GPU Cloud Server
NVIDIA GPUs, from one card to eight
GPU cloud servers come with NVIDIA GeForce RTX 4090 and RTX 5090, or data-centre L40S, A100 and H100 cards — up to eight per server — in our Texas, Munich and Ningxia GPU regions. From a quick experiment to round-the-clock inference, there is a fit.
Starting at
from$10.28/day- GPU models
- NVIDIA RTX 4090 · RTX 5090 · L40S · A100 · H100
- GPU memory per card
- 24–80GB
- Cards per server
- 8
- GPU regions
- Texas / Munich / Ningxia
- Bandwidth
- Overseas 1–20 Gbps shared · dedicated add-on
- Billing
- Daily / monthly
Features
The NVIDIA line-up
RTX 4090 and RTX 5090 suit inference and image generation; L40S, A100 and H100 take on large-model training.
Daily or monthly
RTX 4090, RTX 5090 and L40S can be rented by the day and released when the experiment ends; monthly is better value for sustained inference.
Frameworks ready at boot
Images with CUDA, PyTorch, TensorFlow, vLLM, ComfyUI and LLaMA-Factory preinstalled, so you skip the setup.
Use cases
- 01
Training and inference
Run deep-learning training and online inference workloads.
- 02
Graphics and rendering
Process image, video and rendering tasks.
Configure and price
Pick a family, region and size — the price updates as you go.