NVIDIA H200 HGX SXM5 8-GPU Baseboard – 8x H200 141GB AI Accelerator Board

The NVIDIA HGX H200 SXM5 8-GPU is a next-generation AI and HPC accelerator platform built for the most demanding data center workloads. This baseboard is fully populated with 8x NVIDIA H200 141GB SXM5 GPUs, delivering a total of 1,128GB of HBM3e GPU memory for large-scale AI training, generative AI, inference, HPC, and data-intensive computing.

The H200 platform is designed for workloads that require more memory capacity and higher memory bandwidth, making it especially valuable for large language models, retrieval-augmented generation, scientific computing, model serving, and advanced analytics. It is commonly deployed in high-density GPU servers and AI infrastructure where memory capacity is a major performance factor.

Important: This listing is for the NVIDIA H200 8-GPU baseboard only. It is not a complete server and does not include chassis, CPUs, memory, storage, power supplies, cooling infrastructure, or system-level configuration unless otherwise specified. | 1 year standard NVIDIA warranty included.

Availability, Pricing & Custom Configuration Notice

Due to market volatility, pricing and lead times may vary. Current estimated lead time is approximately one week, but availability can change quickly based on supply, demand, and configuration requirements. For bulk orders, custom requests, integration support, or configuration planning, please Request A Quote or Contact Catalyst Data Solutions before ordering.

Features and Benefits

• Fully populated 8-way NVIDIA H200 SXM5 GPU baseboard
• 8x H200 141GB GPUs for a total of 1,128GB HBM3e GPU memory
• Designed for generative AI, LLMs, inference, HPC, and data-intensive workloads
• HBM3e memory supports larger models and higher-throughput AI workloads
• Built for NVIDIA HGX-compatible server platforms
• New/sealed NVIDIA-branded hardware
• Ideal for enterprise AI infrastructure and high-density GPU clusters

Technical Specifications

• Product Type: GPU Baseboard
• Platform: NVIDIA HGX H200
• GPU Model: NVIDIA H200 SXM5
• GPU Count: 8
• Memory Per GPU: 141GB
• Total GPU Memory: 1,128GB
• Memory Type: HBM3e
• Form Factor: SXM5 Baseboard
• Interconnect: NVIDIA NVLink / NVSwitch Architecture
• Condition: New / Sealed
• Warranty: 1-Year Standard NVIDIA Warranty
• Product Scope: Baseboard Only, Not a Complete Server

Common Use Cases

• Large language model training and deployment
• Generative AI and foundation model workloads
• AI inference and model serving at scale
• High-performance computing and scientific simulation
• Large dataset analytics and data-intensive compute
• Enterprise AI infrastructure and GPU cluster expansion
• Cloud GPU and accelerated computing platforms

Data Sheet

📂 NVIDIA H200 Data Sheet

Compatibility & Popular Deployment Environments

The NVIDIA HGX H200 SXM5 8-GPU Baseboard is commonly deployed in supported HGX-compatible server platforms. Compatibility depends on the server chassis, thermal design, power delivery, firmware, motherboard support, and system integration requirements.

Commonly Compatible Platform Types

• NVIDIA HGX H200-compatible GPU servers
• 8-GPU SXM5 AI server platforms
• Enterprise generative AI systems
• High-performance computing clusters
• OEM HGX platforms from major server manufacturers

Popular Deployment Environments

• Generative AI and LLM infrastructure
• AI inference and model serving clusters
• Research and scientific computing environments
• Cloud GPU compute platforms
• Data center GPU compute nodes
• Enterprise AI and analytics platforms

Frequently Paired With

• HGX-compatible 8-GPU server chassis
• High-wattage redundant power supplies
• High-performance CPUs and DDR5 system memory
• NVMe storage arrays
• InfiniBand or 100G/200G/400G Ethernet networking
• Enterprise cooling infrastructure