NVIDIA took the stage at the Flash Memory Summit this week to show off storage technology built for the data demands of artificial intelligence. The company unveiled open-source cuFile APIs and a platform called Storage-Next, both aimed at speeding up how AI workloads access and move data.
Why Storage Matters for AI
Training large AI models requires moving massive datasets between storage and compute. That transfer can become a bottleneck, slowing down training and inference. NVIDIA's new tools are designed to remove that bottleneck by optimizing the data path. The company says the solutions address the surging data demands of AI workloads.
What's in the New Tools
The cuFile APIs are open-source, meaning developers can integrate them into their own storage systems. Storage-Next is a platform that combines hardware and software to improve data throughput. Both were demonstrated at the summit, which runs through Thursday in Santa Clara, California.
NVIDIA is making the cuFile APIs available now. Storage-Next is being shown at the company's booth. The company expects these solutions to help data centers handle the growing volume of data that AI workloads generate.
