AWS Summit New York 2025: Key AI and Cloud Infrastructure Announcements

On July 16, AWS held its annual New York City summit, focusing on scaling AI capabilities and enhancing cloud infrastructure. Swami Sivasubramanian, AWS VP of Agentic AI, presented the day's keynote and discussed recent updates on delivering production-ready AI agents at scale.

Here’s a breakdown of the most notable updates:

1. Amazon Bedrock AgentCore (Preview)
   AWS introduced AgentCore, a framework aimed at simplifying the deployment and management of AI agents. It supports integration with open-source tools and models and includes features like memory handling and identity controls for enterprise use.

2. Amazon Nova Customization via SageMaker
   Users can now customize Amazon Nova foundation models throughout the training lifecycle using SageMaker. This includes pre-training, fine-tuning, and alignment options, allowing for more tailored AI solutions across industries.

3. AWS AI League
   A new initiative designed to help organizations build AI skills through competitive learning. Participants engage in hands-on activities involving model tuning and prompt engineering using AWS tools.

4. Expanded AWS Free Tier
   New users can receive up to $200 in credits—$100 upon registration and another $100 by engaging with specific AWS services like EC2 and Bedrock.

5. TwelveLabs Video Models on Bedrock
   Video understanding models from TwelveLabs are now accessible via Amazon Bedrock. These models support tasks like video search, scene classification, and summarization.

6. Amazon S3 Metadata Enhancements
   S3 now offers more detailed metadata visibility for all stored objects. This includes live inventory and journal tables that support SQL-based analysis and update automatically.

7. Amazon S3 Vectors (Preview)
   A new storage option designed for vector data, enabling scalable querying and integration with AI tools. AWS claims significant cost savings compared to traditional methods.

8. New SageMaker Features
   Three new capabilities were added:

   • Integration with QuickSight for dashboards and sharing

   • Support for unstructured data from S3

   • Automated data onboarding from Lakehouse environments
     These aim to unify data management and visualization across formats.

9. Amazon EventBridge Logging
   EventBridge now includes detailed logging to help developers monitor and debug event-driven applications more effectively.

10. Blue/Green Deployments in Amazon ECS
    Built-in support for safer container deployments with rollback capabilities, reducing the need for custom tooling.

11. Amazon EKS Scaling
    EKS now supports up to 100,000 nodes per cluster, enabling large-scale AI/ML workloads. This includes compatibility with up to 1.6 million AWS Trainium chips or 800,000 NVIDIA GPUs.

These updates underscore AWS’s ongoing push to make AI more scalable, accessible, and integrated across cloud environments.