AWS Introduces Cutting-Edge AI and Cloud Innovations at re:Invent 2025

At re:Invent 2025, Amazon Web Services (AWS) unveiled a range of next-generation technologies aimed at transforming enterprise cloud and AI capabilities. The announcements included the Graviton5 CPU, Trainium3 UltraServers, Amazon Bedrock AgentCore, AI Factories, and an expanded Amazon Nova model lineup.

The Graviton5 processor delivers up to 25% higher performance with 192 cores per chip and a cache five times larger than its predecessor, powering more than half of new AWS CPU capacity. Leading companies such as Adobe, Airbnb, Epic Games, Formula 1, Pinterest, SAP, and Siemens are already benefiting from its cost-efficiency and performance advantages. The new Graviton5-based Amazon EC2 M9g instances are designed to handle complex workloads faster and more sustainably, helping organizations reduce costs without compromising performance.

AWS is also expanding the Nova model family with four new models optimized for reasoning, multimodal processing, conversational AI, code generation, and agentic tasks. Nova Forge introduces “open training,” enabling organizations to combine pre-trained model checkpoints with proprietary datasets. Early users report significant improvements, including up to 90% reliability for browser-based UI automation and faster development cycles, with companies like Reddit and Hertz seeing substantial productivity gains.

The company also launched three frontier agents—Kiro autonomous agent, AWS Security Agent, and AWS DevOps Agent—designed to operate autonomously for extended periods, effectively acting as AI team members. Organizations such as Commonwealth Bank of Australia, SmugMug, and Western Governors University are leveraging these agents to enhance software development, security, and operational workflows.

To support large-scale AI workloads, AWS introduced Trainium3 UltraServers, powered by its first 3nm AI chip and capable of hosting up to 144 chips per system. These servers offer 4.4 times more compute performance and four times greater energy efficiency than their predecessors, reducing AI training from months to weeks. Companies including Anthropic, Karakuri, Metagenomi, NetoAI, Ricoh, and Splash Music have reported up to 50% savings in training and inference costs, while Decart achieves four times faster generative video processing at half the GPU cost.

AWS AI Factories enable enterprises and government organizations to deploy dedicated AI infrastructure in their own data centers, integrating NVIDIA GPUs, Trainium chips, and AWS AI services. This approach allows organizations to leverage existing data center resources while maintaining regulatory compliance and data sovereignty.

Finally, Amazon Bedrock AgentCore introduced new features to improve AI governance and performance. Policy in preview allows teams to define boundaries for agent actions using natural language, while AgentCore Evaluations and Memory enhance monitoring and learning capabilities.

These announcements demonstrate AWS’s continued focus on providing powerful, efficient, and scalable solutions that empower organizations to accelerate innovation, optimize operations, and unlock new business possibilities.