AWS recommends running highly available EKS clusters with worker nodes in node pools that are spread across multiple Availability Zones or Regions for reliability, but this approach can lead to sub-optimal utilization and significant cloud waste and high costs. Using a single node pool with maximum utilization is a more cost-efficient alternative, as it allows teams to avoid the mounting costs of cloud resources by keeping all nodes completely full, eliminating any cloud waste. This approach requires constant monitoring to ensure utilized and requested capacity match, making automation a key factor in its success.