Latest Content

How to Provision Vultr Cloud GPU Instances

Updated on 23 February, 2026

A guide explaining how to set up and deploy GPU-powered virtual machines on Vultrs cloud platform.

Vultr Cloud GPU instances are virtual machines that run with dedicated NVIDIA GPUs designed for AI applications, machine learning, HPC, visual computing, and VDI. Cloud GPU instances include a dedicated GPU device capable of running multiple applications depending on your resource requirements.

Follow this guide to provision Vultr Cloud GPU instances using the Vultr Customer Portal, API, CLI, or Terraform.

  • Vultr Customer Portal
  • Vultr API
  • Vultr CLI
  • Terraform

  1. Navigate to Products and select Compute on the list of product options.

  2. Click Deploy to access the Deploy New Instance page.

  3. Select Cloud GPU.

  4. Choose your desired Vultr location to deploy the instance.

  5. Select a GPU type from the sidebar based on your workload requirements.

  6. Select a plan from the available options based on your GPU, vCPU, memory, storage, and bandwidth requirements.

  7. Click Configure Software.

  8. Select a cloud image to install on your instance based on the following options:

    • Marketplace Apps: Install a prebuilt software stack or application and the recommended operating system image on the instance.
    • Operating System: Install a fresh operating system image on the instance. GPU enabled versions include pre-installed drivers.
    • Upload ISO: Install a specific ISO available in your Vultr account or upload a new ISO image from your workstation to install on the instance.
    • ISO Library: Install a specific ISO image from the Vultr ISOs library.
    • Backup: Recover a specific backup available in your Vultr account to the instance.
    • Snapshot: Install a specific snapshot available in your Vultr account to the instance.

  9. Select optional Server Settings to apply on the instance.

    • SSH Keys: Installs a specific SSH key on the instance.
    • Startup Script: Enables a startup script to execute at deployment or a PXE script to automate the operating system installation.
    • Firewall Group: Activates a Vultr Firewall group to filter incoming network traffic on the instance.

  10. Enter a new hostname in the Server Hostname field and a descriptive label in the Server Label field to identify the instance.

  11. Configure Additional Features for the instance.

    • Instance Connectivity: Select how the instance connects to the internet.
      • Instance(s) with Public IP: Assigns public IP addresses directly to the instance. Under Instance Address, Public IPv4 is enabled by default. Select Public IPv6 to enable IPv6 addressing. After selecting IPv6, you can optionally deselect Public IPv4 to create an IPv6-only instance.
      • Private Instance(s) behind NAT Gateway: Routes internet traffic through a NAT Gateway in a Virtual Private Cloud (VPC) Network. Select an existing VPC Network with a NAT Gateway or click Add VPC Network to create a new one.
    • VPC Network: Connects the instance to a VPC Network in the current location.
    • Automatic Backups: Automatically creates backups for data recovery in case of instance failures.
    • DDoS Protection: Prevents potential Distributed Denial of Service (DDoS) attacks on the instance.
    • Limited User Login: Creates a linuxuser non-root user with sudo privileges as the default user account instead of root.
    • Cloud-Init User Data: Enables Cloud-Init user data to initialize and customize the instance at boot.

  12. Click Deploy to provision the instance.

  1. Send a POST request to the Create Instance endpoint to create a new Vultr Cloud GPU instance. Replace VULTR_LOCATION, INSTANCE_PLAN, OS_ID, INSTANCE_LABEL, and HOSTNAME with your target values.

    console 
    $ curl "https://api.vultr.com/v2/instances" \
  -X POST \
  -H "Authorization: Bearer ${VULTR_API_KEY}" \
  -H "Content-Type: application/json" \
  --data '{
    "region" : "VULTR_LOCATION",
    "plan" : "INSTANCE_PLAN",
    "os_id" : OS_ID,
    "label" : "INSTANCE_LABEL",
    "hostname": "HOSTNAME"
  }'

    Visit the Create Instance API page to view additional attributes you can apply on the Vultr Cloud GPU instance.

  1. Create a new Vultr Cloud GPU instance. Replace VULTR_LOCATION, INSTANCE_PLAN, OS_ID, INSTANCE_LABEL, and HOSTNAME with your target values.

    console 
    $ vultr-cli instance create --region VULTR_LOCATION --plan INSTANCE_PLAN --os OS_ID --label INSTANCE_LABEL --host HOSTNAME

    Run vultr-cli instance create --help to view additional options you can apply on the Vultr Cloud GPU instance.

  1. Ensure the Vultr Terraform provider is configured in your Terraform project.

  2. Add the Cloud GPU instance resource to your Terraform configuration.

    terraform 
    terraform {
    required_providers {
        vultr = {
            source  = "vultr/vultr"
            version = "~> 2.26"
        }
    }
}

provider "vultr" {}

resource "vultr_instance" "gpu" {
    region      = "del"              # Target deployment region
    plan        = "vcg-a100-12c-120g-80vram" # Cloud GPU plan ID (A100/A40/A16)
    os_id       = 2284               # Ubuntu 24.04 LTS x64
    label       = "gpu-instance-1"
    hostname    = "gpu-instance-1"
    enable_ipv6 = true
}

output "public_ip" {
    value = vultr_instance.gpu.main_ip
}

  3. Apply the configuration and observe the following output:

    Apply complete! Resources: 1 added, 0 changed, 0 destroyed.

Comments