ToolsSeptember 15, 2026via AWS Machine Learning Blog
Announcing instance preference lists for Amazon SageMaker AI training jobs
Why it matters
A practical productivity feature for ML practitioners running training jobs at scale. Eliminates manual capacity-watching and retry loops — a small but real operational win for teams juggling GPU scarcity and cost.
Key signals
- Instance preference lists: up to five instance types per training/processing job
- Automatic failover to first available capacity — no manual retries required
- Eliminates need for custom capacity-watching scripts
- SageMaker AI product feature, AWS announcement
- Amazon SageMaker AI adds instance preference lists for training and processing jobs
- Users can specify ordered list of up to 5 instance types
- SageMaker automatically launches on first available type
- Eliminates manual retry loops and capacity-watching scripts
- Feature announced September 15, 2026
The hook
SageMaker cuts GPU-hunting friction: specify five instance types, let AWS pick the first one with capacity.
Amazon SageMaker AI now offers instance preference lists for training and processing jobs. Specify an ordered list of up to five instance types, and SageMaker AI automatically launches on the first type with available capacity, eliminating manual retry loops and capacity-watching scripts.