NVIDIA GB200 NVL72
by NVIDIA · enterprise · Last verified 2026-03-17
The NVIDIA GB200 NVL72 is a liquid-cooled, rack-scale system designed for exascale AI. It connects 36 Grace Blackwell Superchips, comprising 72 B200 GPUs and 36 Grace CPUs, via fifth-generation NVLink to function as a single massive GPU for training and inferencing on trillion-parameter models with unprecedented performance and energy efficiency.
https://www.nvidia.com/en-us/data-center/gb200-nvl72/ ↗C+
C+—Average
Adoption: C+Quality: A+Freshness: A+Citations: B+Engagement: F
Specifications
- License
- Proprietary
- Pricing
- enterprise
- Capabilities
- Trillion-parameter model training, Real-time LLM inference, Second-Generation Transformer Engine, FP4 and FP6 precision support, Fifth-generation NVLink fabric, Liquid-cooled rack-scale design, Integrated Grace CPU and Blackwell GPU, High-bandwidth memory (HBM3e), Decompression engine for data processing
- Integrations
- NVIDIA CUDA Platform, NVIDIA AI Enterprise Software Suite, TensorRT-LLM, PyTorch, TensorFlow, JAX, Standard data center liquid-cooling infrastructure, Ethernet and InfiniBand networking fabrics
- Use Cases
- [object Object], [object Object], [object Object], [object Object], [object Object]
- API Available
- No
- Tags
- gpu, data-center, training, inference, blackwell, grace-blackwell, rack-scale, hpc, supercomputing, liquid-cooling, nvlink, generative-ai
- Added
- 2026-03-17
- Completeness
- 0.95%
Index Score
58.8Adoption
50
Quality
100
Freshness
98
Citations
75
Engagement
0