NVIDIA · Blackwell · Data centre
NVIDIA B200
NVIDIA's Blackwell data-centre GPU, sold in 8-GPU HGX B200 and DGX B200 systems, adding FP4 support and fifth-generation NVLink.
01 / SPECIFICATIONS
What the hardware holds
- Memory
- 180 GB
- Memory bandwidth
- 7.7 TB/s
- FP8 dense
- 4,500 TFLOPS
- FP16 / BF16 dense
- 2,250 TFLOPS
- Variant
- B200 (HGX B200)
- Memory type
- HBM3e
- Form factor
- SXM (HGX B200 8-GPU baseboard)
- Interconnect
- NVLink 5 1.8 TB/s
- Max power
- 1,000 W
Source notes for B200 (HGX B200)
NVIDIA Blackwell datasheet (Oct 2025), 'Individual Blackwell GPU Specifications', HGX B200 column - per-GPU values printed directly, no division needed (system totals for 8 GPUs: 1.4 TB memory, 62 TB/s, 72 PFLOPS FP8 sparse). Datasheet lists FP16/BF16 Tensor 4.5 PFLOPS and FP8/FP6 9 PFLOPS with sparsity (footnote 2: dense is one-half); dense = 2.25 / 4.5 PFLOPS. TDP configurable up to 1,000 W. Also PCIe Gen5 128 GB/s. Linked from https://www.nvidia.com/en-us/data-center/hgx/.
Compute figures are dense, without sparsity. NVIDIA source for B200 (HGX B200) Checked 8 Nov 2026.
02 / MEMORY FIT
How many do you need to hold a model?
| Model size | FP16 / BF16 | FP8 | INT4 |
|---|---|---|---|
| 8B parameters | 1 GPU16 GB of weights | 1 GPU8 GB of weights | 1 GPU4 GB of weights |
| 70B parameters | 1 GPU140 GB of weights | 1 GPU70 GB of weights | 1 GPU35 GB of weights |
| 405B parameters | 6 GPUs810 GB of weights | 3 GPUs405 GB of weights | 2 GPUs203 GB of weights |
GPUs = parameters × bytes per parameter × 1.2 headroom ÷ 180 GB, rounded up. This covers weights only. Long contexts, large batches, and training need much more memory.
03 / OFFERS
Reviewed on-demand rates
| Provider | Configuration | Status | Source | ||
|---|---|---|---|---|---|
 Runpod | Variant not stated1× B200 Pod · 180 GB | $7.99$7.99 ÷ 1 GPU | $7.991 GPU | Stock not verified | Rate card |
 Nebius | B200 (HGX B200)HGX B200 · published per GPU rate | $8.50Published per GPU-hour | Minimum GPU count varies | Stock not verified | Rate card |
 CoreWeave | B200 (HGX B200)8× HGX B200 node · 180 GB/GPU · North America | $8.60$68.80 ÷ 8 GPUs | $68.808 GPUs | Stock not verified | Rate card |
Per GPU-hour divides a machine's hourly price by its GPU count, so single GPUs and 8-GPU nodes compare fairly. The billed price is what one machine costs per hour. Storage, network transfer, and minimum terms can add to the bill.
04 / SELLERS
Who lists the B200
- RunpodGPU cloud · Self-serve · Reviewed rate
LambdaGPU cloud · Self-serve- CoreWeaveGPU cloud · Sales-led · Reviewed rate
- NebiusGPU cloud · Self-serve · Reviewed rate
CrusoeGPU cloud · Self-serve
Vast.aiMarketplace · Self-serve
Lambda, Crusoe, Vast.ai list the B200 without a rate we have reviewed. Ask them for a quote, especially for clusters of 8 or more GPUs.
05 / COMPARE WITH