# Create an AI server
Source: https://docs.infraista.com/en/compute/ai-servers/create-an-ai-server



**Before you start:** your account setup must be complete, including a payment method
([Create your account](/en/getting-started/create-your-account)).

1. Go to **Compute › AI Servers** and select **Create AI server** (or **Deploy** on a
   [GPU package](/en/compute/ai-servers/choose-a-gpu-package)).
2. **GPU plan**: pick the GPUs, CPU, RAM and disk. **Show all GPU plans** lists every plan.
3. Choose **Hourly** or **Monthly** billing:
   * **Hourly**: pay for every hour it runs. Stop it any time and compute stops.
   * **Monthly**: a fixed monthly price, charged when the server is ready and renewed each month.
4. **Operating system**: pick a Linux system or Windows. GPU drivers come pre-installed.
5. **Location**: pick the region closest to you or your data.
6. **Access**: an **SSH key** (recommended, Linux) or a **Root password** we generate and email you.
7. **Details**: a **Name** (also the hostname) and, optionally, a **Project**.
8. Check the **Summary** and select **Create AI server**.

<Shot name="compute/ai-servers/create" alt="The Create an AI server page with GPU plans, billing, operating system, location, access and the summary" />

The page shows the deployment's progress. You can leave it: the server keeps building and we email
you when it's up. Billing starts once the server is running. If the deployment fails, you aren't
charged for it.

## Related [#related]

* [Connect to an AI server](/en/compute/ai-servers/connect-to-an-ai-server)
* [Switch between hourly and monthly billing](/en/compute/instances/monthly-billing)