CZ-B300
Czech Republic
NVIDIA B300 1,024 GPU cluster
A European facility with 128 nodes and 1,024 B300 GPUs, 800 Gbps GPU Fabric, and direct liquid cooling for large LLM training and HPC workloads.
1,024
Total GPUs
128
Nodes (8x B300 per node)
800Gbps
GPU Fabric
4.25 and up
USD/hr/GPU (5 year term)
Compute nodes
128 nodes with 8 NVIDIA B300 GPUs per node.
- GPU
- 128 units · 8x B300/node
- GPU memory
- Per the standard NVIDIA B300 specification (not stated in the source data)
- CPU
- Dual Intel 6776P
- Memory
- DDR5 6400, 3 TB or 4 TB per node
- Storage
- OS: 2x 960 GB SATA SSD; data: 4x 7.68 TB NVMe U.2
Network and cooling
Direct liquid cooling supports high bandwidth training across multiple nodes.
- Network (GPU Fabric)
- 800Gbps
- Network (External)
- Same as GPU Fabric (not listed separately)
- Cooling
- Direct liquid cooling
Note
The Tier rating, contracting entity, and data center name are provided during contracting. Pricing varies by the 3 TB or 4 TB memory configuration and payment structure. The formal quote sets the final terms.
Other facilities
Compare other facilities
TW-H100
NVIDIA H100 facility with available inventory
New Taipei, Taiwan · 24 GPU
From USD 2.30, available for immediate delivery
JP-CHIBA-H100
NVIDIA H100 plan with no down payment
Chiba, Japan · 24 GPU
From approximately USD 2.20 with no down payment
SE-H200
NVIDIA H200 facility for mixed workloads
Sweden · 296 GPU
From USD 2.88, 141 GB per GPU
Is this facility right for your model?
Share your training scale and launch schedule. Our engineering team will respond with a node configuration and estimated price range.