Two sponsors, same sector, same quarter, same lender. Both want roughly the same money.
Sponsor A brings a memorandum of understanding with a company everyone in the room recognises. It is signed. It records an intention to purchase “up to” a stated volume at “market-related prices”, subject to definitive agreements.
Sponsor B brings a term sheet with a mid-tier industrial buyer nobody outside the sector has heard of. It is unsigned. It fixes an annual volume, a price formula tied to a published index with a floor, a seven-year term, and what the buyer owes if it fails to take delivery.
Sponsor A believes he holds the stronger document. He holds the stronger name.
Sponsor B gets the term sheet from the lender.
This outcome surprises almost every sponsor it happens to, and the surprise is expensive. It sends them back into the market to do the one thing that will not help: find a better counterparty. Months later they return with a more impressive logo attached to the same unmodellable sentence, and receive the same haircut, for the same reason, which nobody has yet explained to them.
Conventional practice misreads what the lender is doing
The market treats offtake risk as counterparty risk. On that reading the remedy is obvious — upgrade the counterparty. Get the supermajor. Get the sovereign. Get it signed.
Both signature and brand are proxies. Sponsors reach for them because they are legible: you can tell at a glance whether a document is signed and whose logo sits on it. What neither proxy tells you is the thing the credit committee is actually trying to establish, which is whether the project’s revenue can be written down as a schedule.
A credit model is not a judgement. It is an arithmetic object with fields in it. The analyst opens the offtake document to populate those fields. If the document does not populate them, the analyst does not populate them with optimism; the analyst populates them with the merchant case — the price the project would get selling into the open market with no contract at all. That is the haircut. It is not a penalty for a weak counterparty. It is the arithmetic consequence of a document that said nothing the model could hold.
The Offtake Matrix
Two variables decide whether an offtake supports debt, and sponsors consistently optimise the wrong one.
| Counterparty credit | Vague obligations | Fixed obligations |
|---|---|---|
| Strong counterparty credit | The comfort trap Feels safest. Models as nothing. | Bankable Debt sized against contracted revenue. |
| Weak counterparty credit | Not a document Nothing to price, nothing to cover. | Financeable with credit support LC, escrow, guarantee, substitution. |
The diagonal is uninteresting. The two off-diagonal boxes carry the whole argument, and they run opposite to intuition. The top-left box — an impeccable name attached to loose language — feels safest to the sponsor and models as nothing. The bottom-right box — a modest name attached to firm obligations — is a financing problem with known solutions.
A credit model has fields. A letter of intent fills none of them
Four fields decide whether a revenue line can be contracted rather than assumed.
Volume. How much must the buyer take, and what happens to the obligation if the buyer’s own demand falls away? The World Bank’s model power purchase agreement does not leave this to good faith: the purchaser is bound to take “all of the capacity available at any time from the Facility at the Point of Delivery,” and the seller’s exclusivity is stated as a right, not an expectation.
Price. Not a price — a formula. An index, a reference period, an escalation mechanic, and where the sponsor can get it, a floor. “Market-related” is not a price formula. It is a placeholder for a negotiation that has not happened.
Tenor. The World Bank’s model agreement runs its term as a defined number of years from a defined commercial operation date, because lenders size debt against a contracted period, not a hoped-for relationship. A contract shorter than the loan leaves a refinancing risk the sponsor will be asked to carry.
Remedy. What is owed when the obligation is not met. The model agreement specifies liquidated damages for capacity shortfall, capacity-payment reductions that escalate month by month for availability failures, and a termination-payment table that names what the lenders are owed at termination. Nigeria’s own regulator does the same thing on the gas side: the Domestic Gas Delivery Regulations 2022 price non-delivery of allocated volumes at US$3.50 per MMBtu. Whatever one thinks of the number, it is a number. It can go in a model.
Volume, price, tenor, remedy. Take your own offtake document and try to fill those four fields from it. Most sponsors, doing this honestly for the first time, fill one.
Weak credit has mitigants. Missing obligations do not
This is the part that changes what a sponsor should do on Monday.
Counterparty credit weakness is a solved problem in project finance. The toolkit is standard and the World Bank’s model agreement carries most of it: a standby letter of credit from a bank of specified rating, replenished within a set number of business days of any draw; development security posted on a defined schedule; escrow arrangements that release on stated conditions without either party’s further consent; parent guarantees; and, where the buyer genuinely cannot be supported, substitution of the offtaker.
