Find GPUsH100 PCIe

Configure your instance

Choose the resources and software for your workload.

Hopper

1 Configuration
2 Your account
3 Deploy

Compute

H100 PCIe · 80 GB HBM2e per GPU

GPUs per node

1 × H100 PCIe26 vCPU · 240 GB RAM2,000 GB local NVMe2 × H100 PCIe52 vCPU · 480 GB RAM4,000 GB local NVMe4 × H100 PCIe104 vCPU · 960 GB RAM8,000 GB local NVMe8 × H100 PCIe208 vCPU · 1920 GB RAM16,000 GB local NVMe

Pricing plan

Spot $0.450/ GPU-hourLower price. A 2-minute notice before interruption.On-demand $0.690/ GPU-hourContinuous capacity for uninterrupted workloads.

Region

Keep your compute close to your data.

US EastVirginia · 100 Gbps40 GPUs availableUS WestOregon · 100 Gbps24 GPUs availableEU CentralFrankfurt · 100 Gbps32 GPUs available

Software template

Choose the tools for your workload.

Environment

Ubuntu 22.04, PyTorch 2.5, CUDA 12.4, cuDNN, JupyterLab, SSH.

Explore all 9 templates

Storage & identity

Local NVMe is included. Add persistent storage for data you want to keep.

Persistent disk (GB)Instance name Optional

Persistent storage: $0.08/GB-month, including while the instance is stopped. Set to 0 to use only the included local disk.

Whole node / hour$0.900