The Offtake Agreement: Demand Contracted in Advance

An offtake agreement is a contract signed before construction under which a buyer commits to take the output, or to pay for availability, at an agreed price. The contract converts demand from a forecast into an obligation. Tapti Crossing Infrastructure Private Limited has no such contract: its Rs 310 crore of revenue depends on how many vehicles cross.

The contract makes very little sense until the problem it was invented to solve has been felt, so the problem comes first. Somebody wants to build something large and expensive. The money has to be spent first, all of it, over several years, before a single rupee comes back. And the only thing standing behind the repayment is a stream of revenue that does not yet exist, from customers who have not yet arrived, at an asset that has not yet been built. Every part of that sentence is a forecast. A lender is being asked to lend against a document that contains no history at all.

Consider how an ordinary person would answer that. If a cousin asked for a loan to start a catering business, the first thing worth knowing is who is going to buy the food. If the answer is a signed standing order from one office canteen, four hundred plates every working day at a rate written down in advance, something settles. If the answer is that the food will be very good and people will surely come, the opposite happens. Nothing about the food changed between those two answers. The demand changed from a hope into a commitment.

An offtake agreementA contract under which a buyer agrees in advance to take the output a project will produce, or to pay for the project being available to produce it, at a price the contract sets. is the catering standing order, written at the scale of an asset that costs hundreds of crores. The agreement is a contract signed before the asset exists, between the project and the party who will buy what the project produces, fixing in advance that the buying will happen and at what price. On a power station it is a purchaser of electricity. On a processing plant it is a purchaser of the processed material. On some kinds of road and rail it is a public counterparty paying for the asset being kept open. The mechanism is the same in every one of those settings. The counterparty, the price and the term change from one deal to the next.

The worked example throughout is Tapti Crossing Infrastructure Private Limited, an invented single-asset toll road company formed to build and operate one crossing, funded by Rs 1,260 crore of borrowing and Rs 540 crore of equity against a project cost of Rs 1,800 crore. In the modelled year it collects revenue of Rs 310 crore, spends Rs 62 crore running the crossing, and therefore earns Rs 248 crore of earnings before interest, tax, depreciation and amortisation (EBITDA). Its debt service in that year is Rs 182.70 crore, being Rs 119.70 crore of interest at the project's own contracted rate of 9.5 per cent on Rs 1,260 crore plus Rs 63 crore of scheduled principal. The cover ratio on those figures is 1.36 times. Tapti Crossing Infrastructure Private Limited has no offtake agreement, and the absence is the most useful fact about it: an offtake agreement shows its work most clearly in a project that has to manage without one.

What exactly does an offtake agreement commit a buyer to?

Two forms, and they behave differently enough that treating them as one thing will get a reader into trouble. The first is take or payA contract form in which the buyer either takes the agreed quantity of output or pays for it anyway, so the project is paid whether or not the buyer wanted the output that period.. The buyer agrees to take an agreed quantity of what the project produces, and if it does not take that quantity, it pays for it anyway. The name is the whole mechanism. The second is an availability paymentA payment made for the asset being kept ready and usable to an agreed standard, rather than for any quantity actually produced or consumed.. The buyer pays for the asset being available and working to an agreed standard, whether or not anything is produced or consumed at all.

The difference between the two is where volume risk finally comes to rest, and that is the only question worth asking on first meeting a contract of either kind. Under take or pay, the contract usually covers a stated quantity, so anything above or below that quantity is still exposed and some volume risk stays with the project. Under an availability payment, the payment is not attached to a quantity in the first place, so almost all of the volume risk moves away. A project paid for being open does not care how many people walked through the door, only that the door was open and in the condition the contract specified.

Set both against the case where there is no contract at all. This crossing sits in exactly that case. There, every rupee of revenue is a collected revenueMoney that arrives because a user chose to use the asset and paid at the time, rather than because a counterparty promised in advance to pay it.: it arrives because somebody chose to use the asset and paid. Nobody promised any part of it. There is nobody else to hold the volume risk, so the project keeps all of it.

