by breaching another.

Key facts
What it obliges
Hold deliverable energy at or above a contracted floor by doing work — added racks or containers — rather than by paying damages for the shortfall
Versus a capacity warranty
An energy-retention warranty permits fade along a guaranteed curve and pays when the curve is missed; a CMA fixes the floor and owes the augmentation that holds it. Where a CMA exists it layers on top of the curve rather than replacing it
Where it sits
The occasional third contract alongside the BESA/BSA and the LTSA — and just as often a clause inside one of them
Definition source
Contractual, not statutory: the floor, triggers, cure periods, term, remedies and caps mean what the specific agreement says
Trigger and measurement
A periodic capacity test at a named protocol and boundary, with the trigger written against a measured result below the floor plus margin, or a forecast crossing from trended state-of-health data
Physical preconditions fixed at COD
Reserved land and foundations, feeder positions, a compatible DC bus voltage window, PCS and transformer headroom, and separation that still complies when new containers land
Added equipment is certified separately
UL 9540 (system), UL 1973 (cell to rack), IEC 62619 (cells and batteries) plus UL 9540A data for the product supplied, installed under the NFPA 855 edition in force at the time of the addition
Outage interaction
Installing augmentation takes blocks offline — unless the availability guarantee excludes or caps those hours, one obligation is discharged by breaching another
Scale of what is promised
Cumulative augmentation commonly runs 15-40% of the original DC build over a 15-20 year term, carried as a dedicated reserve rather than ordinary O&M
Credit is the substance
The obligation is worth the promisor's ability to perform in year twelve — parent guarantee, letter of credit, funded reserve or step-in rights

Who owns the added modules, and what warranty they carry

Title deserves an explicit clause: whether added racks become the owner's property on installation or on termination, whether they are removable, and what happens if the agreement ends early with augmentation already delivered. The matching question is what warranty the new equipment carries.

Suppliers commonly start a fresh degradation clock for newly added capacity while the original fleet stays on its original curve, so the plant ends up with two curves and one floor — and the contract has to say how the blended fleet is measured, because the capacity test integrates the whole plant and does not care which racks are which.

The engineering constraints arrive with the racks. Cells in series carry the same current, and racks paralleled on a shared DC bus are pulled to the weakest string's window, so mixing fresh and faded capacity behind one PCS gives back part of what was just bought; vendors typically require augmentation racks on a separate DC bus or a dedicated PCS input.

The added product is certified in its own right — UL 9540 as a system, UL 1973 for the battery from cell to rack, IEC 62619 for cells and batteries where those apply, with UL 9540A fire-test data for the product actually supplied — and installed under the NFPA 855 edition the authority having jurisdiction enforces at the time of the addition, which may be stricter than the one the original plant was permitted under. Cell formats turn over every few years, so a year-ten addition almost never uses the original part.

The credit behind a twenty-year obligation

A promise with work attached is worth the promisor's ability to do the work in the year it is called. Counterparty analysis is therefore the substance of a CMA review rather than a footnote to it: whether the signing entity is the same thin project-level subsidiary that signed the supply agreement or has a parent guarantee behind it, whether there is a letter of credit, a funded reserve or an escrow that survives the supplier, and whether the owner holds step-in rights and a priced alternative if the supplier is gone.

Pricing structure carries the same exposure — an obligation held as an annual fee, as pre-agreed unit prices per rack, or as a fixed-price option with an expiry are three different bets on a future cell price, and fixed-price augmentation options are valuable and rare.

Lenders read it as an input to the degradation forecast rather than as a covenant. A CMA that holds converts the retention curve in the base case from a declining line into a flat one, which is worth real debt capacity. A CMA whose counterparty cannot be relied on in year twelve converts nothing, and the model reverts to the warranty curve plus an owner-funded augmentation reserve. Carrying that reserve anyway is the cheap version of the same insurance: cumulative augmentation over a 15-to-20-year term commonly runs 15 to 40 percent of the original DC build, which is not a number to meet late.

Where it collides with the other contracts

Augmentation is commonly a separately priced option under a long-term service agreement rather than automatic scope, while the guaranteed degradation curve in that same document may quietly assume it happens. If the CMA assumes the LTSA does the work and the LTSA prices it as an option, the gap is funded by whoever notices last. Write the interface line by line: which document owns the test, which owns the trigger, which owns the labour, which owns the parts, and which absorbs the outage hours.

On the revenue side the floor has to be the quantity the offtake actually enforces. A CMA holding usable DC energy at the racks while the tolling agreement pays for delivered energy at the point of interconnection leaves the whole conversion and auxiliary chain unprotected, and that chain is several percent of the plant. One quantity, one meter, one condition set, written identically in both documents.

Common misconception

With a capacity maintenance agreement in place, capacity is the supplier's problem for twenty years — the owner needs no reserved footprint and no augmentation budget.

In reality: The obligation still has to land somewhere physical. If land, foundations, feeder positions, DC bus headroom and code-compliant separation were not reserved at COD, the supplier's promise collides with the as-built plant and becomes a construction project inside an energised site. The work is priced, not gifted — as an annual fee, as unit prices, or as an option — so it sits in the model either way. And it is worth the supplier's ability to perform in year twelve, which is why lenders read the credit support behind a CMA before they read its floor.

Go deeper

Capacity maintenance agreement, in context.

The Grid-Scale BESS course covers capacity maintenance agreement — and the rest of the system — from the ground up, the way it actually gets deployed.

Browse the course