SanDisk Adopts Long-Term NAND Contracts Driven by AI Infrastructure Demand

SanDisk is transitioning its business model from volatile spot-market pricing to multi-year supply agreements because data center demand for AI-driven NAND flash memory is now the company's primary growth engine. This shift aims to stabilize revenue and align production capacity with the long-term needs of hyperscale infrastructure providers.

Data center demand now accounts for more than half of the total addressable market for NAND. Management describes this as a fundamental change for both pricing structures and customer behavior. By moving away from transactional, spot-market negotiations, the company is securing more predictable long-term utilization of its manufacturing facilities.

"NAND is the most scalable semiconductor technology," said David Goeckeler, Chief Executive of SanDisk, during the Goldman Sachs Communacopia conference. "Inference workloads will need more bandwidth and endurance as model context lengths grow and systems rely more on key-value cache storage and retrieval-augmented generation (RAG)."

Securing Future Bit Supply

The move toward contractual stability is already underway. SanDisk has signed eight long-term agreements with major data center customers over the last nine months. These contracts cover approximately two-thirds of the company's projected fiscal 2028 bit supply, providing a hedge against the historical cyclicality of the memory market.

Product Development and HBF Timeline

To meet the performance requirements of modern AI systems, SanDisk is accelerating the development of High Bandwidth Flash (HBF). The technology is currently in the die-taping phase, which is the final stage of design before physical manufacturing begins. The company is targeting customer sampling for 2025, with full commercialization planned for the subsequent years.

Why the Shift Matters for Memory Markets

Historically, the NAND industry has been defined by boom and bust cycles driven by spot pricing, where manufacturers frequently faced oversupply or shortages based on consumer electronics demand. By pivoting toward the data center segment, SanDisk is attempting to treat memory as a core infrastructure component rather than a commodity.

This strategy relies on the technical necessity of RAG, a framework where AI models retrieve data from external sources to improve response accuracy. As these systems scale, they require massive, high-endurance storage architectures that can handle frequent read-write cycles, effectively locking in demand for the next generation of NAND products.