Availability Guarantees: What the O&M Contractor Is Actually On the Hook For
A solar O&M agreement says the contractor guarantees 98% or 99% availability.
That sounds excellent.
Then the main transformer fails, the plant is down, and the owner discovers the outage is outside the measured availability definition or the damages are capped at a fraction of the lost revenue.
The percentage was never the whole guarantee.
For finance, availability is a much more useful concept when it is reduced to three questions:
What equipment is measured? Which downtime counts? What happens financially when the guarantee is missed?
Availability and performance ratio answer different questions
DOE/FEMP separates the metrics cleanly.
Performance ratio compares actual output with model output using the measured resource and temperature conditions.
Availability asks whether the system was operational and capable of producing when the resource and grid were available.
So:
PR problem:
Plant was available, but converted the resource poorly.
Availability problem:
Plant could not operate when it should have been able to.
A plant can have a good PR and poor availability.
If it performs beautifully during the hours it is online and spends too many hours offline, PR can look fine while annual energy suffers.
That is why EPC performance testing does not close the operating-risk discussion.
The first part of this series ends where availability begins.
Read the exclusions before the percentage
First Solar's public filings describe its O&M availability guarantee in unusually plain language.
The guarantee is that the system will be available to generate a specified percentage of possible energy after adjusting for factors outside First Solar's control, including weather, curtailment, outages, force majeure and other conditions. The filing explicitly says this is not a guarantee that the system will actually generate a specific amount of energy; it is a guarantee that the system will be operational for factors within the service provider's control.
That is the right way to think about the instrument.
An O&M contractor can reasonably take responsibility for:
- response and repair of covered equipment;
- preventive maintenance;
- monitoring;
- operator-caused downtime;
- failures within the agreed scope.
It cannot sensibly guarantee the sun, grid dispatch, force majeure or every owner-controlled item.
So the percentage is only meaningful after the excluded downtime is understood.
I would summarize the formula as:
Measured availability
=
available time
/
(total time - excluded time)
The negotiation is often in the denominator.
Availability can be measured by hours or by lost energy
Not every contract uses the same formula.
DOE/FEMP presents availability as a time-series concept: operational time relative to total time, after the relevant conditions are defined.
First Solar has publicly described effective availability using an energy-based formulation:
MWh Generated
/
(MWh Generated + Adjusted MWh Lost)
with lost MWh adjusted for factors outside the service provider's control.
Those approaches can produce different results.
A plant-level energy-weighted metric naturally gives more weight to an outage during a high-production period.
A pure time metric treats each unavailable hour equally unless the agreement adds weighting.
So the diligence summary should capture the exact formula, not merely the percentage.
Measurement point matters more than 1%
A plant-level guarantee and an inverter-level guarantee are not the same thing.
Suppose all inverters are healthy but the main transformer fails.
At the POI, the plant is unavailable.
At an inverter-level measurement point, much of the equipment being measured may still be "available" even though the project cannot deliver a single MWh.
This is why the diligence question should not be:
Is the guarantee 98% or 99%?
It should be:
98% or 99% of what?
The contract or technical exhibit should identify:
- equipment included;
- measurement point;
- partial-availability treatment;
- data source;
- treatment of communication failures;
- exclusions.
A lower percentage measured across the full delivery chain can protect more economic value than a higher percentage measured across a narrower equipment scope.
Time availability can hide the revenue impact
A one-hour outage at 2 a.m. and a one-hour outage at noon are both one unavailable hour.
For solar revenue they are not remotely equivalent.
That creates a weakness in simple time-based availability metrics.
Assume the project is unavailable for 100 hours in the year.
If the hours are concentrated at night or low-irradiance periods, the energy impact may be small.
If they are concentrated in high-irradiance summer afternoons, the lost MWh can be much larger.
For the finance model, I would calculate two numbers:
Contract availability
Energy-weighted lost production
The contract decides whether the O&M provider owes damages.
The second number tells you what the owner actually lost.
Do not assume the first is a good proxy for the second.
Data quality can decide whether the guarantee can be enforced
Availability is calculated from operating data.
That raises a practical contract question: which data source controls when systems disagree?
Possible sources include:
- SCADA;
- inverter logs;
- revenue meter;
- weather station;
- utility outage logs;
- operator tickets;
- third-party monitoring platform.
If SCADA says the inverter was unavailable and the operator's local log says it was available, the damages calculation needs a hierarchy.
The O&M agreement should address data retention, reporting, meter / sensor failure and dispute resolution.
For a lender, this sounds operational until there is a large claim.
Then it becomes collectability.
A guarantee that cannot be measured cleanly is weaker than the headline percentage suggests.
Partial availability needs an equipment-weighting rule
A 100 MW plant does not jump only between 100% available and 0% available.
If one 5 MW inverter block is unavailable, the project is partially available.
The agreement therefore needs a weighting method.
A simple capacity-weighted approach might treat a 5 MW outage for 10 hours as:
5 MW / 100 MW × 10 hours
= 0.5 equivalent full-plant unavailable hours
An energy-based method can weight the same outage by expected production during those hours.
