Micron's take-or-pay contracts face $1T test if AI demand cools

The same long-term contracts that gave Micron Technology and Oracle "unprecedented revenue visibility" may prove unenforceable if AI demand falters — a risk the Bank for International Settlements warns could amplify a $1 trillion supply-chain overhang.

The AI boom has produced a cascade of long-term supply agreements across semiconductors, cloud computing and hardware manufacturing, with companies touting the contracts as evidence of durable revenue growth. But a Wall Street Journal analysis published this month questions whether those commitments would hold if demand reverses, citing the industry's own history of voiding similar pacts during the last downturn.

"Long-term contracts help improve the market's perception of the memory industry," an SK Hynix executive told analysts in April, after the company's New York listing. The reality, according to the analysis, is that such agreements have rarely been enforced when buyers no longer need the product.

Micron has been the most aggressive in locking in customers. Its "strategic customer agreements" typically run five years with take-or-pay provisions — meaning buyers must pay even if they don't take delivery. Chief Executive Officer Sanjay Mehrotra told analysts last month that these contracts will account for more than half of Micron's revenue going forward. The company's stock has roughly tripled this year, mirroring gains at SK Hynix and a near-doubling at Samsung Electronics.

The risk extends well beyond memory chips. Oracle reported $638 billion in remaining performance obligations as of last quarter, a figure Chief Financial Officer Hilary Maxson described as providing "excellent visibility into our future revenue growth, backed by long-term contract customer commitments." Across Google, Microsoft, Amazon and Oracle, aggregate revenue backlog has more than doubled since mid-2025, swelling by over $1 trillion.

Why Contracts Break When Demand Turns

The logic against enforcement is straightforward. If a customer's AI workloads shrink before a contract expires, chip suppliers face a choice: force delivery of products the buyer cannot use, or renegotiate. Forcing delivery merely shifts inventory to the customer's warehouse, delaying future purchases and straining a relationship that may span decades.

History provides a clear precedent. During the pandemic-era chip shortage, long-term commitments proliferated as buyers scrambled for supply. When the shortage flipped to a glut, nearly all those contracts were renegotiated or voided. Microchip Technology launched a "Preferred Supplier Program" in 2021 requiring long-term commitments, only to cancel it when demand reversed. "We will not force customers to buy anything they don't need," Chief Executive Steve Sanghi said in November.

BIS Warns of Amplified Overinvestment

The Bank for International Settlements, in its annual economic report this month, identified the contract structure as a potential amplifier of AI-driven overinvestment. Companies seeking to lock in future capacity through long-term agreements may become more vulnerable if demand falls short of expectations, the BIS said. Lenders and investors funding those contracts could face unexpected losses.

The supply chain's layered structure compounds the risk. AI developers such as OpenAI sign compute contracts with cloud providers including Oracle and CoreWeave; those providers order chips from manufacturers like Nvidia and Micron; chipmakers contract with Taiwan Semiconductor Manufacturing Co. for wafers; and TSMC commits to equipment purchases from ASML Holding. A demand shortfall at any link could cascade through the chain.

For investors, the question is whether the market is pricing in any probability of contract renegotiation. Micron trades at elevated multiples reflecting the "unprecedented visibility" narrative. If AI infrastructure spending decelerates — whether from a macroeconomic shock, a technology breakthrough that reduces compute requirements, or simply a digestion period after the current buildout — the contracts underpinning that narrative could prove far less binding than advertised. The BIS warning suggests this is not a tail risk but a structural feature of how the AI supply chain has been financed.

This article is for informational purposes only and does not constitute investment advice.