The three cases appear at street level. A tailor with a signed contract to supply four hundred school uniforms at a fixed rate per uniform, paid whether or not every child collects, is on take or pay. A watchman paid a monthly wage to keep a gate manned, whatever traffic goes through the gate, is on an availability payment. A tea stall on a footpath is on neither: it earns exactly what it sells that day. All three can end the month with the same money in hand. The three did not carry the same risk to get there, and the difference in that risk is what everything below turns on.

Why is the contract signed before anything has been built?

The ordering reads backwards the first time and is worth slowing down on. Signing first is not an administrative accident but the whole design. A buyer signs a commitment to purchase output from a plant nobody has built yet, at a site where nothing has been poured, from a company that has never produced anything. Read as a commercial decision that looks odd. Read as a financing decision it is the only order that works.

Follow the money and the sequence becomes obvious. The project needs Rs 1,260 crore of borrowing before construction. The lenders will only advance that against something. There is no trading history, no second business and no other asset, so the only thing that can be offered is the future revenue. Future revenue in the form of a forecast is an opinion. Future revenue in the form of a signed contract with an identified payer is a claim on somebody. The contract has to exist before the asset because the contract is what makes the asset fundable, and a contract signed after completion would be arriving to solve a problem that was already survived or already fatal.

Go back to the stallholder. A woman running one snack stall wants to open a second one. If she walks into a lender's office with the observation that the area is busy, she is asking somebody to share her optimism. If she walks in with a standing weekly order from an office canteen, signed, quantity stated, rate stated, she is asking somebody to look at the canteen. The second conversation is short and the first one usually does not happen at all. Notice the timing in her case too: the canteen commits to buying from a stall that has not been built. The canteen is not being generous. The canteen wants the supply, and the supply will not exist unless somebody funds it.

So the sequence runs in a fixed order. First the contract, committing a buyer to something not yet in existence. Second the money, advanced by the project lenders against that commitment rather than against a trading record. Third the construction, now funded and able to proceed. Fourth the output, finally making the promise in step one capable of being performed. Reverse any two of those and the financing does not happen. The fixed order is also why negotiating the contract is not a document exercise running alongside the build; the negotiation is the gate the build has to pass through.

What question does the contract replace for the lender?

Here is the cleanest way to see what an offtake agreement actually does. An offtake agreement does not make a project safe. An offtake agreement changes the question a lender has to answer first, from a question no evidence can settle into a question evidence can settle.

Without a contract, the project lender's first question is whether enough users will come. No document stands behind that question. Only a forecast answers it: how many vehicles, at what growth, at what price, for how long. Every input is a judgement, the judgements compound, and nobody in the room can produce a piece of paper that settles any of them. The exposure has a name, demand riskThe risk that the number of users, or the quantity bought, turns out lower than the level the financing was built on.. Demand risk is genuinely hard to underwrite: the thing being underwritten is the aggregate behaviour of thousands of people who have made no promise to anybody.

With a contract, the first question becomes whether one identified buyer will pay. Documents do stand behind that question. The buyer's accounts can be read, what else it has committed to can be examined, how it has behaved when it was under pressure before can be traced, and a view can be formed that is grounded in evidence rather than in a growth assumption. The two investigations are completely different, and only the second one can be conducted from accounts and documents, so lenders want the contract badly enough to accept a lower price for the output in exchange for it.

The technical name for what has happened is credit substitutionReplacing a claim whose quality depends on how a business performs with a claim whose quality depends on a named payer's ability to pay.. The quality of the project's revenue used to depend on the project's own commercial success. Now it depends on somebody else's ability to pay. The revenue line has been substituted with a claim on a third party. Credit substitution is why, in practice, a financing with a strong offtake contract is often discussed as though the borrower were the buyer rather than the project, and why the buyer's own condition becomes the first thing an experienced reader looks at.

The household version is exact. Suppose the decision is whether to lend to a relative who has just opened a shop. Question one: will the shop do well? Nobody can answer that one. Question two, if the relative produces a one year contract to supply a school canteen: will the school pay its bills? The school's accounts answer that one. The loan has not been made safe. The loan has been made examinable. Examinable is not the same as safe, and it is a great deal better than nothing.

