The storage contract locks your volume, not your price

Sandisk's 10-Q for the quarter ended 2 October 2026 describes a storage contract that fixes what a customer must buy and only part of what it pays. The agreements, which the company calls New Business Models, “generally commit us to deliver, and our customers to purchase, a stated volume of products, mostly over multi-year periods”, and they carry “pricing mechanisms consisting of fixed and variable components”. Sandisk expects them to become “our predominant way of doing business”. The same note puts a number on what stands behind them: $12.8 billion of collateral issued or held at third-party financial institutions against $88.9 billion of remaining performance obligations, about 14 cents for every dollar of them. Sandisk recognises none of it, because it has no unconditional right to the money until a customer defaults.

This reaches the CIO who signs a multi-year storage commitment before the volumes arrive, the treasurer who finds the bank lines behind it, and the general counsel who reads the default clause. It reaches any owner of a retention schedule costed on the assumption that a terabyte gets cheaper every year.

01Volume grew a fifth while revenue grew three fifths

Seagate shipped 220 exabytes in the quarter ended 2 October 2026 against 182 a year earlier, 20.9% more, and took 61.0% more revenue for them: $4,233 million against $2,629 million. Nearline drives, the tier that holds backups and archives rather than the working set, moved from 160 exabytes to 198 and are 90% of what the company ships. The filing gives the cause in its own words, “favorable pricing actions undertaken by the Company and an increase in nearline exabytes shipped”, and its chief executive gave it a name on 27 October 2026: “value-based pricing strategy”, held alongside “maintaining supply discipline”.

WD reported the same quarter a week earlier and guided higher: revenue of $4.38 billion, up 55% year on year, a GAAP gross margin of 57.8% against 43.5%, and revenue of $5.1 billion at a non-GAAP gross margin of 60.5% for the December quarter at the midpoint of its range. Its chief financial officer, Kris Sennesael, put “improved pricing” beside “robust exabyte growth” as the drivers. Sandisk's first fiscal quarter carried $11,031 million of revenue against $2,308 million and an 85.0% GAAP gross margin against 29.8%. Micron's fiscal 2026 revenue was $133.19 billion against $37.38 billion, at a gross margin of 80.7% against 39.8%.

02The volume is guaranteed and the price inside it is not

What the contract states is a volume. What it does not fix is the whole price for the term: the mechanism has a fixed and a variable component, so the part that floats keeps floating while the volume is already committed. Sandisk does not present that as a favour, and does not need to. The same paragraph says the agreements are “supported by financial guarantees that are intended to provide additional protection in the event a customer does not satisfy their contractual purchase obligations”, and that, “as NBMs are expected to become our predominant way of doing business”, the company believes the model will “contribute to greater predictability of revenue, support production planning, and enhance supply assurance for our customers”. Supply assurance is what the buyer is paying for; reduced cyclicality is what the supplier is buying.

03The collateral sits at a third-party bank and does not appear on Sandisk's books

Sandisk gives the aggregate: “The aggregate amount of collateral issued or held by third-party financial institutions was $12.8 billion as of October 2, 2026, representing the maximum potential proceeds available to the Company in the event of customer default or breach.” Against $88.9 billion of remaining performance obligations, of which $87.4 billion has not yet been billed, that is about 14 cents of standing collateral for each dollar of contracted volume. The company recognises none of it, because the right to the funds arises only on default or breach. What the buyer gives up in exchange is not only the option to buy later at the spot price: it is the use of the capital for the whole term. Refund liabilities, $1.6 billion against $1.5 billion a quarter earlier, “primarily include security deposits collected under NBMs that must be refunded at the end of the contract term”, so in that part of the book the cash is already with Sandisk, refundable to the customer after it has satisfied all performance conditions.

04The gain holds while the exabytes stay scarce

Seagate's chief executive put the two halves in one sentence on 27 October 2026: a HAMR-based roadmap “enables us to address growing exabyte demand through areal density gains while maintaining supply discipline”. Areal density is how a drive maker adds capacity without adding a factory, so the discipline is a decision rather than a constraint. That is what makes the position worth a 57.3% gross margin and what makes it fragile: the discipline is a decision each supplier makes separately, and the price holds only while they all keep making it.