Every one of those instruments does the same job. It sits behind an obligation and makes the obligation more likely to be honoured.
None of them can be applied to an obligation that was never written. You cannot post a letter of credit against “market-related prices”. You cannot escrow a volume expressed as “up to”. There is no instrument in the market that credit-enhances a sentence.
That asymmetry is the whole insight, and it inverts the sponsor’s ranking. A weak buyer with firm obligations is a structuring problem: cost it, cover it, price it into the deal. A strong buyer with loose obligations is not a structuring problem, because there is nothing to structure around. It is a drafting problem that the sponsor has spent months not solving while looking for a better logo.
Conditional is not the same as unspecified
Here is where most sponsors reasonably object, and where the useful distinction lives.
No serious offtaker signs binding purchase obligations before a project has reached financial close. The buyer will not commit to take volumes from a plant that does not exist, on a date nobody can guarantee. Sponsors know this, lenders know this, and the sponsor concludes — wrongly — that specificity is therefore unavailable until after the money arrives. Chicken, meet egg.
The confusion is between two different things a document can lack. A conditional obligation is fully specified and switched on by defined events: volume, formula, term and remedy all stated, contingent on commercial operation by a longstop date. An unspecified obligation states no terms at all and defers them to a later agreement. The first is modellable — the analyst runs it as contracted revenue from the condition date, and tests the case where the condition fails. The second is not modellable at any date.
Sponsors ask their counterparties for the wrong thing. They ask for a signature, which the counterparty cannot give, is embarrassed to refuse, and replaces with a warm, vague letter that costs the counterparty nothing precisely because it commits them to nothing. The right ask is narrower and much easier to grant: keep every condition you need, and tell me the four fields. Counterparties agree to that far more often than sponsors expect, because it transfers no risk to them before the conditions are met.
The regulator has begun making the same distinction, and in stronger terms than most sponsors realise. When Nigeria’s electricity regulator issued Order NERC/2024/058, the Order on the Transition to Bilateral Trading in the Nigerian Electricity Supply Industry, on 25 July 2024, it moved the market away from a single central offtaker toward direct contracting between generators and distributors. Having done so, it did not leave the content of those contracts to the parties. It ruled:
“Within 3 (three) months of the commencement of bilateral contracting, all GenCos capacity under bilateral contracts shall be supported by firm GSAs on a ‘take or pay’ basis with adequate provision for liquidated damages.”
Order NERC/2024/058, 25 July 2024
Read that as a lender would. In a single sentence the regulator has mandated a volume basis — take-or-pay — and a remedy — liquidated damages — on the supply side of every bilaterally contracted plant in the market. It did not require generators to find better-rated gas suppliers. It required their contracts to say something specific. A market of bilateral contracts is only bankable if the bilateral contracts populate fields.
Sponsors optimise the axis they can see
Why does the error persist among competent people?
Because brand is observable and specificity is not. A sponsor can tell whether they have the supermajor. They cannot easily tell whether clause 6 is modellable, and nobody in their advisory chain is incentivised to tell them: the introduction broker is paid on the meeting, not on the term sheet, and the counterparty’s own commercial team has no reason to volunteer that the letter they just issued is worth nothing in a credit model.
So the sponsor optimises the visible axis, and the invisible one decides the outcome. The cost is rarely a rejection — rejections are cheap and fast. The cost is a slow, polite process in which the lender sizes the debt off a merchant case, the gap comes back as equity the sponsor does not have, and the better part of a year disappears.
There is an uncomfortable corollary for anyone holding a portfolio of these letters. A drawer full of MoUs from recognisable names is not evidence of commercial traction. It is evidence of conversations. The two are priced very differently.
Back to the two sponsors
Sponsor B’s document was unsigned. It was also complete: a volume, a formula, a term, a remedy, and a set of conditions naming exactly what had to happen before any of it bound anybody. The lender could model it, stress it, and size against it. Sponsor A’s document was signed by a company with a market capitalisation larger than Sponsor A’s entire sector, and populated no field at all.
The lender was never pricing the counterparty. It was pricing whether the revenue could be written down as a schedule.
Before your next lender meeting, do not schedule another call with the counterparty. Take the document you already have, put four columns on a page — volume, price, tenor, remedy — and fill them from the text. Whatever you cannot fill is what your lender will replace with the merchant case, and it is the only thing worth going back to your counterparty about.