Loading...
The cloud's spare capacity at up to 90 percent off, with the catch that it can vanish.
Quick Definition
Spot instances are spare cloud capacity offered at steep discounts, often up to 90% off on-demand, but can be reclaimed with little notice. They are ideal for fault-tolerant, flexible, or batch workloads and are a powerful lever for reducing compute cost.
Spot instances are spare cloud capacity sold at steep discounts, commonly 60 to 90 percent below on-demand prices, with the condition that the provider can reclaim them at short notice, typically two minutes' warning. Same hardware, fraction of the price, no guarantee of tenure.
Spot is a fit question, not a bravery question. Workloads that tolerate interruption, batch jobs that checkpoint and resume, stateless web tiers behind load balancers with mixed capacity, CI runners, big data and training jobs, capture the discount safely. Workloads that cannot lose a node mid-task should pay for on-demand or committed capacity instead.
Example. A team moves its CI fleet and nightly data pipelines to spot with automatic retry. Interruptions occur a few times weekly and cost minutes of rerun time; the compute bill for those workloads drops 72 percent.
On Kubernetes, a dedicated spot node pool plus Karpenter makes adoption nearly operational-free. The Kubernetes Cost Management Guide covers the patterns, and the AWS Auto Scaling primer covers mixed-capacity groups.
It varies by instance type, region, and demand. Diversifying across types and zones sharply reduces simultaneous interruptions.
Single-instance databases, stateful services without replication, and anything where a two-minute eviction causes data loss or customer impact.
Commonly 60 to 90 percent off on-demand. Across a suitable workload portfolio, spot is often the single largest discount available.