05What survives a price reversal is the term

The structure cuts the other way in a downturn, and the honest reading includes it. A buyer who fixed part of the price printed in that quarter has protection if the market falls, and a supplier that signed has given up part of the fall. Sandisk says so itself: the agreements “do not eliminate the risks associated with customer demand, market conditions, or operational execution”. What does not reverse with the price is the commitment. The volume does not shrink because the price turns out to be high, the deposit returns at the end of the term rather than the end of the need, and the bank line stays committed throughout. A buyer who is right about the price and wrong about the volume has still bought the term.

06The retention schedule was written on an assumption that has stopped holding

The unit to carry into your own plan is $4.80. That is how much more revenue per blended terabyte Seagate Technology took in the quarter ended 2 October 2026 than a year earlier, on its Form 10-Q: $19.24 million of revenue per exabyte shipped against $14.44 million, 33.2% higher, both Mosaic Singularity analysis, October 2026, on a blend that is 90% nearline by volume. A retention schedule is a cost decision taken once and renewed by default: how long the corpus is kept, how many copies, which tier holds the cold half. Every one of those defaults was priced when a terabyte got cheaper each year. Multiply the $4.80 by the terabytes in your estate, add about 14 cents of collateral for every dollar of volume you commit, and the archive tier stops being the cheap place to put a decision you have not made yet.

07The decision is what the commitment is for, and it sits in four seats

The question in front of a buyer is not whether storage is expensive. It is how much capacity to lock, for how long, at what proportion of fixed price, and with whose capital. Those answers belong to different people, and the collateral is arranged with a bank rather than the supplier, so it lives in a document the procurement file does not hold.

08What the board puts to management

• Which storage commitments fix a volume without fixing a price, and what share of each is variable?
• How much collateral or letter of credit capacity does the commitment require, for how many years, and which facility provides it?
• What are the specified events of default, who holds the collateral, and who releases it once we have performed?
• If the price falls by a third, what does the remaining term cost us, and can it be renegotiated or assigned?
• Which retention tiers were costed on deflation, and what does holding them at the current price per terabyte do to the plan?
• Who signed the last storage commitment, and did that seat see the collateral terms before signing?
• Resolve, at the sitting that approves the capital plan and in any case by 31 March 2027, that no storage commitment fixing a volume for a multi-year term is signed without a capped variable component, and that the collateral and letter of credit capacity standing behind those commitments is reported to the audit committee at each quarter end. Two facts make that a board resolution rather than a procurement note: the contract fixes the volume and only part of the price, and $12.8 billion of the buyer's capital sits at third-party banks in a document the procurement file does not hold.

09The six filings behind every number here

• Seagate Technology, Form 10-Q for the quarter ended 2 October 2026, accession 0001137789-26-000309.
• Seagate Technology, Form 8-K, Item 2.02, Exhibit 99.1, filed 27 October 2026, accession 0001137789-26-000300.
• WD, Form 8-K, Item 2.02, Exhibit 99.1, filed 22 October 2026, accession 0001628280-26-066882.
• Sandisk, Inc. Form 8-K, Item 2.02, Exhibit 99.1, filed 29 October 2026, accession 0001628280-26-068243.
• Sandisk, Inc. Form 10-Q for the quarter ended 2 October 2026, accession 0001628280-26-068601.
• Micron Technology, Form 8-K, Item 2.02, Exhibit 99.1, filed 30 September 2026, accession 0000723125-26-000018.

10What to watch before 31 March 2027

• WD's December quarter, guided to $5.1 billion of revenue at a 60.5% non-GAAP gross margin. If that prints and the guidance after it is higher again, any commitment signed now is priced before another rise.
• Sandisk's collateral line in its next 10-Q. $12.8 billion stood against $88.9 billion of obligations as of 2 October 2026; if obligations grow while the collateral falls, the terms were a peak condition rather than a new one.
• Any supplier that offers a fixed price for a multi-year term without collateral. That offer ends this reading.
• The first enterprise buyer to disclose a deposit or a letter of credit under one of these agreements. It puts the other side of the trade on the record.
• 31 March 2027. If no supplier has announced a further increase by then and Sandisk's collateral has fallen, the reading that storage terms tightened permanently fails: the price rise was a cycle and the collateral was its peak.