Spot Instance
Spot instances are cheap, reclaimable cloud compute for interruptible workloads, offering major savings when applications tolerate sudden termination.
A spot instance is unused cloud compute capacity sold at a large discount, often up to ninety percent below on-demand price. The catch is that the provider can reclaim the capacity at short notice when it is needed elsewhere. Spot instances are ideal for workloads that tolerate interruption and can resume or restart cleanly.
How It Works
Providers make idle capacity available as spot, with prices that vary by region, instance type, and availability. When demand for that capacity rises, the provider sends a short interruption notice, often a couple of minutes, before reclaiming the instance. Applications handle this by checkpointing progress, draining work, and gracefully shutting down. Common patterns include diversifying across multiple instance types and zones to reduce simultaneous reclamation, and mixing spot with on-demand or reserved capacity so a baseline stays stable. Container orchestrators and batch schedulers integrate spot handling to reschedule interrupted work automatically.
Why It Matters
Spot instances are one of the largest cost levers in cloud computing. Batch processing, data analytics, CI pipelines, rendering, machine-learning training, and stateless web tiers can run far cheaper on spot. The trade-off is operational: workloads must expect interruption and avoid holding critical state in memory. For steady, uninterruptible workloads, reserved instances or savings plans offer predictable discounts instead. A balanced fleet often blends all three pricing models.
Related Terms
Spot instances contrast with reserved instances and pair with autoscaling and elasticity as a core FinOps cost-optimization technique.