Spot Instances

Spot instances are spare cloud compute sold at a steep discount — often 60-90% below on-demand price — in exchange for one catch: the provider can reclaim the machine with little warning (typically a couple of minutes) when it needs the capacity back. AWS calls them Spot, Google calls them Spot VMs, Azure calls them Spot Virtual Machines. Why it matters: for interruptible, stateless, or batch work — CI runs, data processing, background jobs, GPU inference at scale — spot capacity can slash your compute bill dramatically. Practical note: never put stateful or latency-critical workloads (your primary database, checkout flow) on pure spot. Design for interruption: checkpoint long jobs, drain connections on the termination signal, and mix spot with on-demand or reserved capacity so a mass reclaim doesn't take you fully offline. Many managed platforms and Kubernetes autoscalers can blend spot and on-demand nodes automatically, giving you most of the savings with a safety floor.