Spot Instances V2

Same infrastructure. Up to 91% off.

Predictable preemptible compute with improved availability and graceful termination. Spot V2 instances are ideal for stateless, fault-tolerant, and batch workloads.

SPOT VM LIFECYCLE & PRICINGOn-DemandSpot Pricing-91%NODE A (STABLE)BATCH_JOB_XYNODE A (PREEMPTED)30S NOTICEEVICTINGBATCH_JOB_XYMIGRATINGNODE B (RESTORED)RE-QUEUEDFault-TolerantAuto-Recovery

Up to 91%

Savings

30 second

Termination notice

Auto-requeue

Restart policy

Improved SLOs

Availability

Smarter preemptible compute.

All the improvements that make spot instances production-ready.

Up to 91% discount

Access the same machine types at a fraction of on-demand pricing. Dynamic pricing based on supply and demand.

30-second termination notice

V2 provides a 30-second warning before preemption via metadata and ACPI signal, giving workloads time to checkpoint.

Auto-requeue

Spot V2 instances automatically requeue terminated workloads. Jobs resume from the last checkpoint when capacity returns.

Flexible capacity pools

Specify multiple machine types and zones. Our scheduler finds the cheapest available option across all your preferences.

Improved availability

V2 spots have better availability guarantees than legacy preemptible VMs. Historical preemption rates available via API.

GPU spot support

Use Spot V2 with GPU instances for ML training at extreme discounts. Automatic checkpointing for NCCL training jobs.

Getting started

Launch your first instance in three steps. CLI, console, or API — your choice.

Terminal
ur compute instances create batch-worker \
  --machine-type=e2-standard-8 \
  --provisioning-model=SPOT \
  --instance-termination-action=DELETE \
  --zone=eu-west1-b

Designed for fault-tolerant workloads.

Batch, CI/CD, ML training — save up to 91% on compute that can handle interruptions.

Large-scale batch computation

Run ETL pipelines, data processing, and MapReduce jobs on hundreds of Spot instances. Automatic requeue handles preemptions transparently.

View tutorial

Suggested configuration

E2-standard-8 · Spot V2 · MIG autoscaler

Estimate your costs

Create detailed configurations to see exactly how much your architecture will cost. Pay for what you use, down to the second.

Configuration 1

Estimated: $220.24/mo

Platform & Architecture

Compute Resources

GB

Storage

GB

Cost Optimization

Hibernation SupportSave instance state on preemption
Config 1 cost$220.24

Cost details

$220.24

Up to 90% savings. 2-minute termination notice.

Configuration 1
$220.24
4 vCPU × 16 GB Compute$210.24
Persistent Storage$10.00

Works seamlessly with

Instance Groups
Kubernetes Engine
Batch API
Cloud Monitoring
Cloud Storage
Cloud Logging
VPC Networks
Cloud Build

Frequently asked questions

Cut your compute bill.

Same infrastructure, up to 91% less. Start using Spot V2 today.