Return Policy

We have a 15-day Return Policy

Give Us A Call

(888) 282-1350

Email Us

Purchase Orders & Tax-Exempt

For schools, universities, government agencies, and qualified businesses — better pricing, NET-30 terms, and tax-exempt orders welcome.

✓ Tax-Exempt ✓ W-9 on Request ✓ NET-30 via Resolve

Request a quote →

Verified Global Seller

· since 2014

Independently rated on the world’s largest marketplaces. Check our reputation for yourself before you buy.

Amazon Walmart eBay Mercado Libre

Amazon across 7 markets — U.S. · Canada · Mexico · Brazil · Europe · Australia · Middle East (incl. Turkey).

Serving buyers across North America, South America, Europe, Africa & the Middle East.

Export-compliant. We verify international buyers (KYC) and ship in accordance with U.S. BIS export controls. We do not ship to embargoed, sanctioned, or restricted destinations.

NVIDIA DGX B200 AI Server

SKU B200

Ahorre 24 %

$858,000.00

$858,000.00 - $858,000.00

$858,000.00

$649,000.00

$649,000.00 - $649,000.00

$649,000.00

The NVIDIA DGX B200 AI Server is an enterprise-grade platform powered by eight NVIDIA Blackwell architecture GPUs and dual Intel Xeon Platinum processors, engineered to handle the most demanding generative AI, large language model, and deep learning workloads. Delivering up to 144 PFLOPS of FP4 Tensor Core performance and 1,440 GB of combined HBM3e GPU memory, it accelerates complex model development and high-throughput production inference.

Equipped with 5th generation NVLink interconnects and NVIDIA ConnectX-7 high-speed networking, this 10U rackmount server integrates redundant power architecture and high-bandwidth switching to provide scalable, reliable multi-node AI compute infrastructure.

Key Features

Brand	NVIDIA
Model	DGX B200
Product Type	GPU Server
GPU Architecture	NVIDIA Blackwell
GPU Configuration	8× NVIDIA B200 GPUs
GPU Memory	1,440 GB HBM3e
Aggregate GPU Memory Bandwidth	Up to 64 TB/s
GPU Interconnect	5th Generation NVIDIA NVLink
NVLink Bandwidth	Up to 14.4 TB/s aggregate
NVSwitch	2× NVIDIA NVSwitch
FP4 Tensor Core Performance	Up to 144 PFLOPS
FP8 Tensor Core Performance	Up to 72 PFLOPS
CPU	2× Intel Xeon Platinum 8570
CPU Cores	112 total cores
System Memory	2 TB, expandable to 4 TB
Cluster Networking	NVIDIA ConnectX-7, up to 400 Gb/s
Form Factor	10U rackmount
Maximum System Power	Approximately 14.3 kW
Power Supply	6× 3.3 kW power supplies with 5+1 redundancy

Highlights

✓ 8× NVIDIA Blackwell GPUs: Equipped with eight NVIDIA B200 GPUs, delivering 1,440 GB of aggregate HBM3e GPU memory and up to 64 TB/s of combined GPU memory bandwidth for demanding AI and deep learning workloads.
✓ 5th Generation NVIDIA NVLink: Features fifth-generation NVLink connectivity with two NVIDIA NVSwitch systems, providing up to 14.4 TB/s of aggregate GPU interconnect bandwidth for high-speed GPU-to-GPU communication.
✓ Accelerated AI Training: Delivers up to 3× higher AI training performance compared with NVIDIA DGX H100, enabling faster model development, training, and experimentation.
✓ High-Performance AI Inference: Delivers up to 15× higher inference performance compared with NVIDIA DGX H100, supporting demanding production inference and generative AI applications.
✓ Blackwell Tensor Core Acceleration: Provides up to 144 PFLOPS of FP4 Tensor Core performance and up to 72 PFLOPS of FP8 Tensor Core performance for accelerated AI computing.
✓ Large-Scale GPU Memory: With 1.44 TB of aggregate HBM3e memory across eight GPUs, DGX B200 is designed to handle memory-intensive AI models and workloads requiring substantial GPU resources.
✓ Enterprise-Class CPU & System Memory: Incorporates two Intel Xeon Platinum 8570 processors with 112 total CPU cores and 2 TB of system memory, expandable to 4 TB.
✓ High-Speed Cluster Networking: Provides NVIDIA ConnectX-7 networking with connectivity up to 400 Gb/s, enabling high-performance communication for multi-node AI infrastructure.
✓ Integrated NVMe Storage: Includes dedicated NVMe storage for the operating system and internal data-cache workloads, providing fast local storage for the AI computing environment.
✓ Enterprise AI Platform: DGX B200 is designed as a unified infrastructure platform for AI development, model training, fine-tuning, inference, and deployment.
✓ Built for Generative AI & LLMs: Optimized for large language models, generative AI, foundation models, deep learning, recommendation systems, AI analytics, and other compute-intensive workloads.
✓ Scalable AI Infrastructure: Can serve as a high-performance building block for larger NVIDIA DGX BasePOD and DGX SuperPOD environments, enabling organizations to scale AI compute across multiple systems.
✓ Enterprise-Ready Architecture: Combines high-density GPU compute, large GPU memory, high-bandwidth GPU interconnects, high-speed networking, redundant power, and integrated management capabilities in a 10U rackmount platform.