DGX GB200 NVL72
DGX GB200 NVL72 is a NVIDIA AI system with 72 NVIDIA B200 and 36 NVIDIA Grace.
Built from BenchGecko data as of April 14, 2026 · updates when the data changes
DGX GB200 NVL72 is a NVIDIA AI system with 72 NVIDIA B200 and 36 NVIDIA Grace.
Basic
NVIDIA's flagship liquid-cooled rack. 72 Blackwell GPUs + 36 Grace CPUs connected via NVLink 5.0 in a single 72-GPU domain. Designed for trillion-parameter model training. The most powerful commercially available AI system, requiring liquid cooling infrastructure and 120+ kW per rack.
Deep
The dataset lists 720 PFLOPS of FP8 compute and 360 PFLOPS at BF16. It has 13.8 TB of HBM3e. Interconnect: NVLink 5.0. Power: 120 kW, liquid cooled.
Expert
Known deployments: Microsoft Azure, Oracle Cloud, CoreWeave and xAI. Figures come from spec sheets and disclosures in the curated systems dataset, dated on the systems page.
Depending on why you're here
- ·72 × NVIDIA B200
- ·Interconnect: NVLink 5.0
- ·liquid cooling · 120 kW
- ·Compare systems on /systems/nvidia-dgx-gb200-nvl72
- ·Deployed by Microsoft Azure, Oracle Cloud, CoreWeave and xAI
- ·DGX GB200 NVL72 is a big AI computer built from many NVIDIA B200 chips
Frequently Asked Questions
Read the primary sources
- NVIDIA DGX GB200 NVL72 spec sheetwww.nvidia.com
- NVIDIA GTC 2025 keynotewww.nvidia.com