The choice matters most in a plant with:
- uneven inverter block sizes;
- DC overbuild;
- clipping;
- different irradiance across the site;
- tracker or collector outages affecting multiple blocks.
For diligence, "availability measured at inverter level" is still incomplete until the aggregation formula is known.
What the owner still keeps
Consider a project with these outage hours:
Inverter failure 45 hours
Grid curtailment 70 hours
Transmission outage 30 hours
Scheduled maintenance 20 hours
Owner communication outage 10 hours
The O&M agreement may count only part of that 175-hour total against the provider.
If curtailment, grid outages, some scheduled maintenance and owner-caused events are excluded, the contractual availability can remain high while project energy is materially below plan.
That is not necessarily poor drafting.
It is risk allocation.
The owner still needs to model the excluded bucket somewhere else.
A simple bridge is:
| Lost-energy cause | O&M guarantee? | Other protection? |
|---|---|---|
| Operator-controlled equipment downtime | Often yes | O&M damages / cure |
| Grid outage | Usually excluded | Possibly none / IA remedies |
| Curtailment | Usually excluded | PPA deemed generation if applicable |
| Resource | Not availability | Yield risk |
| Owner-caused outage | Excluded | Owner |
| Force majeure | Usually excluded | Insurance / retained risk depending event |
This is the coverage map I would want before financial close.
Bonuses and annual netting can hide when the bad hours occurred
Some O&M structures include upside payments for exceeding an availability threshold and allow bonus and penalty positions to net over the measurement period.
That can be commercially reasonable and economically misleading if the project model only looks at the annual net result.
A strong autumn can offset a poor summer in the contractual calculation.
The project cannot recover the summer revenue later.
So I would keep:
Contract annual availability result
Contract LD / bonus net amount
Actual lost MWh by month
Actual lost revenue by month
separate.
The contract settles the operator relationship.
The monthly lost-energy schedule explains the cash-flow and covenant consequence.
The damages cap is the real ceiling
An availability guarantee can promise compensation for lost production and still transfer only a modest amount of economic risk.
The reason is the cap.
Public filings from solar operators make the commercial logic clear. Availability guarantees sit inside an O&M service business, and underperformance exposure can be limited by contractual remedies. First Solar disclosed that its O&M agreements typically limited total potential losses, including availability LDs, to a percentage of O&M fees.
That makes sense.
An O&M contractor earning a $1 million annual fee cannot economically insure a $20 million annual revenue stream without charging for that risk.
So I would compare:
Expected revenue loss from severe availability event
vs
maximum O&M damages
For example:
Annual project revenue $18m
Severe covered outage revenue loss $2m
Annual O&M fee $1.1m
Availability damages cap $1.1m
Residual owner exposure $0.9m
The guarantee is valuable.
It is not revenue insurance.
That distinction should be visible in the banking case.
A separate SEC-filed Ameresco / SCE structure shows a different calibration: a 97% availability requirement, LDs equal to 0.2% of the applicable EPC price for each whole percentage point below the threshold, a 3% cap for each 12-month period, and a 20% aggregate cap during the agreement. The point is not that either formula is “market.” It is that the percentage, LD formula and cap have to be read together before the guarantee can be translated into a bankable recovery.
Major-component responsibility is part of the availability credit
Two O&M agreements with the same 99% guarantee can transfer different amounts of risk if one includes major-component replacement and the other does not.
The operating scope should identify responsibility for:
- inverter replacement;
- transformer repair;
- tracker systems;
- HV equipment;
- spare parts;
- warranty claims;
- OEM coordination;
- labor outside routine maintenance.
If the operator only diagnoses the failure while the owner funds the transformer and waits for the OEM, the annual availability guarantee may have exclusions or caps that leave most of the economic exposure with the owner.
This is why the scope matrix belongs beside the percentage.
Scheduled maintenance should have a calendar, not only an annual allowance
An O&M agreement can exclude a stated number of scheduled-maintenance hours.
The economic cost depends on when those hours occur.
For solar:
20 hours in low-resource winter periods
≠
20 hours in high-resource summer afternoons
For storage, the same issue appears around high-price or contracted availability windows.
I would therefore schedule planned maintenance in the operating model rather than simply applying an annual availability percentage.
This also gives the lender a better basis for deciding whether the sponsor's expected availability is consistent with the maintenance plan.
Contract guarantee versus banking assumption
A project can have a 99% contractual availability guarantee and still be underwritten by lenders at a lower expected availability.
There is no contradiction.
The guarantee is a contractual floor backed by a remedy.
The banking case is an estimate of what the project will actually achieve.
If the remedy is capped, excluded events are material, or the operator's credit is limited, the lender cannot treat the guarantee as equivalent to physical performance.
I would keep two separate model inputs:
Contractual guaranteed availability
Banking-case expected availability
Then bridge the difference.
If the sponsor wants the lender to move the banking assumption upward, the most useful evidence is not a higher headline percentage in the O&M agreement.
