Jason Hardy is vice president of storage technology at NVIDIA, where he drives storage architecture and ecosystem alignment for accelerated computing and enterprise AI. He’s led the development of high-performance storage systems and AI solutions supporting analytics, enterprise, and industrial workloads. He works across engineering, product, and industry partners to advance storage architectures designed for scale, performance, and reliability in AI environments.
The AI industry focus is shifting from model training to inference systems that deliver customer value. This transition is changing the design of new data centers built specifically for AI workloads, placing new demands on memory and storage. As the inventor of NAND and 3D flash memory, KIOXIA continues to deliver innovations in flash memory, SSDs, and software for every application, and especially AI. KIOXIA BiCS FLASH™ 3D memory deliver advances in performance, density, and reliability, which enable SSDs to address workloads for today and the future. KIOXIA memory and SSD leaders will share the latest advancements in the technology, and how they deliver performance and scaling advantages in new AI-centric data centers.
As LLM model sizes, context lengths, concurrent sessions, and active user counts continue to grow exponentially, the memory and storage hierarchy demands a fundamental redesign. Cohesively co-designing KV cache and SSD storage with context awareness will be critical to building scalable, efficient AI infrastructure. In this presentation, we will share our perspectives on optimizing SSD controller chip and firmware design in concert with KV cache systems — such as CMX — to help shape the future of AI inference.