NVIDIA B200 rental price per hour
192GB HBM3e VRAM · Blackwell architecture · released 2025
The cheapest way to rent an NVIDIA B200 by the hour in June 2026 is about $3.50/hr on Vultr Cloud GPU. We track 9 providers publishing an on-demand price for it , and spot/community rates reach as low as $2.69/hr. Frontier-scale training and inference. The fastest broadly-rentable single GPU in 2026; large memory and big generational throughput gain over H100/H200.
Source: Provider pricing pages. Data as of June 2026.
NVIDIA B200 price by provider
| Provider | Type | On-demand /hr | Spot/community /hr | Notes |
|---|---|---|---|---|
| Vultr Cloud GPU | specialist | $3.50/hr | — | HGX B200, per GPU. |
| Vast.ai | marketplace | $4.23/hr | — | Marketplace median (192GB). Host-set. |
| RunPod | specialist | $5.89/hr | — | Secure Cloud on-demand. |
| Verda (formerly DataCrunch) | specialist | $6.11/hr | — | On-demand; B300 $7.50, GB300 $8.62 also listed. |
| Lambda | specialist | $6.99/hr | — | On-demand; 8x SXM6 also seen $6.69/GPU. |
| Nebius | specialist | $7.15/hr | $3.95/hr | HGX per GPU; reserved $3.95. B300 $7.85 also listed. |
| Together AI | specialist | $8.19/hr | — | GPU Cluster per-GPU; Dedicated Inference 1x $11.95; reserved from $6.79. |
| CoreWeave | specialist | $8.60/hr | — | HGX B200, per GPU. (B300/GB300 contact-sales.) |
| AWS EC2 | hyperscaler | $14.24/hr | $2.69/hr | Derived: p6-b200.48xlarge $113.93/hr / 8 GPUs. |
| Crusoe | specialist | — | — | Contact-sales (no public hourly). |
| Google Cloud | hyperscaler | — | — | a4-highgpu-8g; a single source implied ~$4.28/GPU but it is anomalously low and unverified, so we omit it. |
| Hyperstack | specialist | — | — | Catalog-listed but contact-sales for self-serve hourly. |
| Microsoft Azure | hyperscaler | — | — | ND GB200 v6 (NVL72) has no public per-GPU price. |
| Paperspace (DigitalOcean) | specialist | — | — | Only HGX B300 reserved (12-mo $5.65/GPU). |
Source: Provider pricing pages. Data as of June 2026.
Sorted cheapest on-demand first; "—" means the provider does not publish a price for this GPU (see the note). Snapshot June 2026 — cloud GPU prices change weekly; verify on the provider's pricing page before relying on a figure.
About the NVIDIA B200
Frontier-scale training and inference. The fastest broadly-rentable single GPU in 2026; large memory and big generational throughput gain over H100/H200.
- VRAM: 192GB HBM3e
- Architecture: Blackwell
- Released: 2025
- Cheapest on-demand: $3.50/hr on Vultr Cloud GPU
- Cheapest spot/community: $2.69/hr on AWS EC2
Frequently asked questions
How much does it cost to rent an NVIDIA B200 per hour?
The cheapest on-demand NVIDIA B200 we track is $3.50/hr on Vultr Cloud GPU, with 9 providers publishing a price. Spot/community rates go as low as $2.69/hr. Snapshot June 2026 — cloud GPU prices change weekly; verify on the provider's pricing page before relying on a figure.
Which provider is cheapest for the NVIDIA B200?
Vultr Cloud GPU at $3.50/hr is the cheapest on-demand option in our June 2026 snapshot. Specialist neoclouds and marketplaces usually beat the hyperscalers (AWS/GCP/Azure) by 2-4x on the same GPU.
What is the NVIDIA B200 good for?
Frontier-scale training and inference. The fastest broadly-rentable single GPU in 2026; large memory and big generational throughput gain over H100/H200. It has 192GB HBM3e of VRAM and uses the Blackwell architecture (released 2025).
Is spot or on-demand cheaper for the NVIDIA B200?
Spot, community and interruptible tiers can cut the NVIDIA B200 price roughly in half — the cheapest discounted rate we track is $2.69/hr versus $3.50/hr on-demand. The trade-off is your instance can be reclaimed, so use spot for fault-tolerant or checkpointed jobs.
Related GPUs
- AMD Instinct MI300X (192GB HBM3)
- NVIDIA H100 PCIe (80GB HBM2e)
- NVIDIA H100 SXM (80GB HBM3)
- NVIDIA H200 (141GB HBM3e)
Keep exploring
Source & accuracy
Figures are a dated snapshot from each provider's published pricing page (as of June 2026). Hyperscaler per-GPU prices are derived from 8-GPU instance prices (instance price / 8) and include bundled vCPU/RAM; marketplace (Vast.ai) figures are moving medians, not list prices. We do not invent numbers — where a value is unknown we show "—" with a reason. See our methodology. This is an informational comparison, not a quote.
Last updated: 2026-06-21