What risk arrives in place of the one that left?

Nothing was destroyed. Something was exchanged, and what the project received has its own character. Careless readings go wrong exactly there. The project no longer depends on thousands of users deciding independently to turn up. The project depends instead on one party continuing to be able to pay. The new exposure is counterparty riskThe risk that the specific party on the other side of a contract stops being able, or willing, to perform what it promised., and it has a shape that the risk it replaced did not have.

The shape is concentrationWhen a large share of something depends on a single source, so that one failure moves the whole of it at once rather than a small part of it.. Demand risk across thousands of users is distributed: on a bad day some people do not travel, and the revenue falls a little. Every user disappearing on the same morning is very unusual. Counterparty risk is not distributed at all. The payer either pays or does not. There is no version of a payer failing where the project loses fifteen per cent of the contracted revenue and carries on. Demand risk has been converted into counterparty risk rather than eliminated, and the conversion is worth making only where the payer is more predictable than the demand was. Whether the payer is more predictable is a judgement about a specific party, not an arithmetic result.

Say it in a plainer form. A household with two salaries and a household with one salary can have the same income. The two households do not carry the same risk. The second has one event that takes everything, and the first has two events that take half each. A tailor whose entire month comes from a single uniform order has swapped the uncertainty of walk-in customers for the certainty of one order and the concentration that goes with it. If the school delays payment for three months, the walk-in tailor next door is in better shape, despite having no contract at all.

Concentration is also why the word secure, attached to contracted revenue, needs handling carefully, and why neither a contracted structure nor a collected one is preferable in itself. Contracted revenueRevenue a project is entitled to because a counterparty promised in advance to pay it, as distinct from revenue it collects from whoever turns up. is secure in exactly one sense: the obligation to pay it exists in writing. Whether the money arrives is a question about the party who signed, and no amount of drafting turns that into a certainty. A contract is a claim, and a claim is worth what the person it is against is worth.

What does a contract do to the range of outcomes rather than to the middle?

This is the reading most treatments get backwards, so it is worth stating flatly before anything else. A contract does not usually raise the revenue a project expects. A contract narrows the spread of outcomesThe distance between the worst and the best result a set of assumptions can produce, as distinct from the single result sitting in the middle of them. around whatever revenue was expected. If anything, a buyer signing a long commitment in advance normally expects to pay something less than the going rate for the privilege of fixing it, so the middle can move slightly down rather than up.

Why does narrowing the range matter so much more than moving the middle in this particular structure? Because of how thin the structure is. Take Tapti Crossing Infrastructure Private Limited exactly as it stands. EBITDA of Rs 248 crore against debt service of Rs 182.70 crore leaves Rs 65.30 crore of spare cash. The spare is 35.7 per cent of the debt service. Rs 65.30 crore of spare sounds comfortable said aloud. Now convert it into the language of revenue. Operating cost of Rs 62 crore does not fall when traffic falls, so every rupee of lost revenue is a rupee of lost EBITDA. Revenue of Rs 244.70 crore produces EBITDA of exactly Rs 182.70 crore and a cover ratio of exactly 1.00 times. Rs 244.70 crore is 21.1 per cent below Rs 310 crore.

A geared structure is decided at the bottom of its range and not in the middle of it, so narrowing the range from below is worth far more here than lifting the middle would be. The top of the range is pleasant and changes only what the sponsors receive after everybody else is paid. The bottom of the range is where the structure either holds or does not. An arrangement that leaves the expected revenue untouched and removes the bad outcomes has done something enormously valuable. A reader looking only at the central figure will see none of it. The central figure did not move.

The household version once more. Two people earn the same average monthly income. One is salaried; the other does piece work and has a good month and a bad month in turn. If they both have a fixed loan instalment that takes most of an average month's income, the salaried one is fine and the one on piece work misses the instalment in every bad month. Their averages are identical. Their ability to service a fixed obligation is not, and no measure that looks only at the average reveals that.

What happens to the band if part of that Rs 310 crore is contracted?

