What a Take-or-Pay Clause Costs When Your Build Forecast Is Wrong

A take-or-pay clause costs the buyer the shortfall between the volume committed in the agreement and the volume actually ordered, settled at the contract unit price for the packs and, depending on how the shortfall paragraph is drafted, for cell material, reserved capacity, and unrecovered tooling amortization. The clause itself is rarely the expensive part. The expensive parts sit around it: the forecast band printed next to it, the payment schedule that releases cash before there is a drawing to freeze or a first article to accept, and the termination-for-convenience section that reads like an exit right up until you reach the sentence about work in progress and finished inventory.

The first shortfall notice I had to explain upstairs was one page. Two columns, a subtraction, and a total I had already spent in my head three quarters earlier.

That total was not fraud, and it was not a clause somebody slipped in at the end. It was arithmetic that had been sitting in the document since before the first pack was built. What made it expensive was that nobody in the room, me included, had read the same number the same way.

I read supply agreements backwards now. I start at the money clauses and walk toward the technical ones, because the technical ones are where everybody's attention goes and the money clauses are where the damage is stored.

A number that appears three times and doesn't match

Here is what I check first. In a real negotiation the annual build volume shows up in at least three places: the program slide deck, the non-binding forecast exhibit, and the volume commitment language. Sometimes a fourth place, a capacity reservation fee schedule.

If those numbers agree, fine. If they don't, and they frequently don't, the one that governs is whichever one the supplier's side can point at. That is usually not the slide.

I sat in a review once where the program forecast said roughly 1,400 units a year and the take-or-pay floor said 900. Everyone nodded at 1,400. The 900 was the number with teeth, and it had been negotiated by someone who had already rotated off the program. When the build plan came in at 600, the invoice was written against 900. Nobody was wrong. Everybody was reading a different page.

Paying for work before there is anything to accept

Calendar milestones versus artifact milestones

The oldest bad structure is a payment schedule keyed to dates. Thirty percent at PO, thirty at design complete, thirty at first shipment, ten at some future date nobody remembers to close out.

Say the words design complete out loud to three engineers and you get three answers. That is the whole problem. A calendar milestone pays on a date. An artifact milestone pays on a thing that either exists or does not.

What I want instead, and what I push for on every pack program now:

• A design freeze payment released against a released, revision-controlled drawing set, plus a locked cell part number, plus a BMS firmware specification at a stated revision. If the drawing set is not released, the milestone is not met and the money stays where it is.
• A PPAP or equivalent approval payment released against a signed approval at a defined submission level, with a named signer on the buyer's side. Not buyer acceptance in its sole discretion, and not supplier notification. If open issues remain on the dimensional report, it is not approved.
• A first article acceptance payment released against a first article inspection report and a functional test at temperature, run to a procedure that was written down before the sample was built.

Why the sequencing matters more than the percentage

Once cash has moved on a calendar, the supplier has no financial reason to close your PPAP open issues. You are asking a program manager on their side to prioritize a punch list that pays them nothing. That is not a moral failing on their part. It is the schedule you wrote.

I hold back a meaningful percentage, think single digits to low double digits of pack price, until the first production lot has run and incoming inspection data has held for a stated period. I write the holdback release as an actual date with a signature, not as upon final acceptance.

NRE is the sharpest one. NRE paid in full at design freeze is money spent before a drawing exists. Split it against artifacts: schematic release, firmware freeze, thermal validation report, first article. Each installment is one more thing the supplier has to finish before the next payment clears.

Where the band does the damage

A volume commitment is almost never one number. It is tiers or bands, each with a price step. Up to 500 units at one price, 501 to 1,500 at a lower one, above 1,500 at the lowest. That structure is reasonable. Volume price breaks are real.

The trap is what attaches to the bottom of the band. If the take-or-pay floor is set at the lowest volume in the band you forecast into, and the band rolls forward annually, you can end up obligated at a floor that was never meant to be your plan.

The protection I insist on is a rolling forecast with three zones, written in weeks, not adjectives.

Firm zone: the next several weeks, non-cancelable, the supplier buys material. Semi-firm zone: the weeks after that, cancellable with a defined charge. Planning zone: everything beyond, advisory only, no liability.

The charge in the semi-firm zone is where I spend my negotiating time. The default ask from the supplier side is full unit price for anything canceled in that window. That is not a material loss, that is margin on a pack they never built. I push for material at documented cost plus a defined handling percentage, with the cell purchase orders as the evidence. Sometimes I get it. Sometimes I get it capped.

And I always ask whether a reschedule counts as a cancellation. If the agreement says any reduction in a release is a cancel, then moving a build out six weeks costs money. That is worth arguing about while nobody is angry.

What take-or-pay actually obligates

Strip it down. The buyer is committing to pay for a quantity, not to take a quantity.

