AWS EC2 Spot Interruption Rates Hit Historic Low — Under 5% for Most Instance Types

AWS has published updated Spot Instance interruption frequency data showing interruption rates below 5% for the majority of instance types across most regions. The improvement is attributed to AWS capacity expansion and improved Spot capacity management algorithms. The Spot Advisor (aws.amazon.com/ec2/spot/instance-advisor) shows real-time interruption frequency per instance type and region.

Best-performing families in us-east-1 (under 2% interruption): m5.large, m5.xlarge, m5.2xlarge, c5.xlarge, c5.2xlarge, r5.large. In ap-southeast-1 (Singapore), m6i and c6i families show under 3% interruption rates. Previously problematic types like g4dn.xlarge have improved to under 5% in most regions. With under 5% interruption rates, a Spot fleet using 6-10 instance types across 3 AZs experiences near-zero actual interruption events.

Savings remain compelling: 60-70% vs On-Demand. A 50-instance m7i.2xlarge fleet at On-Demand costs $177,000/year. Same fleet on Spot: approximately $62,000/year — a $115,000/year saving. With modern interruption handling (2-minute notice, graceful drain), Spot is now viable for production web tiers with proper load balancing and autoscaling groups.

TCOIQ Action: Implement Spot Instances in stages. Start with CI/CD build agents (zero-cost to implement, 70% savings, interruptions handled automatically). Then ML training jobs (checkpoint every 10 minutes). Then batch processing. Finally, production web tier with On-Demand fallback (80% Spot, 20% On-Demand baseline). Use the aws-node-termination-handler Kubernetes controller for automatic graceful drain.