Availability Guarantees: Two Documents, Two Definitions, One Fee-Sized Remedy

A solar project usually carries two availability guarantees. The seller gives one to the offtaker in the PPA. The O&M contractor gives one to the owner in the service agreement. They are different numbers, measured on different bases, at different points in the plant, with different exclusions — and the sponsor sits between them, owing on one and collecting on the other.

The one that pays is capped at the O&M contractor's annual fee. On a utility-scale plant that ceiling is roughly a tenth of the annual revenue the guarantee exists to protect. It is the smallest remedy in the contract stack and it sits against one of the more probable failure modes.

This is the fifth post in the series and the second on performance. The previous one showed that a PR test excludes downtime by design, which means availability has to be caught somewhere else. This is where. You will get the two definitions and why they do not reconcile, what the exclusions actually remove, why time-based measurement understates the revenue loss, and the arithmetic on a fee-capped remedy.

ℹ️ Note: This describes how these mechanisms work in practice. It is not legal or investment advice — definitions, exclusions and caps vary substantially between contracts and markets.

What Does "Availability" Actually Mean?

The proportion of time the plant was capable of operating, after removing hours nobody is holding the operator responsible for. It measures readiness, not output — a plant can be 100% available all night.

That distinction is the source of most of the confusion. Availability answers "was the equipment able to run?" Performance ratio answers "how efficiently did it convert what it received?" Energy answers "what did it actually produce?" Three questions, three metrics, and only the third one pays the debt.

The previous post covered why PR cannot catch downtime: the guaranteed PR is adjusted to exclude periods of non-production, precisely so the contractor is not penalised for an outage it did not cause. That exclusion is technically correct and it leaves a hole. Availability guarantees exist to fill it.

Why Are There Two Availability Guarantees, and Why Don't They Match?

Because they are solving different problems for different parties, and neither was drafted with the other in view.

	PPA availability guarantee	O&M availability guarantee
Given by	Seller (the project)	O&M contractor
Given to	Offtaker	Owner
Typical level	90–95% mechanical availability	99%+ over a year
Measured at	Plant level	Inverter level
Basis	Time-based (mechanical) for corporate offtakers; output-based for utilities	Time-based, calculated annually
Remedy	Set amount per percentage of shortfall, or converted to an output shortfall	Lost revenue reimbursed, capped at the annual fee

On the PPA side, Stoel Rives' guide to utility-scale PPAs notes that "mechanical-availability percentages usually range from 90 to 95 percent, but they may have a lower percentage in the first measurement period," and that mechanical-availability guarantees "are standard in PPAs where the offtaker is a corporate or industrial user," having become rarer for utilities with RPS obligations to satisfy.

On the O&M side, Solar Best Practices puts the benchmark considerably higher: "A best practice is a Minimum Guaranteed Contractual Availability of 99%+ over a year at least at inverter level," with availability "calculated at inverter level, on an annual basis."

At first glance that is comfortable — you owe the offtaker at 90–95% and you are owed by the operator at 99%. A four-to-nine point buffer.

The buffer is less clean than it looks, because the two numbers are not measuring the same object. Inverter-level availability asks whether each inverter was capable of running, weighted across the fleet. Plant-level mechanical availability asks whether the facility could deliver. A plant with every inverter healthy but a failed main transformer, a tripped breaker or an out-of-service collector circuit can report excellent inverter availability and zero plant availability. The measurement point sits upstream of some of the equipment most likely to take the whole site down.

ℹ️ Note: When diligencing the pair, check the measurement point before the percentage. A 99% guarantee measured at a point that excludes the high-voltage equipment is a weaker instrument than a 97% guarantee measured at the meter.

What Hours Don't Count?

The exclusions, not the headline percentage, are where an availability guarantee is actually negotiated. Stoel Rives lists the standard three: an availability guarantee applies "after excluding hours lost to force majeure, curtailments, and a certain amount of scheduled maintenance and other excused non-available hours."

Each of those does real work.

Force majeure removes events outside anyone's control, and its definition should be checked against the EPC and PPA definitions — asymmetry between documents was the trap in post 3 and it recurs here.

Curtailment is the significant one for a merchant-exposed or basis-exposed plant. Hours the grid refused to take are excluded from the operator's obligation, correctly, since the operator did not cause them. But those hours are still lost revenue, and they are now the responsibility of nobody in the O&M chain. Whether the offtaker pays for curtailed output is a PPA question — deemed generation — and a separate post in this series.

