Deploy a GPU VPS Instance
Infinodes supports GPU-backed VPS deployments for compute-intensive workloads such as AI inference, model training, zk proof generation, or high-throughput indexing. These servers are powered by dedicated NVIDIA GPUs hosted on our own infrastructure.
⚙️ Available GPU Plans
| Plan | GPU | vCPUs | RAM | Storage | Use Cases |
|---|---|---|---|---|---|
| GPU-Lite | NVIDIA T4 | 4 | 16GB | 100GB SSD | LLM inference, model hosting, basic AI |
| GPU-Standard | NVIDIA A4000 | 8 | 32GB | 250GB NVMe | Fine-tuning, training, stable diffusion |
| GPU-Pro | NVIDIA A6000 | 16 | 64GB | 500GB NVMe | Large model training, zkSNARK proving |
GPU availability may vary. Contact support if a plan is unavailable.
🧠 Use Cases
- LLM Inference (LLama.cpp, Ollama, etc.)
- AI bots, chat UIs, voice synthesis
- Stable Diffusion / image generation
- ZK circuit compilation (SnarkJS, Circom)
- ML model fine-tuning
🛠️ Deployment Steps
1. Login to Dashboard
Visit infinodes.com/dashboard and log in using your preferred auth method.
2. Go to VPS → Deploy GPU Instance
Select the GPU VPS tab from the dashboard and click Deploy.
3. Configure Instance
- GPU Plan: Choose from T4, A4000, or A6000
- Region: Pick preferred data center
- OS Image: Ubuntu 24.04 (default, CUDA preinstalled)
- SSH Key: Paste your public key for access
- Auto-shutdown: Enable to save costs (optional)
4. Launch & Monitor
Click Launch VPS. You’ll receive:
- Public IP and SSH port
- GPU driver status (NVIDIA SMI output)
- GPU usage metrics in the dashboard
🔐 Security Considerations
- SSH-only login with keys (password disabled)
- UFW + fail2ban enabled by default
- GPU isolation is enforced per container
- Optional VPN-access or private tunnel (on request)
📁 GPU Environment Details
- CUDA: Installed and verified (version varies by plan)
- NVIDIA Drivers: Preinstalled and updated
- Common packages: Python, pip, virtualenv, PyTorch, TensorFlow (optional install)
💬 Need Custom Setup?
We support:
- Custom AMIs / base images
- Persistent volumes
- Docker preloads
- Jupyter Lab setup with password protection
Open a support ticket to request custom configurations.