Now work it, as a labelled hypothetical and nothing more. Tapti Crossing Infrastructure Private Limited has no offtake agreement and none is being proposed for it. To show what a contract does, suppose some share of the same Rs 310 crore arrived under a contract at a fixed price, and the remaining share kept varying. A contract on the revenue line touches neither operating cost nor debt service, so hold them at Rs 62 crore and Rs 182.70 crore throughout.

The varying part needs a band, and the source of that band matters. No traffic forecast for this crossing exists, so nothing justifies any particular width. A band of plus or minus 20 per cent on the uncontracted part is used below purely as a drawing choice, so the shape of the effect can be seen. The band is not a forecast and is not derived from anything. The 20 per cent is not a property of this crossing or of roads in general.

With nothing contracted, revenue runs from Rs 248 crore to Rs 372 crore, EBITDA from Rs 186 crore to Rs 310 crore, and cover from 1.02 times to 1.70 times. With half contracted, revenue runs Rs 279 crore to Rs 341 crore and cover 1.19 times to 1.53 times. With three quarters contracted, cover runs 1.27 times to 1.44 times. With all of it contracted, cover is 1.36 times and does not move at all.

The five lines are read down rather than across. The middle never moved: every row is centred on 1.36 times. The arithmetic that produces 1.36 times does not know or care where the Rs 310 crore came from. The width fell in a straight line, from 0.68 to 0.51 to 0.34 to 0.17 to nothing, exactly in proportion to the share left uncontracted. And the bottom edge rose from 1.02 times to 1.36 times. On a structure with Rs 65.30 crore of spare, that rise in the bottom edge is the entire value of the contract, and it is invisible to anybody who was watching the central figure.

One more reading worth taking from the table above, before the control is moved. At nothing contracted, the low end leaves EBITDA of Rs 186 crore against debt service of Rs 182.70 crore, a spare of Rs 3.30 crore. At half contracted the same low end leaves Rs 34.30 crore, at three quarters Rs 49.80 crore, and at all contracted Rs 65.30 crore. The project went from having almost no room in a bad year to having its full room in every year, without a single rupee being added to the middle.

Why does this crossing have no offtake agreement at all?

Because it is a toll road, and a toll road is paid by the people who use it. Tapti Crossing Infrastructure Private Limited collects Rs 310 crore from whoever crosses. No buyer of the output stands behind that figure, no party has promised any part of it, and no clause anywhere entitles the company to a rupee it did not collect from a vehicle.

The Rs 310 crore can therefore be described plainly. The figure is a collection rather than an entitlement. The figure also has no traffic forecast standing behind it, no concession period, no debt tenor and no year by year schedule. So there are two absences here rather than one: nothing contracts the revenue, and nothing in the record projects it either. Both belong in any note written about this project, and neither can be estimated around.

A great many real assets are financed exactly this way, so the absence is not an oversight but the point: the same 1.36 times means something different here than it would with a contract behind it. A structure that lives on collection is not defective. A collection structure is a different animal, underwritten differently, and it is the more common case for user-paid infrastructure. Reading one payer's accounts and stopping there is exactly what a lender to this crossing cannot do.

The difference between a right and a promise sits here too, and it is worth marking because the two are easy to run together. Tapti Crossing Infrastructure has the right to operate the crossing and to collect from it. A right to collect is not a promise that anybody will pay. The concession that grants the right, and what a concession does and does not carry, is set out under the concession agreement.

Would a contract change the cover ratio at all?

Answer it with the arithmetic rather than with an opinion, because this is the point where a reader's instinct and the numbers separate. Revenue of Rs 310 crore less operating cost of Rs 62 crore is EBITDA of Rs 248 crore. Debt service is Rs 182.70 crore. Rs 248 crore divided by Rs 182.70 crore is 1.36 times. Now put the identical Rs 310 crore under a contract. EBITDA is still Rs 248 crore. Debt service is still Rs 182.70 crore. The cover is still 1.36 times.

The ratio is identical and the confidence in it is not. A cover ratio quoted without saying where the revenue comes from has left out what decides how much weight the ratio can carry. The whole lesson is compressed into that one line. Nothing in the division knows the source of the numerator. A ratio is a division, and a division is indifferent to provenance in a way that a reader must not be.

