8 providers · per GPU-hour
NVIDIA B200 cloud pricing
8 offers from 8 providers, 7 with a published on-demand price. Per GPU-hour, US$. Prices checked 28 Sept 2026.
- Lowest on-demand
- $5.04/GPU-hr
- Lowest spot
- $3.95/GPU-hr
- Memory
- 180 GB HBM3e
Spec sheet: B200 ↗
| # | Provider | Offer | On-demand | Spot | Committed | As published (original currency) · source | Actions |
|---|---|---|---|---|---|---|---|
| 1 | UpCloud | B200 x1 1× B200 · 12 vCPU · 240 GB RAM | $5.04/GPU-hr | — | — | ||
| 2 | Hyperstack | B200 x1 1× B200 | $6.00/GPU-hr | — | — | ||
| 3 | Verda (formerly DataCrunch) | B200 x1 1× B200 | $6.29/GPU-hr | — | — | ||
| 4 | E2E Networks | B200 x1 1× B200 · 32 vCPU · 400 GB RAM | $6.99/GPU-hr | — | — | ||
| 5 | Nebius | B200 x1 1× B200 | $7.15/GPU-hr | $3.95/GPU-hr | — | ||
| 6 | CoreWeave | NVIDIA HGX B200 (8x) 8× B200 · 128 vCPU · 2048 GB RAM | $8.60/GPU-hr | $4.26/GPU-hr | — | ||
| 7 | Oracle Cloud Infrastructure | B200 x1 1× B200 | $14.00/GPU-hr | — | — | ||
| 8 | Vultr | B200 x8 8× B200 | Contact for pricing | — | $1.99/GPU-hr Committed (4-yr) |
What is the B200 used for?
The B200 is a data-centre GPU positioned by its maker for large-scale model training, built on the Blackwell architecture and released in 2024. Multi-GPU servers link cards over NVLink 5th gen (1.8 TB/s per GPU), which matters when a job is split across several GPUs.
How large a model fits on one B200?
With 180 GB of memory, one card holds the weights of a model up to about 90 billion parameters at 16-bit precision, or about 180 billion at 8-bit. That is weights only: inference also needs room for the KV cache, and training needs several times more for gradients and optimiser state, so larger jobs are split across multiple GPUs.
| Spec (as published) | B200 |
|---|---|
| Architecture | Blackwell |
| Form factor | SXM (HGX/DGX B200) |
| Memory | 180 GB HBM3e |
| Interconnect | NVLink 5th gen (1.8 TB/s per GPU) |
| Precisions listed | FP4, FP8 |
| Released | 2024 |
| Source | maker spec sheet ↗ |
Need B200 capacity for a longer run?
Committed and cluster pricing is often negotiated. Tell us the GPU count, duration and region.