STRATA Reserve capacity
Strata/GPUs/B200

Rent NVIDIA B200 192GB

Blackwell192 GB HBM3e8.0 TB/s

The B200 is NVIDIA's Blackwell flagship. It roughly doubles H100 compute, adds FP4 for cheaper inference and pairs it with 192 GB of very fast memory.

On requestreserved clusters, 3–36 months
NVIDIA B200 192GB GPU

Specifications

GPU memory192 GB HBM3e (≈ 180 GB usable)
Memory bandwidth8 TB/s
BF16 dense≈ 2,250 TFLOPS
FP8 dense≈ 4,500 TFLOPS
Low precisionFP4 support
InterconnectNVLink 5, 1.8 TB/s

What the B200 is best for

How renting works

On-demand GPUs start in minutes and are billed per minute from a prepaid balance. Dedicated B200 clusters are allocated to you alone for 3 to 36 months at a lower, locked rate. Pay by card, bank transfer, USDT or USDC.

Questions

How much does it cost to rent an NVIDIA B200? Strata GPU vs Lambda: which GPU cloud should I choose? How many GPUs do I need to run DeepSeek V3 or R1 (671B)? How much does a reserved 8×H100 cluster cost per month? Where can I rent a dedicated 8×H100 or H200 cluster for 3 to 12 months? Can I rent H100 or B200 month‑to‑month? Which GPU to rent for models needing 80 GB, 130 GB or more VRAM

FAQ
How much faster is B200 than H100?

Roughly 2× in dense BF16 and FP8 compute, with 2.4× the memory bandwidth. Real speedups depend on the model and how well the software uses FP8 and FP4.

How do I get B200 pricing?

B200 capacity is allocated mainly as reserved clusters. Send a request with cluster size, term and region and we'll quote available capacity.

Does my CUDA code run on B200?

Yes. B200 runs the standard CUDA stack; recent versions of PyTorch, vLLM and TensorRT-LLM support Blackwell.

Other GPUs