Skip to main content

Create your first GPU VM

Overview​

Following this guide, you can create and connect to a GPU VM end-to-end in under five minutes.

Prerequisites

Step 1: Create the VM​

  1. From the left-hand menu, choose Compute > Virtual Machines.

  2. Click the Create VM button in the top right.

    ECI portal home

  3. On the Basic Info step, fill in the following:

    FieldDescription
    NameA recognizable name (e.g. my-gpu-vm)
    ZoneThe zone to deploy in (choices vary by organization)
    Instance typePick a GPU type (e.g. G-NHHS-80 = H100 × 1)
    Pricing typeOn-demand is recommended for your first try
    ImageFor GPU workloads, pick an image whose name contains (AI/GPU) (e.g. Ubuntu 24.04 LTS (AI/GPU) (50 GiB) — CUDA and drivers preinstalled)
    Username1–32 characters; lowercase letters, digits, _, and - only; must start with a lowercase letter or _ (e.g. elice)
    Password10–256 characters; must include 3 or more of uppercase, lowercase, digits, special characters. No sequential or repeating characters, and the email local part is not allowed

  4. On the Network step, choose the virtual network and subnet. If you need external access, either create a new public IP or attach an existing one. The network interface is created at this step, and the public IP attaches to that interface.

  5. On the Storage step, confirm the OS disk size and attach additional block storage if needed.

  6. On the Review + Create step, confirm the settings and estimated cost, then click Create.

    Review + Create screen

Without a public IP, external SSH won't work

To connect from outside, you must allocate a public IP. If you only plan to use the in-portal web console, you don't need one.

Set firewall rules at the same time

The ECI firewall allows all traffic by default. A VM with a public IP attached will, with no further changes, expose its SSH port to the entire internet. Add a rule at the virtual-network level that narrows inbound port 22 to your own IP and blocks everything else. See Firewall rules for details.

Step 2: Start the VM​

  1. After creation completes, you're taken to the VM detail page automatically (status: Ready).
  2. Click the Start button in the top right.
  3. Wait until the status changes to Running (usually 1–2 minutes).

Starting the VM

Step 3: Connect from the web console​

You can open a terminal directly from the browser, with no public IP or SSH setup required.

  1. On the VM detail page, click the Web Console tab.
  2. Sign in with your username and password.

Web console

To connect via SSH

With a public IP assigned, run ssh <username>@<public-ip> from a terminal. For SSH key authentication and more, see VM access methods.

Step 4: Verify the GPU​

nvidia-smi

If you see the GPU model name and driver version, you're good to go.

Next steps​