A100: price per hour on every cloud
An hour of one A100 (80 GB) costs from $1.49 on DeepInfra. 8 clouds rent it on demand.
| Cloud | Per GPU-hour | Cheapest instance | GPUs in it | Instance, an hour | Region | Instances with it |
|---|---|---|---|---|---|---|
| DeepInfra | $1.49 | a100.1x | 1 × 80 GB | $1.49 | global | 1 |
| Crusoe | $2.00 | a100-pcie.1x | 1 × 80 GB | $2.00 | global | 2 |
| HF Endpoints | $2.50 | aws.nvidia-a100.x1 | 1 × 80 GB | $2.50 | global | 8 |
| AWS | $2.74 | p4d.24xlarge | 8 × 40 GB | $21.96 | us-east-1 | 2 |
| Azure | $3.40 | Standard_ND96asr_v4 | 8 × 40 GB | $27.20 | eastus | 5 |
| Google Cloud | $3.48 | a2-megagpu-16g | 16 × 40 GB | $55.74 | us-central1 | 9 |
| OCI | $4.00 | BM.GPU.A100-v2.8 | 8 × 80 GB | $32.00 | global | 1 |
| IBM watsonx | $5.80 | deploy-on-demand.1xA100 | 1 × 80 GB | $5.80 | global | 4 |
What one of them can run
- Qwen3 30B A3B Instruct 2507, which needs 76.3 GB
- OpenAI gpt-oss 120B, which needs 73 GB
- Qwen3.8 27B, which needs 69.5 GB
- Gemma 3 27B, which needs 68.6 GB
- Gemma 4 26B A4B, which needs 64.5 GB
- Mistral Small, which needs 58.9 GB
- Phi-4, which needs 36.6 GB
- Llama 3.1 8B, which needs 20.1 GB
List prices collected 22 November 2026, refreshed every three hours. No negotiated rates or taxes.