Scheduled maintenance is excluded only up to an agreed allowance. The size of that allowance, and crucially when it may be taken, is worth as much as a point of guaranteed availability. Maintenance scheduled into December costs a fraction of maintenance scheduled into June.

The residue after those exclusions is what the operator is actually on the hook for: unplanned outages within its control. That is a much smaller slice of total downtime than the guarantee's headline suggests.

Why Does Time-Based Availability Understate the Loss?

Because availability counts hours and revenue counts megawatt-hours, and an hour of solar production in June is worth several times an hour in December. Time-based measurement treats them identically.

The consequence is that identical availability figures can conceal very different revenue outcomes. A 2% annual outage spread evenly across the year costs roughly 2% of production. The same 2% concentrated into high-irradiance midday hours in peak season costs substantially more — and reports as the same number.

This is not a theoretical concern. Outage duration correlates with the difficulty of the repair, and inverter or transformer failures under thermal stress are more likely in the hot, sunny periods that also carry the most revenue. The measurement basis systematically flatters exactly the scenario that hurts most.

There is a well-established fix — weighting availability by expected generation rather than by clock time, so an unavailable hour is scored by the energy it would have produced. Some O&M agreements adopt it; Solar Best Practices' benchmark guidance is framed on a time basis calculated annually at inverter level. Where the agreement is time-based, the exposure is still modellable, and it belongs in the base case as a distinct line rather than as an assumed equivalence.

How Big Is the O&M Remedy, Really?

Capped at the annual fee. Solar Best Practices is explicit on both the recovery and the ceiling: "if the Minimum Guaranteed Availability is less than the measured availability, all the revenue lost due to the availability shortfall should be reimbursed to the Asset Owner by the O&M service provider" — and "the amount of yearly LDs should be capped at 100% of the O&M annual fee," with reaching that threshold typically triggering termination rights for the owner.

Read those two sentences together. The recovery principle is generous: all the lost revenue. The ceiling is the operator's fee for the year. On a utility-scale plant those are not remotely the same order of magnitude — annual O&M cost is a small single-digit percentage of annual revenue, so the guarantee makes you whole right up to a limit set by roughly a tenth of what it is protecting.

That is not a drafting failure either. An O&M contractor earning a fee cannot underwrite the asset's entire revenue; no operator would sign it and none could fund it. But it does mean the instrument should be understood for what it is: a performance incentive with a modest financial backstop, not revenue insurance.

The termination right attached to cap exhaustion is arguably worth more than the money. Reaching the cap means the operator has failed badly enough to owe a full year's fee, and the value of being able to replace them exceeds the value of the final dollar of damages.

How Do the Bonus and Penalty Bands Work?

Symmetrically, and with netting. Where availability exceeds the guarantee, Solar Best Practices describes the upside sharing: "if the measured availability exceeds the Minimum Guaranteed Availability, the additional revenue will be divided between the Asset Owner and the O&M service provider per previously agreed shares." And critically: "bonuses can be offset against LDs and vice versa."

The netting is the part to model. An operator that has a poor first half and an excellent second half may end the year owing very little, because the bonus periods absorb the penalty periods. Whether that is the right outcome depends on when the downtime fell — and the netting mechanism, being time-based, does not know that a bad June and a good November are not equivalent.

Two drafting details determine how much the band is worth: the measurement period over which bonus and penalty are netted, and whether the sharing percentages are symmetric. An annual netting period with a generous upside share is a materially weaker instrument than quarterly netting with a modest one.

What Does Replacing the Operator Actually Cost?

More than the damages you just collected, which is why the termination right at cap exhaustion is usually more valuable as leverage than as an action.

Cap exhaustion means the operator owes a full year's fee and you may terminate. Before exercising that, three things need checking, and none of them appear in the O&M agreement's damages clause.

The transition is a coverage gap. A new operator mobilises into a site it has not maintained, with a backlog it did not create and spares it did not order. Availability during the handover is typically worse than the availability that triggered the termination, and the incoming operator will not guarantee a period it cannot control.

