Amazon SageMaker introduces Scale Down to Zero for AI inference to help customers save costs

Amazon SageMaker has introduced a new "Scale Down to Zero" feature for AI inference endpoints, designed to help customers reduce costs for running machine learning models.

• Allows inference endpoints to automatically scale to zero instances during periods of inactivity
• Particularly beneficial for applications with variable traffic like chatbots and content moderation systems
• Endpoints can quickly scale back up when traffic resumes
• Configuration possible through AWS SDK for Python, SageMaker Python SDK, or AWS CLI
• Supports scenarios with predictable or intermittent inference traffic

The feature is now generally available across all AWS regions where SageMaker is supported, offering a cost-effective solution for managing generative AI deployments.