“Corey Sanders suggests that because network latency is a smaller component of total inference time for AI workloads, there is greater flexibility for deploying workloads across different geographic regions to improve availability and manage capacity bursts.”