Why Storage Matters to AI in 2025

Explore the critical role of storage in AI systems through 2025 in this 48-minute Tech Field Day presentation by Solidigm. Dive into the six-stage AI data pipeline—ingestion, data preparation, model training, fine-tuning, inference, and archiving—and understand how each stage places unique demands on hardware infrastructure. Learn how storage choices between direct-attached and network-attached solutions significantly impact AI workloads, with network storage gaining importance due to increasing network bandwidth capabilities and growing model complexity. Director of Market Development Ace Stryker and Metrum AI CEO Steen Graham demonstrate a real-world collaboration showing how SSD-based offloading enables running 70 billion parameter models on less powerful hardware, potentially matching DRAM performance in retrieval augmented generation while reducing costs and hardware requirements. Recorded in Santa Clara on April 23, 2025, this presentation provides valuable insights for organizations looking to optimize their AI infrastructure strategy.