Engineers love to say "global" when they mean "two regions in the same cloud provider." This talk skips the continent-counting contest and focuses on what actually breaks: cross-border latency spikes, data sovereignty laws that contradict each other, and the uncomfortable truth that your "eventually consistent" database becomes "eventually never" when a cable gets dredged up. We'll walk through real production failures from systems that genuinely span disparate locations, no spinning globes, no "just use Kafka." You'll learn: - Why asymmetric latency (not distance) is your real enemy, and how to measure it before it measures you - A decision framework for active-active vs. active-passive that doesn't rely on cloud vendor marketing - One working pattern for data placement when you don't control where your users are