Lending to a multi-offtaker solar project: how lenders assess an aggregated portfolio

By SolarQuant Editorial. Published 2026-10-05. Last updated 2026-10-05.

Most term sheet guides assume one offtaker. When a solar project sells to several corporate buyers, or to an aggregator or trader that resells the power, the lender analyses a portfolio of contracts. The model then needs a row per offtaker, a lender's case haircut and a test for losing the largest buyer.

Why does a lender treat several offtakers as a portfolio?

A lender treats several offtakers as a portfolio because the loan is repaid from the pooled revenue of all the contracts. No single buyer's credit decides the loan, so the lender judges the mix: who pays, how much each pays, and for how long.

Public notes describe two structures. Several buyers sign near-identical contracts for shares of one project's output, or one buyer signs a single contract and resells to others (Norton Rose Fulbright). When one buyer defaults, the others keep paying, but buyers will rarely accept liability for each other.

An aggregator or trader changes the question again. A South African bank's note says aggregators usually have no substantial balance sheet, so their business model becomes key. It says aggregators should give financiers insight into both the generation and the offtake portfolios (Nedbank CIB).

This article varies the series example. The 40 MWp plant and the year 1 revenue of USD 5.6m stay the same. The single offtaker is replaced by five invented offtakers.

The lender looks through the project company to five offtakers
Figure 1. Structure of the invented example · one project company, five offtakers, one lender group

In our reading, the lenders take security over every contract and sign a direct agreement with each offtaker. A training paper says a direct agreement gives lenders notice of a project company default and the right to intervene before termination (CLDP). Each offtaker posts its own payment security and is tested as part of the whole.

How do lenders measure the credit quality of an offtaker portfolio?

Lenders grade each offtaker, then weight the grades by each offtaker's share of revenue. The result is one portfolio measure, read alongside the share of revenue at each grade.

A buyers' association describes the practice for aggregated deals. Developers analyse each offtaker and give the group a common credit rating, and some have used a weighted average (CEBA). We found no public formula from a rating agency, so the method below is our reading.

The table is the input sheet for the example. Every figure in it is invented. Volumes add up to 70 GWh a year, which assumes an invented yield of 1,750 kWh per kWp.

Offtaker Volume (GWh a year) Tariff (USD per MWh) Credit grade Contract ends (end of year) Payment security Security (months of invoices)
Offtaker 1 26 75 A 15 Bank guarantee 2
Offtaker 2 17 80 B 10 Letter of credit 3
Offtaker 3 12 84 B 12 Bank guarantee 3
Offtaker 4 8 85 C 7 Cash deposit 4
Offtaker 5 7 86 C 5 Cash deposit 4

Volume times tariff gives year 1 revenue of USD 5.6m, the same as the single-offtaker example. Grades A, B and C are an invented internal scale, scored 1, 2 and 3.

Portfolio score=∑iRevenueiTotal revenue×Scorei\text{Portfolio score} = \sum_{i} \frac{\text{Revenue}_i}{\text{Total revenue}} \times \text{Score}_i

The example. The revenue-weighted score is 1.88, a little better than grade B. Grade A earns 34.8% of revenue, grade B 42.3% and grade C 22.9%.

What is a concentration limit?

A concentration limit caps the share of revenue that may come from one offtaker, one credit grade or one sector. It stops a portfolio from behaving like a single-offtaker deal in disguise.

The example. We found no public source for typical limit levels, so the limits here are invented. Assume a limit of 40% of revenue from any one offtaker and 25% from grade C offtakers. Offtaker 1 earns 34.8% and grade C earns 22.9%, so both tests pass.

What it changes in the model. The limit links to cover ratio headroom. In year 1 the example has cash flow available for debt service (CFADS) of USD 4.2m and debt service of USD 3.231m. Revenue can fall 5.8% before the 1.20x lock-up and 11.5% before the 1.10x default, if costs and tax stay fixed.

So any offtaker with more than 11.5% of revenue can take the project to default level if it leaves unreplaced. That is our reading of why lenders pair a concentration limit with replacement rules. The ratio levels are explained in the article "The cover ratio ladder".

How does the remaining contract life compare with the loan tenor?

