AWS re:Invent 2025 biggest announcements

Dec 2

Key Takeaways and Biggest Announcements from re:Invent

The core of the keynote was the launch of new services designed to empower developers to build, deploy, and manage AI agents.

AWS Frontier Agents: AWS unveiled a new class of autonomous, scalable, and long-running AI agents designed to work as an extension of a software development team, specifically for building, securing, and operating software.
Amazon Nova Forge: This pioneering service allows customers to build their own custom frontier models by mixing their private data with Amazon-curated datasets, all deployable via Amazon Bedrock.
Amazon Nova 2 Models: An upgraded suite of Amazon’s in-house foundation models, featuring an all-new version available in three tiers for various workloads. A new capability, Nova Sonic, was introduced to power natural, human-like, speech-to-speech conversations within services like Amazon Connect.
Agentic Capabilities in AWS Transform: New AI-powered capabilities for rapid modernization of any code or application. AWS Transform now learns an organization's patterns and automates transformations across legacy tech debt, including full-stack Windows modernization.

To support the massive compute requirements of Agentic AI, AWS announced significant updates to its custom silicon and deployment options.

Trainium 3 UltraServers: The latest generation of AWS’s custom AI chip is now available, delivering 4.4x the compute power of the previous generation. These 3nm AI chips enable customers to train and deploy their most ambitious AI models faster and at a lower cost.
AWS AI Factories: A new offering that transforms a customer’s existing data center into a high-performance AI environment. This essentially deploys dedicated, customer-specific AWS AI infrastructure (a private cloud region) directly into the customer's facility, enabling rapid AI application development at scale.
Project Rainier Activation: The activation of one of the world's largest AI compute clusters, delivering nearly half a million Trainium 2 chips to support massive-scale AI development.

The keynote also introduced new features for fundamental AWS services to improve performance, cost, and developer experience.

AWS Lambda Managed Instances: A new feature that offers the serverless simplicity of AWS Lambda while running on Amazon EC2 compute, providing access to specialized hardware and cost optimizations through EC2 pricing models.
Database Savings Plans: Introduced to help customers optimize costs and manage database workloads more efficiently across services like Amazon RDS.
Amazon S3 Vectors: Now generally available, this capability significantly expands the scale of vector storage and querying, supporting up to 1 billion vectors per index for large-scale generative AI applications.

Connect on LinkedIn

https://www.linkedin.com/in/amycolyer/