NVIDIA H200 vs NVIDIA B200
What each costs to rent by the hour across the providers we track — NVIDIA H200 on 53 providers, NVIDIA B200 on 39.
As of , the NVIDIA B200 has a median rate card of $7.50 per GPU-hour versus $4.54 for the NVIDIA H200 — $2.96 (39%) more for 36% more memory.
Side by side
| NVIDIA H200 | NVIDIA B200 | |
|---|---|---|
| Architecture | Hopper | Blackwell |
| VRAM | 141 GB | 192 GB |
| Form factor | SXM | SXM |
| Median rate card, per GPU-hour | $4.54 | $7.50 |
| Cheapest listed rate | $2.09 Beam, on-demand, from captured | $1.45 Google Cloud, spot, GPU only captured |
| Median per GB of VRAM, per hour | $0.032 | $0.039 |
| Providers listing it | 53 | 39 |
Providers that list both
| Provider | NVIDIA H200 | NVIDIA B200 | B200 vs H200 |
|---|---|---|---|
 Beam | $2.09 from captured | $4.11 from captured | +$2.02 (+97%) |
 Dataoorts | $2.29 Orion Series captured | $5.10 Atlas Series captured | +$2.81 (+122%) |
 HPC-AI.COM | $2.68 from captured | $7.50 from captured | +$4.82 (+180%) |
 Bitdeer AI | $2.76 from captured | $4.69 from captured | +$1.93 (+70%) |
 Koyeb | $3.00 captured | $5.50 captured | +$2.50 (+83%) |
 Northflank | $3.14 GPU only captured | $5.87 GPU only captured | +$2.73 (+87%) |
 GMI Cloud | $3.20 from captured | $5.00 from captured | +$1.80 (+56%) |
 Deep Infra | $3.49 captured | $4.89 captured | +$1.40 (+40%) |
 Hyperstack | $3.99 captured | $6.00 captured | +$2.01 (+50%) |
 Hyperbolic | $3.99 from captured | $5.99 from captured | +$2.00 (+50%) |
 Cerebrium | $4.20 GPU only captured | $6.01 GPU only captured | +$1.81 (+43%) |
 HexGrid | $4.20 Secure Cloud captured | $7.50 Secure Cloud captured | +$3.30 (+79%) |
 Barrack AI | $4.45 captured | $6.89 captured | +$2.44 (+55%) |
 Modal | $4.54 GPU only captured | $6.25 GPU only captured | +$1.71 (+38%) |
 Daytona | $4.54 GPU only captured | $6.25 GPU only captured | +$1.71 (+38%) |
 Hugging Face | $5.00 aws captured | $9.25 aws captured | +$4.25 (+85%) |
 RunPod | $5.29 Secure Cloud captured | $7.99 Secure Cloud captured | +$2.70 (+51%) |
 E2E Networks | $5.29 captured | $7.99 captured | +$2.70 (+51%) |
 GPU.ai | $5.29 from captured | $7.99 from captured | +$2.70 (+51%) |
 Nebius | $5.40 captured | $8.50 captured | +$3.10 (+57%) |
 Verda | $5.89 captured | $8.36 captured | +$2.47 (+42%) |
 Together AI | $5.99 captured | $8.19 captured | +$2.20 (+37%) |
 fal | $6.00 captured | $7.99 captured | +$1.99 (+33%) |
 Lyceum | $6.19 captured | $8.79 captured | +$2.60 (+42%) |
 CoreWeave | $6.30 captured | $8.60 captured | +$2.29 (+36%) |
 Amazon Web Services | $7.91 us-east-1 captured | $14.24 us-east-1 captured | +$6.33 (+80%) |
 Fireworks AI | $8.00 GPU only captured | $13.00 GPU only captured | +$5.00 (+63%) |
 Google Cloud | $9.31 us-west1, GPU only captured | $16.11 us-west1, GPU only captured | +$6.80 (+73%) |
 Oracle Cloud (OCI) | $10.00 captured | $14.00 captured | +$4.00 (+40%) |
 EmpirioLabs | $10.70 captured | $12.20 captured | +$1.50 (+14%) |
Medians are median rate cards — each provider once, at its cheapest on-demand rate, community capacity and marketplace floors out — the same figure as on each card's page; the cheapest listed rate may be spot or a community tier and is labelled as such. Compare providers directly on the overview, or see each GPU's full table: NVIDIA H200, NVIDIA B200. On your own page: embed this comparison (/embed/gpus/h200-vs-b200, an iframe, no script).
Data updated: