VMI containers and cloud deployment

Virtual Machine Images (VMIs) on NGC are complete VM images for GPU workloads on cloud platforms — pre-configured with OS, NVIDIA drivers, CUDA toolkit, and AI frameworks.

Supported cloud platforms:

• AWS — EC2 p3, p4, p5 GPU instances
• Azure — NC/ND GPU VMs
• Google Cloud — A100/H100 GPU instances
• Oracle Cloud — GPU shapes

Deploying a VMI:

1. Go to NGC: ngc.nvidia.com → Catalog → VM Images
2. Select your cloud provider
3. Follow NGC's launch instructions for your account
4. The VMI is pre-configured — CUDA, cuDNN, NCCL, Docker, Container Toolkit all included

VMIs are certified by NVIDIA, ensuring driver/software compatibility and optimized performance for the target cloud hardware.