What Is Customer Concentration Risk? GT Advanced's One Buyer
Customer concentration risk is the danger that a business depends so heavily on one buyer that that buyer's decisions — to order less, rewrite the terms, or stop altogether — can break it. Revenue share is only the first measure. The real exposure sits in the contract: who must buy, who must build, and who can walk away without paying for it.
In late 2013, GT Advanced Technologies, a New Hampshire maker of furnaces used to grow industrial sapphire,1 announced a multi-year agreement to supply sapphire to Apple.2 Apple would advance roughly $578 million in installments;3 GT would operate a plant in Mesa, Arizona, leased from an Apple affiliate,4 and supply consumer-electronics sapphire to that customer alone. The shares soared. In October 2014, less than a year later, GT filed for Chapter 115 — the court-supervised reorganization a company enters when it cannot meet its obligations. What the filings exposed was not a surprise customer loss but an asymmetry that had been in the agreement from the start.
The questions below are the ones owners should put to any analysis of a business with a dominant customer, whether buying it, holding it, or still running it. This is a historical illustration of a research method, not a recommendation of any company or security.
How much of the profit and cash, not just the sales, sits with the biggest account?
A good answer names the customer and shows several years of three things: sales, profit after the direct cost of delivering the work, and cash actually collected. It says whether the share is rising. Large accounts often buy at the thinnest margins, so a business with 30 percent of sales but 60 percent of profit from one buyer is far more exposed than the revenue line suggests.
A weak answer is a single rounded percentage, usually paired with the reassurance that lenders accept it. Public companies must disclose any customer at 10 percent or more of revenue — but are not required to disclose the customer's name7 — and that threshold is a disclosure trigger, not a safety threshold, and it is routinely mistaken for one. Watch, too, for a buyer counted as three: a parent, a subsidiary and a distributor who all send the same purchase orders.
Does the contract actually require the customer to buy anything?
A good answer reads from the clause itself. Is there a minimum volume? A take-or-pay provision, meaning the customer pays for agreed quantity whether or not it takes delivery? A termination fee? If the customer can go to zero at no cost, the business has a hope, not a backlog.
A weak answer uses the word partnership and describes the relationship rather than the terms. "They've been with us for years" is not a clause, and money paid up front is not the same as a commitment to buy.
GT's agreement set out how much sapphire GT had to stand ready to produce and set no minimum the customer had to purchase. The advance was repayable over several years, either as credit against purchases or in cash. An advance that must come back in cash when the customer does not buy is debt with a customer's name on it.
Who is left holding the costs that only one customer can use?
A good answer lists the plant, the custom equipment, the dedicated people and the inventory that exist for this account alone — and states plainly who owns them and who pays if the volume never arrives. It separates the building, which somebody else may own, from the payroll and the repayment schedule, which the business cannot escape.
A weak answer treats the customer's funding of the build-out as proof the supplier's money is safe. Funding is not the same as risk.
The Mesa plant belonged to the landlord. The operating burden stayed with the tenant, and the project's costs ran well past the money advanced. Cost that only one customer can absorb is concentration risk, whether or not it has shown up yet in the revenue line.
Can price, specification or timing still move after signing?
A good answer walks through price steps, who may change engineering specifications, and who pays for delay, cancellation or scrapped material. Where the two sides already disagree, it gives both accounts and explains which clause decides.
A weak answer rests on the idea that the relationship is strategic and the customer will be reasonable. Reasonableness is not a contract term, and a far larger counterparty rarely volunteers after the fact to share a loss it never agreed to share.
GT told the bankruptcy court that the contracted prices would not have produced a profit8 and that crystals of the size required could not be grown on the agreed schedule or at the agreed cost.9 Apple's account was that it had kept funding the project after milestones were missed and received only a fraction of the sapphire committed.10 The dispute does not need a winner to be useful: a write-up that cannot describe the rights, or both versions of what they did, is not yet something to act on.
Whose product calendar decides this business's year?
A good answer admits that the customer's launch schedule, design choices and engineering tolerances set the supplier's outcomes — and then describes what the business hears from the customer's engineers and the customer's other suppliers. Asking around a customer's ecosystem is the oldest form of scuttlebutt research, and it is especially telling here, because a supplier's peers and rivals often know how a program is going before the supplier's own investors do.
A weak answer reports the customer's enthusiasm as of the last meeting.