It is:
- tight exclusions;
- broad equipment coverage;
- strong historical fleet data;
- credible response-time obligations;
- real spare-parts strategy;
- meaningful uncapped or well-secured remedies;
- strong operator credit.
Those are the things that make 99% believable.
Spares turn a service-level promise into a physical response
A four-hour response commitment does not fix a failed inverter if the replacement power stack is six weeks away.
The O&M plan should therefore connect response obligations to critical-spares strategy:
Failure mode
Critical spare required
Spare on site / regional / OEM
Lead time
Who funds inventory
Who owns obsolescence risk
Maximum restoration time
This matters more as equipment ages and OEM support changes.
For lender diligence, the operating model should distinguish "technician response" from "asset restoration."
Only the second one restores CFADS.
Repeated short outages can be worse than one long outage for some revenue streams
Availability metrics compress the year into one number.
The market can care about the pattern.
A plant with 98% availability generated by one seven-day outage has a different operating profile from a plant with dozens of short trips.
Repeated trips can create:
- lost peak-price hours;
- imbalance exposure;
- repeated restart risk;
- warranty concerns;
- higher operator workload;
- more uncertainty for contracted availability obligations.
The same annual percentage can therefore support different banking assumptions.
If the revenue stack is sensitive to specific operating windows, the lender should review outage distribution as well as annual availability.
Response time matters too
Availability is an annual metric.
Operations happen event by event.
A contract that guarantees 99% annually but has weak response and repair obligations can still create long individual outages.
For debt service, the timing can matter.
A 10-day outage concentrated in a low-revenue month may be manageable.
The same outage during a high-price or high-resource period can create a covenant problem even if annual availability eventually recovers.
So I would read the service-level mechanics alongside the annual guarantee:
Remote response time
Site mobilization time
Critical spare availability
Repair time commitment
Escalation process
Major-component replacement responsibility
The annual guarantee tells you the year-end score.
Those provisions tell you how the operator gets there.
Replacing the operator is not a clean remedy
If the provider repeatedly misses the guarantee, termination and replacement may eventually be available.
That is useful leverage.
It is not a zero-cost solution.
A new operator inherits:
- unfamiliar plant history;
- open work orders;
- warranty interfaces;
- spare-parts position;
- data / SCADA handover;
- potentially disputed root causes.
Financed projects can also require lender consent to replace the O&M contractor or amend a material O&M agreement.
So the decision is not:
bad operator → terminate
It is:
cost of keeping current operator
vs
transition cost + expected performance of replacement
The damages cap can be less valuable than the leverage created by a credible replacement right.
The severe case should test the cap and the covenant together
Assume:
Annual P50 revenue $18.0m
Annual O&M fee $1.1m
Availability damages cap $1.1m
Severe covered outage loss $2.0m
Banking DSCR before outage 1.35x
If the operator pays the full $1.1 million cap, the project still retains $0.9 million of operating loss.
That residual can be enough to move a marginal debt-service period below lock-up or default thresholds.
So the severe availability case should show:
Lost revenue
+ O&M damages received
= net CFADS reduction
→ DSCR after claim
A guarantee can pay exactly as drafted and still fail to protect the covenant.
Availability, warranty and insurance overlap without fully replacing one another
A major component failure can touch three contracts at once.
The O&M provider may be responsible for diagnosis and restoration.
The OEM warranty may fund replacement equipment.
Property-damage or machinery-breakdown insurance may respond to a covered physical event.
Those protections do not have identical waiting periods, deductibles, exclusions or cash timing.
For a severe transformer or inverter failure, the owner can therefore face:
lost revenue
+ insurance deductible
+ uninsured waiting period
+ O&M damages cap
+ warranty logistics
+ temporary replacement / mobilization cost
The availability guarantee is one recovery stream inside that stack.
It should not be modelled as the entire response to equipment failure.
Put the operating risk on one page
For diligence, I would put this on one page:
GUARANTEE
Guaranteed availability:
Measurement point:
Measurement period:
Included equipment:
Excluded downtime:
Scheduled-maintenance allowance:
SERVICE
Response times:
Repair obligations:
Critical spares:
Major-component responsibility:
REMEDY
LD / shortfall formula:
Annual cap:
Aggregate cap:
Security:
Termination threshold:
MODEL
Sponsor availability:
Banking availability:
Revenue impact of exclusions:
Revenue impact above damages cap:
That separates the legal instrument from the operating assumption.
And it prevents a sentence like "99% availability guaranteed" from doing more work in the credit memo than the contract can support.
That completes the short performance-guarantee path: EPC acceptance tests tell you whether the completed plant met the contractual construction-performance standard; O&M availability tells you how much operating uptime the service provider actually stands behind.
Neither one guarantees the project's full revenue case.
Sources: DOE/FEMP — Optimizing Solar PV Performance for Longevity · DOE/FEMP — Operate and Maintain an Existing PV System · SEC — First Solar availability-guarantee disclosure and measurement formula · SEC — Ameresco / SCE availability and capacity guarantee disclosure · SEC — Ameresco EPCM performance-guarantee exhibit