AI infrastructure is no longer just a contest of accelerators and raw compute.
As models handle longer context windows and more agentic workflows, the pressure is shifting toward the memory and storage systems that feed them.
Sandisk will use Future of Memory and Storage 2026 in Santa Clara to show how its latest NAND work is being positioned for that next phase of AI inference.
At a glance
- Sandisk will showcase next-generation NAND, HBF technology and enterprise SSD developments at FMS 2026 in Booth #607.
- The focus is on AI inference workloads that need more capacity, bandwidth and power efficiency across data centers and cloud environments.
- Sandisk will host a keynote on Wednesday, August 5, and join an HBF panel with Google and SK hynix on Thursday, August 6.
NAND moves closer to the center of AI inference
The next wave of AI deployment is making storage architecture more consequential.
Larger models, extended context windows, retrieval-augmented generation and agentic applications all increase the amount of data that has to be stored, moved and accessed quickly.
Sandisk’s message for FMS 2026 is that NAND is becoming a more active part of AI system design rather than a passive capacity layer.
Khurram Ismail, Sandisk’s chief product officer, framed the shift as one where efficient data access has become as important as compute for scaling inference workloads.
That is a practical point for cloud operators and data center teams, where performance per watt, capacity density and system simplicity can shape the economics of running AI at scale.
BiCS10, HBF and enterprise SSDs take the spotlight
One of the key technologies on display will be BiCS10 QLC NAND, Sandisk’s next-generation 3D NAND designed to improve density, performance and power efficiency for AI, cloud and data center storage.
The company will also demonstrate HBF, or High Bandwidth Flash, a NAND-based memory technology developed for the changing requirements of AI inference at scale.
HBF is especially notable because it sits in the conversation around the memory wall, where AI workloads can be constrained by data movement and bandwidth rather than compute alone.
Sandisk will also show enterprise SSD developments aimed at demanding AI use cases, including high-endurance configurations for KV cache workloads, future PCIe Gen 6 technology and ultra-high-capacity E3.S designs enabled by reduced DRAM placement.
Those details point to a broad strategy rather than a single product reveal.
The company is addressing fast data lakes, model storage, RAG pipelines, KV cache behavior and generated-content repositories as connected parts of the same AI data cycle.
The AI Data Cycle framework gets an update
Sandisk will present an updated version of its AI Data Cycle framework, which maps storage requirements across the AI lifecycle.
The framework spans raw data archives, fast data lakes, training, inference and repositories for generated content.
For infrastructure buyers, that framing matters because AI systems rarely stress storage in just one way.
Archives need capacity, active data lakes need speed, inference needs efficient access, and generated-content repositories add another expanding layer of storage demand.
By tying its NAND roadmap to the full lifecycle, Sandisk is positioning flash as a foundation for AI operations that extend beyond training clusters.
Keynote and panel schedule in Santa Clara
Sandisk’s FMS 2026 keynote is titled “NAND: The Versatile & Scalable Foundation of the AI Era.”
It will take place Wednesday, August 5, at 11:40 a.m. PT in the Mission City Ballroom at the Santa Clara Convention Center.
The session will feature Jim Elliott, Sandisk’s chief revenue officer, Khurram Ismail, chief product officer, and Alper Ilkbahar, chief technology officer.
The company will also join Google and SK hynix for a panel titled “Breaking the Memory Wall with High Bandwidth Flash.”
That session is scheduled for Thursday, August 6, at 9:45 a.m. PT in Ballroom D at the Santa Clara Convention Center.
Visitors to Booth #607 will be able to see demonstrations across HBF, BiCS10 NAND, enterprise SSDs, automotive storage, client SSDs and consumer products.
For anyone tracking how AI hardware is evolving beyond the accelerator, Sandisk’s FMS 2026 presence should offer a useful view of where flash memory and high-capacity storage fit into the next buildout of inference infrastructure.