The weighted average remaining contract life is the average number of contract years left, weighted by each offtaker's share of revenue. In our reading, lenders compare it with the loan tenor to see whether contracted revenue outlasts the debt.

What the term sheet typically says. A business council guide says contracts for new projects often need a duration that covers at least the debt term (WBCSD). A law firm and advisory paper says a smaller buyer may negotiate a shorter contract if the majority offtaker finances most of the project (Baker McKenzie and Schneider Electric). The bank note on aggregators lists the weighted-average life of offtake agreements among its key measures.

Weighted average remaining life=∑iRevenueiTotal revenue×Remaining yearsi\text{Weighted average remaining life} = \sum_{i} \frac{\text{Revenue}_i}{\text{Total revenue}} \times \text{Remaining years}_i

The example. The five contracts have a weighted average remaining life of 11.2 years. The loan runs for 12 years, and its average life, weighted by principal repaid, is 7.2 years. Which of the two the lender tests against varies by deal.

Contracted revenue falls to 53% before the loan is repaid in year 12
Figure 2. Invented example · five offtakers, year 1 revenue USD 5.6m, 12-year loan

Three contracts end before the loan does, and Offtaker 2 is the one that matters most. An average of 11.2 years looks close to 12, but it hides a drop in the last two years of the loan.

How much of the output must be under contract?

The contracted share is the part of forecast revenue, in a given year, that is covered by a signed contract with an accepted offtaker. Lenders test it year by year, because uncontracted output earns an uncertain price.

What the term sheet typically says. A rating methodology ranks merchant sales as the least favourable revenue for a renewable project (CareEdge Global). A term sheet can therefore set a minimum contracted share, a haircut on uncontracted revenue, or both. We found no public source for typical levels.

The example. Assume an invented minimum contracted share of 70% in every year of the loan. The portfolio is fully contracted in years 1 to 5 and stays above 70% until year 10. In years 11 and 12 it holds 52.8%, so the test fails unless Offtaker 2 renews or is replaced.

What changes when the buyer is an aggregator or trader?

When an aggregator or trader buys the output and resells it, the lender also tests whether the aggregator's own customers need the power. The demand cover ratio is contracted end-customer demand divided by the supply the aggregator has committed to buy.

Demand cover ratio=Contracted end-customer demand (GWh)Supply committed by the aggregator (GWh)\text{Demand cover ratio} = \frac{\text{Contracted end-customer demand (GWh)}}{\text{Supply committed by the aggregator (GWh)}}

The bank note describes the logic. Aggregators usually offer a mix of long and short offtake options, and may hold more demand than supply so that demand always exceeds supply. It does not publish a ratio.

The example. Suppose one aggregator bought all 70 GWh and held end-customer contracts for 84 GWh a year. Demand cover is 1.20x, against an invented minimum of 1.10x. Both the 84 GWh and the 1.10x are invented.

The common mistake. Testing demand cover once. End-customer contracts can be shorter than the aggregator's purchase contract, so the ratio needs its own row for every year of the loan.

What payment security does each offtaker give?

Payment security is a bank guarantee, letter of credit, parent guarantee or cash deposit that the project calls when an offtaker fails to pay. In a multi-offtaker deal each offtaker posts its own. In our reading, the amount follows its credit.

A PPA software provider defines the instruments (Pexapark). Under a bank guarantee or letter of credit, a bank pays if the buyer does not. A parent company guarantee puts a stronger parent behind a weak or unrated buyer, and cash in an escrow account is an alternative.

The buyers' association note says each offtaker in an aggregated deal signs its own contract and posts its own credit, commensurate with its rating. A training paper on power project finance lists escrow accounts and a liquidity letter of credit as forms of liquidity support (CLDP).

The example. The input table sets security at 2 months of invoices for grade A, 3 for grade B and 4 for grade C. That totals USD 1.34m, or 2.9 months of portfolio revenue. The month counts are invented.

What it changes in the model. Security is a backstop, not revenue. It covers unpaid invoices while a contract is ended and the volume is resold. Our reading is that it should span the credit period, 60 days in the example, plus the time needed to terminate.

What happens when an offtaker leaves?

When an offtaker defaults or its contract ends, the project must resell that volume to a replacement the lender accepts. Until it does, the lender treats the volume as uncontracted and tests whether the remaining offtakers can carry the debt.

