AWS instances for Aerospike clusters are selected based on a combination of factors including storage, memory, network, CPU, and cost considerations. The optimal instance type depends on the specific use case, performance requirements, and budget constraints. Instance families such as i3en, i3, r5, r5d, c5d, and m5 are considered for their unique characteristics, including storage capacity, throughput, memory-to-disk ratio, network performance, and CPU power. The instance type selection process involves evaluating the cost per GB of RAM, SSD storage, and vCPU, as well as considering the impact of clustering and scaling on performance and cost. A conservative approach can be taken initially, with room for adjustment based on real-world production metrics, allowing for rolling updates and zero downtime to optimize performance and cost.