DDN has joined the list of Nvidia AI Data Platform implementers with its Enterprise AI HyperPOD, a switch-on-and-go Infinia object storage-based system.
The turnkey system’s compute hardware uses Supermicro’s AI-optimized servers, and Nvidia’s RTX PRO 6000 Blackwell GPUs, BlueField-3 DPUs, and AI Enterprise technologies, including its NIM and NeMo Retriever microservices. HyperPOD will integrate with Oracle Cloud Infrastructure (OCI) and Google Managed Lustre, and coming Nvidia technologies (think BlueField-4). Infinia provides sub-millisecond latency, supports more than 100,000 AI calls per second, and is purpose-built for Nvidia’s H100s, GB200s, and BlueField DPUs.
DDN CTO Sven Oehme said: “With DDN Enterprise AI HyperPOD, built on Supermicro Edition, accelerated by Nvidia, we’re delivering a turnkey platform that keeps GPUs 95 percent busy, transforms infrastructure into outcomes, and scales from edge deployments to sovereign AI factories.”
DDN claims HyperPOD provides 2 PB capacity in 1 RU, and accelerates data ingestion by 22x and the KV cache stage of inference by 18x. It comes in four configuration sizes:
- XS (4 GPUs, 0.5+ PB): Instant deployment for inference and edge AI.
- Small (32 GPUs, 1+ PB): Cost-efficient AI scaling for enterprise workloads.
- Medium (64 GPUs, 3+ PB): Full-scale enterprise AI factories.
- Large (256 GPUs, 12+ PB): Exascale, sovereign-grade AI for global leaders.
Each one is pre-integrated with Supermicro hardware, Nvidia AI Enterprise software, including the new Dynamo engine for tiered KV caching, and DDN’s Infinia v2.3 software. This “sustains near-perfect GPU utilization,” DDN says.
Sovereign AI cloud provider Singtel has adopted DDN’s AI HyperPOD. Singtel is a Singaporean telecommunications conglomerate and one of Asia’s largest communications groups. DDN suggests other telcos, financial services, and manufacturing businesses could use its AI HyperPOD as well.
The AI HyperPOD is available immediately through Supermicro as turnkey packages.
Comment
It seems no enterprise storage system supplier seriously intent on selling in the enterprise AI market can ignore Nvidia, hence the rush of suppliers to support Nvidia’s latest AI Data Platform reference design, such as Hammerspace, HPE, and DDN. We imagine all the existing AI Data Platform supporters will join them.
It’s a potential gold rush as enterprises like Singtel aim to offer AI LLM and agent services, a sovereign AI MSP, and need to set up GPU+networking+storage+software infrastructure on which to run their AI services. That’s step forward DDN with a scalable, turnkey, converged infrastructure AI service appliance that customers like Singtel can turn on and go. DDN could sell a lot of these boxes.
