One hour
$5.98
1 GPU × 1h · compute only
Blackwell data center GPU
Rent NVIDIA B200 cloud capacity for frontier AI.
Direct answer
NVIDIA B200 is a Blackwell accelerator with 180 GB of HBM3e for frontier training and high-throughput inference. GPURento lists it from $5.98 per GPU-hour with self-service creation in Paris and Frankfurt after the wallet is funded.
Reviewed by the GPURento infrastructure team · September 1, 2026
Cost scenarios
What one B200 costs at the listed rate.
One day
$143.52
1 GPU × 24h · compute only
Seven days
$1004.64
1 GPU × 168h · compute only
Average month · 730h
$4365.40
1 GPU × 730h · compute only
| Memory | 180 GB HBM3e | Configuration documented in the NVIDIA HGX reference. |
|---|---|---|
| Architecture | Blackwell | Newer data center generation than Hopper. |
| On-demand rate | $5.98 / GPU-hour | Public entry rate aligned with the competitor market. |
| Available regions | Paris · Frankfurt | Selectable directly from the funded dashboard. |
Best for
- Frontier-scale training plans
- High-memory inference and long contexts
- Teams evaluating Blackwell migration
- Workloads that can exploit FP4 or newer Blackwell kernels
Not the best fit when
- Budget-sensitive experimentation
- Jobs that fit comfortably on smaller GPUs
- Projects without a validated Blackwell software stack
Decision method
What to verify before you deploy capacity.
The content below separates published specifications, GPURento catalog references and decisions that still require a workload benchmark.
01
Treat B200 as an architecture decision
A Blackwell migration should be validated against your framework, kernels, precision and parallelism strategy. Peak specifications do not predict end-to-end job time when data loading, collectives or unsupported kernels dominate.
02
Specify the system, not only the GPU
For multi-GPU B200 work, host memory, storage throughput, network fabric and GPU topology are part of the product. Record GPU count, model size, checkpoint cadence, dataset location and deadline before deployment.
- Pin the container and driver requirements
- Define the benchmark and acceptance threshold
- Choose storage and region together
03
A visible price and self-service path
The B200 catalog card exposes the $5.98 hourly entry rate and both EU regions. Funding the wallet unlocks creation directly from the dashboard.
Catalog shortlist
Relevant GPU options.
BlackwellAvailable
NVIDIA B200
180 GB HBM3e · $5.98/hr
Frontier training and inference
EU West · ParisEU Central · Frankfurt
HopperAvailable
NVIDIA H200 SXM
141 GB HBM3e · $3.59/hr
Large-model inference and HPC
EU West · ParisEU Central · Frankfurt
HopperAvailable
NVIDIA H100 SXM
80 GB HBM3 · $2.69/hr
Intensive training and FP8 inference
EU West · ParisEU Central · Frankfurt
Questions
Clear answers, including the limits.
Can I rent a B200 from GPURento without contacting sales?+
Yes. The funded dashboard exposes B200 at $5.98 per GPU-hour in Paris and Frankfurt.
How much VRAM does B200 have?+
The B200 configuration referenced on this page has 180 GB of HBM3e memory.
Should I use B200 instead of H200?+
Use B200 when Blackwell changes the economics or memory fit enough to justify its higher hourly price and migration work. H200 can remain more economical for memory-bound Hopper stacks.
Sources and scope
Last reviewed September 1, 2026. GPURento rates are current catalog references; provisioning state remains visible in the workspace, and manufacturer specifications do not substitute for workload benchmarks.
Continue the research
Deployment plan
Deploy B200 capacity as a system.
Choose the model topology, storage, region and container after funding the workspace.