New NVIDIAHGX B300 rack and cluster solutions — now with integrated NVIDIA AI Data Platform support — give enterprises the complete AI infrastructure stack to train, reason, and retrieve at scale.

The Infrastructure Gap Enterprise AI Has Been Waiting to Close

Most enterprises are not held back by a lack of AI ambition. They are held back by infrastructure that was never designed for what AI demands — compute at cluster scale, data that is continuously indexed and instantly retrievable, and a software stack that ties both together without months of integration work.

Supermicro is addressing that directly on two fronts: AI Factory rack and cluster solutions built on NVIDIA HGX B300, and integrated support for the NVIDIA AI Data Platform that connect model workloads to continuously indexed, searchable enterprise knowledge.

Introducing the Supermicro HGX B300 AI Factory Clusters

Three new rack and cluster-scale AI Factory solutions anchor the portfolio, each built on the SYS-822GS-NB3RT — an 8U air-cooled server housing eight NVIDIA Blackwell Ultra GPUs with up to 2.1TB of HBM3e memory per node. These systems are purpose-built for the most demanding AI workloads enterprises face today: high-volume inference, foundation model training, large-scale agentic AI, and FP32 HPC.

The three cluster configurations with Supermicro’s NVIDIA-Certified Systems are designed to meet organizations wherever they are in their AI journey:

• SRS-48AC-4N-B300SX — 4-node / 32-GPU cluster, each GPU delivering 20 petaFLOPS of FP4 sparse inference performance for enterprises scaling their first production AI deployments.
• SRS-48AC-8N-B300SX — 8-node / 64-GPU cluster, optimized for organizations running multiple concurrent AI workloads including inference and model fine-tuning.
• SRS-48AC-32N-B300SX — 32-node / 256-GPU cluster, built for enterprises training foundation models and running large-scale agentic AI pipelines at full production capacity.

All three configurations run NVIDIA Spectrum-X Ethernet with eight integrated NVIDIA ConnectX-8 SuperNICs at 800Gb/s per node, and a NVIDIA BlueField-3 DPU providing intelligent, secure data plane management — with NVIDIA AI Enterprise software to develop and deploy AI in production. These infrastructure cluster configurations are endorsed by NVIDIA, based on the NVIDIA Enterprise Reference Architecture for HGX B300, and validated through Supermicro's complete L12 system and cluster-level testing before shipment — so customers get solutions that are ready to generate revenue from day one.

Closing the Loop: AI Data Platform Meets the AI Factory

Raw compute power alone is not enough. The most capable AI model still depends on having fast, secure access to relevant, up-to-date enterprise data. That is why Supermicro's HGX B300 AI Factory solutions are designed to work hand-in-hand with the NVIDIA AI Data Platform — a reference design that transforms enterprise storage systems into intelligent knowledge retrieval engines for RAG, semantic search, and AI agents.

Running on Supermicro's storage infrastructure, the NVIDIA AI Data Platform continuously ingests, embeds, and indexes unstructured enterprise data — documents, emails, images, video — without duplicating or moving source files. The result: AI applications running on HGX B300 clusters can access the full breadth of enterprise knowledge in near real time, grounded in current data, with existing security and access controls fully intact.

This integration is a meaningful step forward for enterprise customers who have struggled with the gap between their AI compute investments and the quality and relevance of the data those systems can actually access. Supermicro's AI Factory now addresses both sides of that equation.

A Complete Ecosystem of Storage Software Partners

The NVIDIA AI Data Platform on Supermicro is not a single-ISV story. Supermicro storage servers are validated with leading data platform and file system software partners — including DDN, VAST Data, Everpure, IBM Storage Scale, Cloudian, WEKA, and Nutanix — giving enterprises the freedom to deploy AIDP on their preferred storage software stack without sacrificing performance or compatibility.

Whether an organization needs the extreme throughput for NVMe-intensive AI embedding workloads, a unified platform for multimodal data, or a proven parallel file system for regulated industries, Supermicro hardware provides a consistent, validated foundation underneath all of them.

First to Market. Built to Last.

Supermicro has a long track record of being one of the first to production with new NVIDIA accelerated computing platforms, and HGX B300 is no exception — with AI Factory cluster solutions built on NVIDIA Enterprise Reference Architectures. With U.S. production capacity exceeding 5,000 racks per month and complete L12 cluster-level validation before every shipment, enterprises can move from purchase order to productive AI workloads faster with Supermicro than any other vendor in the market.

Supermicro controls quality, compatibility, and supply chain integrity end to end — from the GPU node to the storage system to the network fabric — giving IT teams the confidence to deploy at scale without the risk of multi-vendor integration complexity.

See It in Action

Supermicro experts are available to walk through the NVIDIA HGX B300 AI Factory cluster configurations and NVIDIA AI Data Platform integration at NVIDIA GTC 2026 in San Jose — helping you design the right architecture for your organization's specific workloads and data environment.

Explore the full AI Factory portfolio at supermicro.com/aifactory, or contact our sales team to start your deployment planning.