DDN, Nvidia team up to cut inference costs and boost GPU utilization

DDN and Nvidia have partnered to address the growing need for cost-effective and efficient AI infrastructure, particularly for inference workloads. DDN’s storage systems, including the Infinia and EXAScaler offerings, are being integrated with Nvidia’s hardware and software – specifically BlueField-4 DPUs and the STX reference architecture – to deliver direct GPU-to-data paths, reduce latency, and increase GPU utilization. The collaboration aims to lower the cost per token for AI inference by minimizing GPU idle time and eliminating data bottlenecks.

Central to this effort is DDN’s Horizon platform, an orchestration layer designed to operationalize AI-as-a-Service across the AI lifecycle. Horizon provides a unified control plane for managing compute, data services, and networking, enabling self-service provisioning, tenant isolation, and integrated billing. DDN is also launching IndustrySync Pipelines, pre-integrated AI blueprints tailored for specific verticals like Financial Services and Life Sciences, to accelerate deployment and improve performance.

Furthermore, DDN is partnering with Zadara for sovereign cloud deployments and Aleria to offer a complete Sovereign AI Factory reference architecture based on Nvidia’s Vera Rubin design. These solutions prioritize data residency, security, and auditability for governments and regulated industries, allowing them to leverage advanced AI capabilities while maintaining control over their data and infrastructure. Deployments are already underway in locations like Indonesia and Abu Dhabi, with plans for expansion to multiple regions over the next two years.

(Source：Blocks And Files)