GPU Cloud Price Index
The cheapest tracked on-demand rows across cloud GPU providers, grouped into practical floors for general compute, inference, training, and premium-memory workloads.
Cheapest cloud GPU floors
Updated Jun 21, 2026. These cards are computed from current on-demand medians.
RTX 4090 on Vast.ai
The cheapest tracked on-demand row is $0.40/hr ($293/mo) for RTX 4090 on Vast.ai.
RTX 4090 on Vast.ai
For 24GB+ inference-friendly GPUs, the current floor is $0.40/hr on Vast.ai.
A100 PCIE on Vast.ai
For 80GB+ training-friendly GPUs, the current floor is $1.00/hr on Vast.ai.
Current on-demand leaders
Each row links to the provider/GPU landing page behind the current cheapest tracked price.
| GPU | Provider | VRAM | Median | Monthly | Offers |
|---|---|---|---|---|---|
| H100 SXM | Vast.ai | 80GB HBM3 | $2.40/hr | $1,752/mo | 15 |
| H100 NVL | Vast.ai | 94GB HBM3 | $2.27/hr | $1,656/mo | 9 |
| H100 PCIE | Vast.ai | 80GB HBM3 | $1.64/hr | $1,199/mo | 1 |
| H200 | Lambda | 141GB HBM3e | $2.29/hr | $1,672/mo | 1 |
| H200 NVL | Vast.ai | 141GB HBM3e | $3.46/hr | $2,529/mo | 3 |
| A100 SXM4 | Vast.ai | 80GB HBM2e | $1.14/hr | $834/mo | 6 |
| A100 PCIE | Vast.ai | 80GB HBM2e | $1.00/hr | $731/mo | 2 |
| B200 | Vast.ai | 192GB HBM3e | $4.38/hr | $3,194/mo | 29 |
| L40 | Vast.ai | 48GB GDDR6 | $0.45/hr | $331/mo | 1 |
| RTX 6000Ada | Vast.ai | 48GB GDDR6 | $0.60/hr | $439/mo | 8 |
| RTX 4090 | Vast.ai | 24GB GDDR6X | $0.40/hr | $293/mo | 69 |
| RTX 5090 | Vast.ai | 32GB GDDR7 | $0.53/hr | $390/mo | 129 |
| A10G | Oracle | 24GB GDDR6 | $2.00/hr | $1,460/mo | 1 |
Who owns the cheapest rows?
A provider leader count means that provider currently has the cheapest tracked on-demand row for a GPU.