What the term sheet typically says. Public sources describe the pieces, not a standard clause. A law firm and advisory paper lists a tripartite agreement with each buyer among project financing requirements (Baker McKenzie and Schneider Electric). In our reading, a buyer's right to transfer its share to another buyer also affects bankability.

In our reading, a replacement clause has four parts. They are a time limit to re-contract, a minimum credit grade, a minimum tariff and fresh payment security. The lender approves any replacement that misses one of them.

The example. Remove each offtaker in turn for a full year, with no replacement. Costs and tax are held at the series figures, which is a simplification. A full model recalculates tax, which softens the fall.

Losing the largest offtaker cuts year 1 DSCR from 1.30x to 0.70x
Figure 3. Invented example · year 1 CFADS USD 4.2m, debt service USD 3.231m, one offtaker removed at a time

Losing any one offtaker takes the ratio below the 1.20x lock-up, and only the loss of Offtaker 5 stays above the 1.10x default. Diversification limits the damage, but a loan sized at 1.30x has little room for a year of lost revenue.

If Offtaker 1 is replaced at once at 85% of its tariff, the ratio is 1.21x. With a six-month gap first, it is 0.95x for that year. The 85% and the six months are invented.

What does the lender want reported on the portfolio?

The lender wants a regular portfolio report on each offtaker's volume, payments, credit grade and security, plus the portfolio tests. The bank note on aggregators says they should give financiers insight into the offtake portfolio. We found no public template, so the list below is our reading.

How do you model it?

Model a multi-offtaker project with one input row per offtaker and one set of portfolio tests per period. The debt, reserve and cover ratio sheets stay as described in "How to read a solar project finance term sheet: a modeller's guide".

  1. Build the offtaker table. One row each for volume, tariff, credit grade, contract end date and payment security. The example uses annual periods. Real deals usually use six-month periods.
  2. Compute revenue per offtaker. Volume times tariff, switched off after the contract end date.
  3. Add an uncontracted row. It picks up volume freed when a contract ends.
  4. Apply the lender's case haircut. The example uses invented haircuts of 10% on grade C revenue and 20% on revenue after a contract ends.
  5. Run the portfolio tests each period. Largest share, share by grade, weighted average remaining life, contracted share and demand cover.
  6. Size or test the debt on haircut CFADS. Hold costs and tax fixed for a first pass.
  7. Run the loss test. Remove each offtaker in turn, with a replacement gap and a discounted tariff.
  8. Keep payment security outside CFADS. It is drawn only against unpaid invoices.

The example. The haircut is 2.3% of revenue in years 1 to 5 and rises to 9.4% in years 11 and 12. If the debt stays at USD 25.1m, the ratio is 1.26x in years 1 to 5 and 1.14x in years 11 and 12. That is under the 1.20x lock-up.

Resizing at 1.30x on the haircut CFADS gives debt of USD 23.8m. That is USD 1.25m less than the single-offtaker case, and gearing falls from 62.7% to 59.6%.

What are the common mistakes?

Most of these mistakes model the portfolio as one average buyer, when it is a set of separate contracts that change over time.

Frequently asked questions

Is a multi-offtaker project easier to finance than a single-offtaker one?

Not automatically. Several buyers spread default risk, and one published view is that lenders may like that diversification if little or no revenue comes from weak credits. The lender also takes on more analysis, shorter contracts and replacement risk.

Do the offtakers guarantee each other?

Usually not. The law firm note we opened says buyers will rarely agree to assume liability for other buyers. Each contract stands alone, so the model should treat each offtaker's default separately.

What is a demand cover ratio?

A demand cover ratio compares an aggregator's contracted end-customer demand with the supply it has committed to buy. A ratio above 1.00x means customers have contracted for more power than the aggregator must take. The 1.10x minimum in this article is invented.

Why does the loss of one small offtaker breach lock-up?

Because the debt is sized with a 1.30x ratio, which leaves limited headroom. In the example, year 1 revenue can fall only 5.8% before the ratio reaches 1.20x. The smallest offtaker earns 10.8% of revenue.

Sources

Pages opened on 4 October 2026 and checked again on 5 October 2026. The example project is invented and has no source.