The usual calculation is committed volume minus taken volume, multiplied by contract unit price, over a defined period. Then come the questions that actually decide the number.

Is the period annual, per-quarter, or cumulative across program life? Annual resets are far more forgiving and twice as hard to get.

Does a shortfall in one period carry forward as a credit against future orders, and for how long? A credit that expires is a penalty with better manners.

Does the shortfall calculation use pack price or the material content of the pack? That is the biggest lever and the one most buyers never ask about.

Does an unused capacity reservation fee credit against future purchases, or does it just get banked by the supplier?

I have watched a program pay a shortfall settlement for cells that the supplier then used on another customer's build. Nothing improper happened. The clause allowed it. The buyer's finance team had signed a document that paid for material, not for an outcome.

Two ways to move a price, only one you can live with

A price adjustment tied to a named external index is a mechanism. It has a reference, a measurement frequency, a pass-through base, sometimes a collar so small moves get absorbed, and a documentation requirement. I can model it and defend it to a CFO. I may not like the result, but I know what I am buying.

Open-ended renegotiation rights look friendlier and are worse. Language saying the parties will meet in good faith to discuss adjustments in the event of significant material cost changes is not a pricing clause. It is a scheduled argument with no floor, no cap, and no reference point. And the argument lands at the worst possible moment: after the tooling is built, after PPAP is approved, after you have spent the second-source budget somewhere else.

The same goes for prices subject to adjustment for market conditions. Market conditions is not a number.

My preference is boring on purpose. Index, base, frequency, cap, and a requirement that the supplier furnish the underlying index prints with each request. If a supplier rejects a capped index but wants renegotiation, that tells me which direction they expect the conversation to go.

Termination for convenience is not an exit

This is where the money hides in plain sight.

A termination for convenience clause usually says the buyer can end the agreement on notice. What it does not say at the top is what stays behind. Read down to the payment-on-termination section and you find some combination of unpaid NRE, tooling that must be paid for in full whether or not it transfers, work in progress, finished pack inventory, non-cancelable material on order, and occasionally a percentage of remaining contract value.

Two distinctions matter more than the rest.

First: buyer will take delivery of finished goods versus buyer will pay for finished goods. Taking delivery means you own usable packs with a shelf life ticking. Paying for them and letting the supplier scrap them means you paid for nothing you can deploy. I have seen both written into the same template by different people at the same company.

Second: tooling ownership and removal rights. Can you take the fixtures? Is the test fixture yours? Is the firmware build environment documented well enough that another supplier can use it, or does the code live on one engineer's laptop? Tooling you cannot remove is a lease with extra steps.

I want the termination payment schedule to be a table, not a paragraph. Line items, caps, and a defined valuation for work in progress. Paragraphs get interpreted. Tables get paid.

Exclusivity costs the same dollar as the volume commitment

Exclusivity and volume commitment get negotiated as separate topics and they should not be. They trade against each other on the same ledger.

If I am giving a supplier a multi-year volume floor, that floor is worth something beyond margin, because it de-risks their capacity planning. My ask in exchange is the right to qualify a second source without triggering a breach, plus reasonable access to the pack documentation needed to do it.

The failure mode I have seen is exclusivity language that blocks a second source for the term while the volume commitment runs the whole term. That is all the obligation and none of the optionality. And a second source you cannot afford to qualify, because it means full NRE, full tooling, and a second PPAP line, is not really a second source. It is a slide in a risk review.

There is an honest tradeoff here. A supplier who gets real exclusivity will often price better and allocate capacity first. On a low-volume program where dual qualification doubles a fixed cost that is already disproportionate, that can be the right deal. But the buyer should be choosing it on purpose, with the number in front of them.

What finance should read before anyone signs

I hand the draft to my finance counterpart with a specific list, because finance reads for different things than engineering does and catches what the program team skips.

The shortfall definition, the measurement period, and whether credit carries forward.

The NCNR window and exactly what triggers a cancel charge.

The price adjustment mechanism, its cap, and who bears index documentation.

Tooling ownership, title transfer date, and removal rights.

The termination payment table, line by line.

Any capacity reservation fee and whether it credits against purchases.

Working capital terms, because a milestone schedule that front-loads payments is a financing decision dressed as a commercial one.

Any guarantee, letter of credit, or parent obligation triggered by the commitment size. Those get pulled in as a condition of the volume floor more often than people expect, and they land on the balance sheet.

None of that is a question of what a court would do with the language. It is a question of what the number is before anyone signs.

The last thing I do is go back to that one-page invoice and check the arithmetic by hand against the clause. Usually it is right. Twice it has not been, once because the measurement period was cumulative and once because a reschedule was counted as a reduction.

The forecast will be wrong. Everybody's forecast is wrong. What separates a bad year from a bad contract is whether the wrong number hits a date on a calendar or a band you can flex.