Equipment warranties may be conditioned on the operator. Module and inverter warranties frequently require maintenance to be performed to a defined standard, sometimes by an approved provider, with records to prove it. Switching operators without confirming those conditions can impair the warranties that sit behind the very equipment that failed. Confirm the warranty terms before serving notice, not after.

Lenders have a say. The O&M agreement is normally a material project agreement, and as covered in post 3, material project agreements sit inside the credit agreement's covenant package — Skadden's analysis describes material breach of such agreements as "universally an event of default under a credit agreement." Replacement generally needs lender consent and often a direct agreement with the incoming operator. That is a process measured in months, running concurrently with a plant that is already underperforming.

The practical consequence: price the threat, not the remedy. An operator facing cap exhaustion and a credible replacement has strong incentives to fix the plant, and that behavioural effect is worth more than the final tranche of liquidated damages.

What Availability Does the Lender Assume?

A lower number than the one in the contract, and the gap is worth real debt capacity.

A lender's base case does not take the contractual guarantee as an input. It takes a view on what the plant will actually achieve, informed by the independent engineer, the technology's operating record and the exclusions this post has already described. A 99% contractual guarantee frequently becomes a 97% assumption in the banking case.

Two percentage points sounds immaterial and is not, for the reason every haircut in this category is larger than it looks: operating costs are fixed. Two points off revenue is more than two points off cash available for debt service, because the O&M fee, the insurance, the land rent and the administration costs are unchanged. On a project with a 10% operating cost ratio, a 2% revenue reduction is a 2.2% CFADS reduction, and debt capacity moves with CFADS.

The availability haircut is one of four that typically separate a sponsor's base case from the case the debt is actually sized on — the others being the generation basis, curtailment and the capture rate. Individually each is arguable. Together they routinely remove a fifth of revenue and rather more of the debt.

Three consequences worth planning for.

The guarantee does not close the gap. A sponsor arguing that its 99% contractual guarantee should support a 99% assumption is arguing that a remedy capped at the annual O&M fee is equivalent to the availability itself. It is not, and the cap is precisely why the lender discounts it — a guarantee backed by a tenth of what it protects is an incentive, not a substitute for performance.

The exclusions are the negotiation. Since the lender's assumption is driven by expected actual availability, narrowing the exclusion list does more to move the banking case than raising the guaranteed percentage. An operator guaranteeing 98% with tight exclusions is offering a more bankable instrument than one guaranteeing 99% with generous ones, and only the second number appears on the term sheet.

Contracted availability belongs in the base case, expected availability in the banking case, and the difference belongs in a bridge. Where those two numbers sit in the same cell, nobody can see what the assumption is doing — and it is doing more than most of the terms being argued over.

Size the shortfall on an energy basis, compare it to the fee cap, and show what sits above the cap. Three blocks.

The inputs

All assumptions, labelled as such:

P50 annual energy                    250,000 MWh
Average realised price               $45.00 /MWh
Annual revenue                       $11,250,000
Guaranteed availability              99.0%
Measured availability (time-based)   96.5%
Annual O&M fee                       $1,200,000
Peak-season weighting factor         1.8x

The naive number and the real one

Availability_Shortfall   = 99.0% − 96.5%                     = 2.5 pp

Time_Based_Energy_Loss   = 250,000 × 2.5%                     = 6,250 MWh
Time_Based_Revenue_Loss  = 6,250 × 45                         = $281,250

Energy_Weighted_Loss     = 6,250 × 1.8                        = 11,250 MWh
Energy_Weighted_Revenue  = 11,250 × 45                        = $506,250

The weighting factor is the ratio of expected generation during the outage hours to average generation. It is derivable from the plant's own production profile and the outage timestamps — it is not a guess, though it should be labelled as an assumption until the outage data exists.

The cap test

LD_Cap              = Annual_O&M_Fee                          = $1,200,000
Recovered           = MIN(Energy_Weighted_Revenue, LD_Cap)    = $506,250
Uncovered           = Energy_Weighted_Revenue − Recovered     = $0

Cap_As_%_of_Revenue = 1,200,000 / 11,250,000                  = 10.7%

Covered here. But note the last line: the ceiling is 10.7% of annual revenue. Any availability event costing more than that is uncovered by construction, whatever the guarantee says.

