H100 vs B200: Cloud GPU Pricing
By median on-demand price, H100 is about 61% cheaper per GPU-hour today (H100 $3.30 vs B200 $8.50). Price per hour is not price per result: a faster GPU can finish a job in fewer hours, so compare cost per job on your own workload.
H100
NVIDIA Hopper
- Lowest
- $1.66/hr
- Median
- $3.30/hr
- Highest
- $10.00/hr
- Listings
- 54
- Providers
- 18
- Cheapest on
- Latitude
B200
NVIDIA Blackwell
- Lowest
- $3.74/hr
- Median
- $8.50/hr
- Highest
- $14.24/hr
- Listings
- 17
- Providers
- 8
- Cheapest on
- BoostRun
Cheapest H100 listings
| Provider | GPUs | VRAM | vCPU | RAM (GB) | Region | Availability | $/GPU-hr |
|---|---|---|---|---|---|---|---|
| Latitude | 1 | 80 GB | 16 | 128 | ashburn-usa-7 | Out of stock | $1.66 |
| Latitude | 8 | 80 GB | 128 | 1536 | dallas-usa-1 | Out of stock | $1.79 |
| Horizon | 8 | 80 GB | 160 | 1536 | blonduos-iceland-2 | Out of stock | $1.95 |
| Latitude | 1 | 80 GB | 16 | 192 | dallas-usa-1 | Out of stock | $1.99 |
| Voltagepark | 8 | 80 GB | 104 | 1024 | dallas-usa-5 | Out of stock | $1.99 |
| Voltagepark | 4 | 80 GB | 104 | 464 | dallas-usa-5 | Out of stock | $1.99 |
| Voltagepark | 2 | 80 GB | 52 | 232 | dallas-usa-5 | Out of stock | $1.99 |
| Voltagepark | 1 | 80 GB | 26 | 116 | dallas-usa-5 | Out of stock | $1.99 |
Cheapest B200 listings
| Provider | GPUs | VRAM | vCPU | RAM (GB) | Region | Availability | $/GPU-hr |
|---|---|---|---|---|---|---|---|
| BoostRun | 8 | 192 GB | 96 | 2048 | dallas-usa-4 | Out of stock | $3.74 |
| Excess Supply | 8 | 192 GB | 160 | 1792 | kansascity-usa-4 | Out of stock | $4.51 |
| Lambda | 8 | 180 GB | 208 | 3114.6000000000004 | asia-northeast-1 | – | $7.69 |
| Lambda | 8 | 180 GB | 208 | 1933.2 | asia-northeast-1 | – | $7.69 |
| Lambda | 2 | 180 GB | 52 | 773.2800000000001 | asia-northeast-1 | – | $7.89 |
| Lambda | 1 | 180 GB | 26 | 386.64000000000004 | asia-northeast-1 | – | $7.99 |
| Nebius | 1 | 180 GB | 20 | 224 | us-central1 | – | $8.50 |
| Nebius | 1 | 180 GB | 20 | 224 | me-west1 | – | $8.50 |
On-demand USD per GPU-hour. Spot, reserved and fractional-GPU offers are excluded. Data refreshed 2026-11-18. Prices come from provider APIs and public catalogs; confirm on the provider's site before committing spend.