Powering a new class of enterprise infrastructure for agentic AI.

The NVIDIA AI Data Platform is a customizable reference design that integrates NVIDIA-accelerated computing into enterprise storage to centralize intelligent data handling and deliver AI-ready data, while reducing latency, enhancing data security, and maximizing performance.

These data platforms are built by enterprise storage leaders to improve the performance and accuracy of agentic AI workflows. The platform features NVIDIA Blackwell GPUs, BlueField®-3 DPUs, Spectrum™-X Ethernet networking, and NVIDIA AI Enterprise software.

Fireside Chat With Jensen Huang and NetApp CEO George Kurian

Delivering AI-Ready Enterprise Data With GPU-Accelerated AI Storage

Transform enterprise storage systems to deliver AI-ready data.

The NVIDIA Blackwell architecture provides powerful acceleration for AI workloads, optimizing data processing pipelines for agentic AI. These pipelines enhance ingestion, retrieval, and query workflows by delivering high-performance compute for large-scale AI models.

The NVIDIA BlueField-3 DPU accelerates data access across storage initiators, controllers, and targets, enabling multiple GPUs to efficiently share the same data stores. This optimizes data throughput and access for AI reasoning, reducing CPU overhead, enhancing data availability, and ensuring secure, efficient scaling of AI workloads.

Through RDMA Over Converged Ethernet (RoCE) adaptive routing and optimized congestion control, NVIDIA Spectrum-X Ethernet accelerates storage performance by nearly 50% and reduces communication bottlenecks. With it, enterprises can efficiently scale AI applications while maximizing AI system utilization.

NVIDIA NIM™ is a set of easy-to-use inference microservices for accelerating the deployment of foundation models and keeping data secure. NVIDIA NIM is optimized for enterprise-scale inference.

NVIDIA NeMo™ Retriever enhances enterprise AI with retrieval-augmented generation (RAG). It combines optimized models, vector search, and multi-GPU performance to accelerate and scale data retrieval and improve large language model (LLM) accuracy.

NVIDIA cuVS is an open-source library for GPU-accelerated vector search and data clustering that enables faster vector searches and index builds, letting developers enhance data mining and semantic search workloads.