Which is why the two figures should never travel alone. Written properly, the reading is not cover of 1.36 times on its own but cover of 1.36 times on collected revenue, with no contract and no forecast in the record behind it. The second version is longer, and it is the only one that can be acted on. The first version invites a reader to compare it directly with a 1.36 times somewhere else that was produced under a firm contract, and those two numbers are not comparable however identical they look.

Something similar happens with household budgets and nobody is confused by it there. Two people both say they can afford an instalment of twenty thousand rupees a month on an income of thirty thousand. One has a salaried job and the other drives on daily hire. Same ratio, same instalment, entirely different conversation. Everybody knows to ask the follow-up question in the household case. In a project financing the ratio arrives looking like a finished answer, so the same follow-up question gets skipped surprisingly often.

What does an offtake agreement never reach?

Three limits, and each has ended a project that had a perfectly good contract sitting in the file. Naming them is what keeps the contract from being read as a shield rather than as one arrangement covering one line.

First, it does not fix operating cost. An offtake agreement is about what arrives, not about what leaves. If the cost of keeping the crossing open rises, that increase lands on the project in full, and no clause about output touches it. On this project the arithmetic is unforgiving: Rs 248 crore of EBITDA against Rs 182.70 crore of debt service means a rise in operating cost of Rs 65.30 crore, from Rs 62 crore to Rs 127.30 crore, takes cover to 1.00 times even with revenue perfectly contracted at Rs 310 crore. The canteen order does not stop the price of vegetables rising.

Second, it does not survive its counterparty failing. A contract is a claim against a party, and if that party cannot pay, the project is holding a piece of paper and a queue position. Concentration, described earlier, arrives at its conclusion here, and that is why the counterparty work the contract made necessary is not optional.

Third, it does not bind a party who did not sign it. A public authority that is not a party to the offtake agreement is not constrained by it. Whatever a government does that reaches the project's revenue is a separate matter, sitting outside the contract entirely, and is covered separately.

A project with a firm contract in place can still fail for reasons the contract was never about. Cost and sovereign action therefore appear as separate rows on any risk register rather than being folded into a revenue row that somebody has ticked. The contract narrows one range. The contract leaves every other range exactly where it was.

How do a lender, a sponsor and an analyst each read a contracted revenue?

Three readers, three genuinely different first moves. Watching all three is the fastest way to see what an offtake agreement is worth in practice rather than in principle.

A project lender reads a contract as a change in what has to be diligenced. Before it exists, the work is a traffic or demand study and a long argument about assumptions that nobody can settle. After it exists, the work is on the payer: what it earns, what else it has committed to, how it has behaved when it was squeezed before, and whether the obligation being taken on is large relative to what that payer can carry. The lender also reads the contract for what it does not cover. The uncontracted part is where the cover ratio will actually be decided. On this project, with nothing contracted, that reading is the whole file.

A sponsor's own finance team reads it as a trade between price and range. The buyer signing a long commitment normally wants something for it, and what it wants is usually a lower price than the open market would pay. So the team is deciding whether to accept a smaller middle in exchange for a narrower band. On a structure geared at Rs 1,260 crore of debt against Rs 540 crore of equity, that trade is usually worth making. The equity is the first thing lost when the bottom of the range breaks, and the sponsors keep only what is left after debt service in any case. Recall that Rs 65.30 crore of the toll road's cash reaches the sponsors in the modelled year. Rs 65.30 crore is a single year's figure on a single year's cash, and one year alone cannot be converted into a return over the life of the asset.

An infrastructure analyst reads a contract as the test of whether two projects are comparable at all. Given two projects both showing 1.36 times, the two numbers are the same, so the analyst's first question is not which number is bigger. The first question is where each revenue comes from. One may be contracted with a strong payer, one collected from users with no forecast in the file, and the ratio will not tell them apart. The analyst's second question is what happens to each at the bottom of its own range. The bottom of the range is where the difference finally shows up as a number rather than as a description.

None of the three treats the contract as a conclusion. The contract changes the work; it does not finish it. A reader who wants one number to carry the whole judgement will be disappointed, correctly. No such number exists.