What is Spot Instance?
Also known as: Preemptible VM
A Spot Instance is a discounted virtual machine that a cloud provider can reclaim with little or no notice when it needs the capacity back for on-demand customers.
A Spot Instance (also called a Preemptible VM in Google Cloud or a Preemptible Instance in some contexts) is a compute resource sold at a deep discount because the provider reserves the right to interrupt or terminate it with short notice, typically 30 seconds to 2 minutes. AWS introduced Spot Instances in 2009, and similar models now exist in Google Cloud, Azure, and other platforms. The price is set by supply and demand or is a fixed discount off the on-demand rate, fluctuating dynamically depending on available excess capacity.
The critical difference from an on-demand instance is the interruption risk. The provider signals a reclaim by sending a notice, for example, a two-minute AWS Spot Instance interruption notice via the instance metadata service or a CloudWatch event. The instance can be stopped, hibernated, or terminated depending on the type and configuration. Spot Instances are suitable for stateless, fault-tolerant, or checkpointable workloads: batch processing, data analytics, rendering, CI/CD build agents, and containerized microservices that can tolerate eviction. They are not appropriate for stateful databases, latency-sensitive web servers, or long-running single-node applications without a failover mechanism.
In the wider cloud architecture, Spot Instances form a critical part of cost optimization strategies. Combined with auto-scaling groups, Spot Fleet (AWS), or node pools (GKE), they allow users to provision large compute clusters at 60-90% lower cost than on-demand, while using on-demand or reserved instances as a base layer for reliability. Major cloud providers now offer features like capacity pools, diversified allocation strategies, and persistent requests to improve availability and manage interruption gracefully.
Key facts
- Typical discount is 60-90% compared to on-demand pricing.
- Provider can reclaim capacity with as little as 30 seconds notice.
- AWS, Google Cloud, and Azure all offer equivalent products (Spot VMs, Preemptible VMs).
- Ideal for batch, container, CI/CD, and stateless microservice workloads.
- Cannot be used for stateful applications without checkpointing or failover.
How it works in practice
Related terms
References
More in Cloud Computing
Availability Zone
An Availability Zone is an isolated, fault-tolerant data centre within a cloud region, with independent power, cooling, and networking to prevent single points of failure.
Cloud Bill Shock
Cloud bill shock is an unexpectedly high invoice caused by unmonitored resource usage, often from data egress fees, autoscaled compute instances, or orphaned storage volumes.
Cloud Cost Optimization
Cloud cost optimization is the continuous practice of measuring, analyzing, and reducing cloud spending through rightsizing instances, purchasing reservations, and eliminating waste.
Cloud Migration
Cloud migration is the process of transferring applications, data, and workloads from on-premises infrastructure or another cloud to a target cloud platform.
Cloud-Native
Cloud-native is an approach to building and running applications that exploit the elastic, distributed, containerized, and immutable nature of modern cloud infrastructure.
Edge Location
An edge location is a data center or point of presence deployed by a cloud provider at the network edge to serve content and run compute workloads closer to end users than a full parent region.
Egress Fees
A per-gigabyte charge imposed by cloud providers on network traffic that leaves their data center, availability zone, or internet boundary.
FaaS
FaaS (Function as a Service) is a cloud computing model where users deploy small, stateless functions that the provider executes on demand, automatically scaling and billing per invocation.
Hybrid Cloud
Hybrid cloud is a computing environment that combines a public cloud with a private or on-premises data center, connected by secure networking to enable workload portability and unified management.
IaaS
Infrastructure as a Service (IaaS) is a cloud computing model that provides on-demand access to virtualized compute, storage, and networking resources, which the customer configures and manages from the operating system upward.