In September 2014, Apple launched its new iPhones without sapphire screens.11 The supplier's year had been decided in a room it was not in.
If that customer disappeared tomorrow, what would the assets, the people and the price be worth?
A good answer rebuilds the year without the account and keeps the costs that would remain. It describes a second life for the capacity — other buyers, other products, a resale market for the equipment, transferable skills — and says honestly how long a pivot would take and what it would cost. On a purchase, it states where the risk is going: a lower price, money held back, or an earnout, a later payment contingent on the customer staying.
A weak answer says the customer will not leave. That is a forecast, not a plan. So is pricing last year's profit with the big account still inside it and assuming new ones appear after closing.
Sapphire furnaces configured for one buyer had few alternative uses. Much of what followed the filing was a plan to sell the machines and repay the customer out of the proceeds.
Who has actually read the governing agreement, and what were they permitted to say about it?
This is the question for anyone who must trust an analysis before acting on it. A good answer names who reviewed the contract, quotes the clauses, states what remains unknown, and does not treat confidentiality as a reason the decision-maker cannot hear the economics. Where a quality of earnings review has been done, it should say whether it examined the terms or only the invoices.
A weak answer opens with a famous name and closes with "de-risked by a long-term agreement," no clause attached. "The contract can't be discussed" is the same failure in politer clothing.
GT's agreements carried a $50 million penalty for each breach of confidentiality,12 and much of the early explanation of the filing arrived under seal.13 Analysts who had never seen the terms assigned the business a bright future on the strength of the announcement. Enthusiasm filled the space information should have occupied — which is the whole lesson. An analysis of a concentrated business that has not seen the governing contract is an analysis of a press release.
Using the list
Run these questions before a price is agreed, and again whenever one account starts to dominate a business still being run. Expect the answers in writing, with the clause sitting beside the percentage — trust in an analysis is not a feeling about its author, it is whether the thing can be checked against the obligations. Concentration itself is not disqualifying; unexamined concentration is. (The related problem of holding too much of one stock after a sale is a different kind of concentration and these questions do not answer it.)
Sources
Figures marked with a superscript were checked against these sources on November 14, 2026.
- GTAT-Apple agreement: Evolution of a fiasco - NH Business Review nhbr.com — GT Advanced Technologies was a New Hampshire maker of furnaces used to grow industrial sapphire.
- Apple signs $578M sapphire deal with GT Advanced Technology | AppleInsider appleinsider.com — In late 2013, GT Advanced Technologies announced a multi-year agreement to supply sapphire to Apple.
- GT Advanced Technologies inks game-changing deal with Apple - NH Business Review nhbr.com — Apple would advance roughly $578 million in installments to GT Advanced Technologies.
- storage.courtlistener.com — GT Advanced Technologies would operate a plant in Mesa, Arizona, leased from an Apple affiliate.
- Apple sapphire supplier files for Chapter 11 – Computerworld computerworld.com — In October 2014, GT Advanced Technologies filed for Chapter 11 bankruptcy.
- GT Advanced to exit sapphire production business, Apple to recover $439M prepayment over 4 years | AppleInsider appleinsider.com — GT's Chapter 11 filing came less than a year after the Apple agreement was announced in late 2013.
- Amendments to Regulation S-K Disclosure Requirements - Wilson Sonsini wsgr.com — not required to disclose customer's name
- GT Advanced blames Apple for one-sided sapphire contract that resulted in $461M loss | AppleInsider appleinsider.com — GT told the bankruptcy court that the contracted prices would not have produced a profit.
- GT Advanced COO Blames Apple's Strict Sapphire Contract Terms for Bankruptcy Filing macrumors.com — GT told the bankruptcy court that crystals of the size required could not be grown on the agreed schedule or at the agreed cost.
- Court unseals documents in GT Advanced case, Apple says it 'bent over backwards' to help sapphire supplier appleinsider.com — Apple's account was that it had kept funding the project after milestones were missed and received only a fraction of the sapphire committed.
- Why Apple Failed to Make Sapphire iPhones technologyreview.com — In September 2014, Apple launched its new iPhones without sapphire screens.
- GT Details Apple's 'Oppressive and Burdensome Terms', Reveals What Went Wrong iclarified.com — GT's agreements carried a $50 million penalty for each breach of confidentiality.
- GT says confidentiality pact rules out bankruptcy explanation finance.yahoo.com — Much of the early explanation of GT's bankruptcy filing arrived under seal.