DDN launches AI storage for inference era

The introduction of DDN Enterprise AI HyperPOD marks a shift in enterprise AI storage, focusing on the inference era where LLMs need more data and storage. The collaboration with Supermicro and Solidigm addresses the complexity customers face when integrating separate components, offering a turnkey solution based on Nvidia's reference design. This approach reduces the time to monetize data and supports on-premises deployment for security and governance.

A key advantage is scalability—customers can start with a single rack and expand as needed, avoiding large upfront GPU investments. The use of SSDs is highlighted for their ability to reuse AI workloads without recomputation, improving GPU utilization. This is particularly relevant as AI workloads become storage-bound rather than compute-bound. The panel also noted a growing trend toward sovereign AI, where organizations want control over their data and IP, which the HyperPOD supports through flexible on-premises and hybrid deployment options.