You have to think about its usage patterns, how to will be accessible to other systems, what applications it will host, how it will be named, where it will be placed, how many resources to allocate them and what type, what subnet to be part of, what network security groups to create, how to do operating system updates, what special licenses will be required, how to accommodate for start up times, how to manage base images with pre-installed software, how to create automated process to create these base images, how to ensure compliance and security, how deployments will be performed, how to do zero downtime deployments, etc.
I could go and on. In short, you have to give them names, feed them, pet them, and take care of them - treat them like pets. If they get sick, you nurse them back to health.
I like the promise of Platform as a Service, Serverless, or Kubernetes in help us move away from this type of mentality. It creates options to reduce the operational workload and buy bandwidth to innovate more. Allows us to create more business value to our end users.