Where it breaks

Severe_Shortfall         = 8.0 pp
Energy_Weighted_Loss     = 250,000 × 8% × 1.8 × 45            = $1,620,000
Recovered                = MIN(1,620,000, 1,200,000)          = $1,200,000
Uncovered                = $420,000   → equity, plus a termination decision

A little over eight points of availability exhausts the entire annual remedy. That is the number worth knowing before signature, because it tells you how much of the availability risk was ever transferable.

ℹ️ Note: Model bonus netting explicitly if the agreement allows it. An annual netting period can reduce a genuine mid-year failure to nearly nothing, which is fine commercially and misleading if the model reports the net figure as the exposure.

To run the full version — availability weighted by your own hourly generation profile, outage timestamps mapped against irradiance, and the bonus/penalty band netted on the contract's actual measurement period — prompt Dezzmond with your O&M terms and production data and it will build the schedule.

What Do Lenders Actually Check?

Lenders treat the availability guarantee as a small credit against an operating risk, and they size the rest into reserves.

• Where is availability measured? Inverter level excludes the equipment most likely to take the whole site offline.
• Is it time-based or generation-weighted? Time-based understates revenue loss whenever downtime correlates with irradiance.
• What is the cap as a percentage of annual revenue? The honest measure of how much risk actually transferred.
• What are the exclusions, and how large is the scheduled maintenance allowance? Including whether maintenance windows can be taken in peak season.
• Do the PPA and O&M definitions reconcile? The project owes on one basis and collects on another.
• How is bonus/penalty netted? A long netting period mutes real failures.
• What happens at cap exhaustion? The termination right, and whether a replacement operator is realistically available.

Frequently Asked Questions

What is a typical availability guarantee in a solar O&M contract?

Best practice guidance puts minimum guaranteed contractual availability at 99%+ over a year, measured at least at inverter level and calculated annually. PPA-level mechanical availability guarantees given to a corporate offtaker typically sit lower, around 90–95%.

Why is availability measured at inverter level rather than at the meter?

It isolates equipment the operator controls and maintains. The trade-off is that failures upstream of the inverters — transformers, breakers, collector circuits — can take the site offline without registering fully in the measured figure.

Are curtailed hours counted against the operator?

Normally no. Curtailment sits with force majeure and scheduled maintenance in the standard exclusions, because the operator did not cause it. The revenue is still lost; whether anyone pays for it is a PPA question rather than an O&M one.

What is the cap on O&M liquidated damages?

Commonly 100% of the annual O&M fee, with reaching the cap triggering owner termination rights. Because the annual fee is a small fraction of annual revenue, the cap rather than the guarantee determines how much risk actually transferred.

Does a lender use the contractual availability guarantee?

No. It takes a view on expected actual availability, commonly around 97% against a 99% guarantee, because a remedy capped at the annual O&M fee is an incentive rather than a substitute for performance. Narrowing the exclusions moves that assumption more than raising the guaranteed percentage does.

Can bonuses cancel out liquidated damages?

Yes where the contract provides for netting, and it is common. The measurement period matters: annual netting lets strong months absorb a serious failure in a weak one, which can leave a materially underperforming year showing almost no damages.

Closing: A Guarantee Sized to the Fee, Not to the Risk

Availability is the guarantee that looks most reassuring and transfers the least. The headline number is high — 99% against a PPA obligation of 90–95% — and the recovery principle is generous, promising all revenue lost to a shortfall. Then the cap arrives and sets the whole instrument at one year of the operator's fee.

That is the right answer commercially. An operator cannot insure an asset's revenue out of a service margin. But it means availability risk is mostly retained, not transferred, and it should sit in the model next to reserve sizing and DSCR headroom rather than in the contracts summary as a solved item.

Across this series the pattern has held four times now. The completion dates sit in two documents and drift. The delay LD rate is negotiated rather than derived. The cap exhausts before the longstop arrives. Capacity and PR are guaranteed while annual energy is not. And here, the remedy is capped at the fee rather than at the loss. In every case the contract is doing exactly what it says — the exposure lives in what sits between the documents, and only the model sees all of them at once.

Next in the series: change in law, and who carries tariff and policy risk when the rules move after financial close.

Sources: Stoel Rives — The Law of Solar: Utility-Scale Power Purchase Agreements · Solar Best Practices — Contractual Framework · Solar Best Practices — System Commissioning · Skadden — Lenders' Relationships with Project Counterparties