B100 GPU
The NVIDIA B100 is a Blackwell architecture GPU designed for data center AI and HPC workloads, offering high performance with lower power consumption than B200.
Cloud Pricing
Prices updated daily. Last check: 4/8/2026
Performance
Strengths & Limitations
- 192 GB VRAM capacity supports large model training and inference workloads
- 8,000 GB/s memory bandwidth enables high-throughput data processing
- Second-Generation Transformer Engine with NVFP4 quantization optimizes AI model performance
- Fifth-generation NVLink supports scaling up to 576 GPUs with 130TB/s aggregate bandwidth
- Ultra Tensor Cores deliver 3,500 TFLOPS FP16 performance for AI computations
- NVIDIA Confidential Computing provides security features for sensitive workloads
- Dedicated RAS Engine enhances system reliability in data center deployments
- 700W TDP requires substantial power infrastructure and cooling systems
- High memory capacity may be excessive for smaller AI models and inference workloads
- Server-only form factor limits deployment to data center environments
- Blackwell architecture ecosystem may have limited software optimization compared to mature platforms
Key Features
About B100
Common Use Cases
The B100 is designed for enterprise AI workloads requiring substantial memory capacity and compute performance. Its 192 GB VRAM makes it suitable for training and serving large language models, particularly trillion-parameter models that exceed the memory constraints of smaller GPUs. The high memory bandwidth and Ultra Tensor Cores optimize performance for generative AI applications, while NVLink scaling capabilities support distributed training across multiple nodes. The GPU also serves data analytics workloads that benefit from large memory capacity and the dedicated Decompression Engine for data processing acceleration.
Full Specifications
Hardware
- Manufacturer
- NVIDIA
- Architecture
- Blackwell
- TDP
- 700W
Memory & Performance
- VRAM
- 192GB
- Memory Bandwidth
- 8000 GB/s
- FP32
- 60 TFLOPS
- FP16
- 3500 TFLOPS
- BF16
- 1750 TFLOPS
- FP8
- 3500 TFLOPS
- Release
- 2024
Frequently Asked Questions
How much does a B100 cost per hour in the cloud?
B100 pricing varies by provider, region, and commitment level. Check the pricing table above for current rates across all providers.
What is the B100 best used for?
The B100 excels at large language model training and inference, particularly for models requiring substantial memory capacity. Its 192 GB VRAM and Second-Generation Transformer Engine make it well-suited for generative AI applications, trillion-parameter model serving, and distributed training workloads that benefit from NVLink scaling capabilities.
How does the B100 compare to the H100 for AI workloads?
The B100 features newer Blackwell architecture with Ultra Tensor Cores and Second-Generation Transformer Engine compared to the H100's Hopper architecture. The B100 provides 192 GB VRAM versus the H100's 80 GB, offering significantly more memory for large model training. Both support similar NVLink scaling, but the B100 includes NVFP4 quantization and enhanced Transformer Engine capabilities for improved AI performance.