Auto Scaling helps to adjust the capacity of the infrastructure to maintain steady, predictable performance and cost in the face of variable workload demands. It automates the process by scaling compute resources up or down based on criteria defined by the user, such as CPU utilization or network I/O. This contrasts with just using a standard compute instance service, which does not offer scaling without manual intervention or additional configuration. Serverless services like Lambda are designed for short, stateless computations and are typically managed without the need for scaling settings. Load balancing is a technique to spread workloads across multiple compute resources but does not inherently adjust the resource count.