A globally distributed GPU cloud that offers agent-ready pods for deploying local models like Llama 3 or DeepSeek. On-demand and spot GPU instances for ML training and inference workloads.