# GPU Cloud on CreateOS

Rent GPU instances for AI/ML training, inference, and rendering. Access NVIDIA GPUs on-demand with pay-as-you-go pricing on decentralized infrastructure.

## Available GPUs

### NVIDIA A100

80 GB HBM2e

Large model training

### NVIDIA A10G

24 GB GDDR6

Inference & rendering

### NVIDIA T4

16 GB GDDR6

Cost-effective inference

### NVIDIA L4

24 GB GDDR6

Video & AI workloads

## Use Cases

- Large language model (LLM) training
- Computer vision model training
- Real-time AI inference
- Image and video generation
- Scientific computing
- 3D rendering and simulation

## Features

- On-demand GPU instances
- Pay-per-second billing
- Pre-installed CUDA and cuDNN
- PyTorch and TensorFlow ready
- Jupyter notebook support
- SSH access
- Persistent storage
- Auto-scaling for inference

## Pricing

GPU pricing is usage-based and varies by GPU type. Pay only for what you use with per-second billing. Contact sales for enterprise volume discounts.

## How It Works

1. **Select GPU type**  
   Choose the GPU that matches your workload requirements.

2. **Deploy your code**  
   Push your ML code, Dockerfile, or use a GPU-enabled template.

3. **Run workloads**  
   Train models, run inference, or process data on powerful GPUs.

4. **Scale as needed**  
   Scale up for training, scale down for inference, or run spot instances.