All the major highlights from AWS’s flagship tech event re:Invent 2025

AWS re:Invent 2025: Enterprise AI Takes Center Stage

The latest AWS re:Invent conference has concluded, and the overarching theme was clear: empowering businesses with advanced AI solutions. Amidst a flurry of announcements and keynote presentations, Amazon Web Services emphasized how artificial intelligence is reshaping the enterprise landscape.

Empowering Customers with Customizable AI Agents

This year’s event spotlighted enhancements that allow users to tailor AI agents to their specific needs. Notably, AWS introduced an agent capable of learning from user interactions and operating autonomously for extended periods. Dr. Werner Vogels, Amazon’s CTO, closed the event with a keynote focused on supporting developers and addressing concerns about AI’s impact on engineering roles.

AI Agents: Unlocking Business Value

The conference, running through December 5, began with AWS CEO Matt Garman highlighting the transformative potential of AI agents. He stressed that these agents are moving beyond simple assistants, now able to automate complex tasks and deliver tangible business results.

“AI assistants are evolving into agents that can execute tasks and automate processes for you,” Garman explained. “This is where organizations are starting to see significant returns on their AI investments.”

Keynotes and Customer Success Stories

On December 3, the focus remained on AI agents, with in-depth discussions on customer experiences. Swami Sivasubramanian, AWS’s VP of Agentic AI, delivered an optimistic keynote, stating, “We’re in an era of unprecedented change. For the first time, we can express our goals in natural language, and agents will devise the plan, write the code, utilize the necessary tools, and implement the solution. This accelerates innovation and removes barriers to turning ideas into impactful outcomes.”

Highlights and Ongoing Updates

While AI agents dominated the agenda, AWS revealed a host of other updates. Below is a summary of the most notable announcements. Stay tuned for ongoing updates as more insights emerge from AWS re:Invent.

Werner Vogels’ Farewell Keynote

Amazon CTO Werner Vogels delivered his final re:Invent keynote, clarifying that he isn’t leaving Amazon but believes it’s time for new voices after 14 years. He spent over an hour engaging with a packed audience, ending his speech with a memorable “Werner, out” and a mic drop.

AI and the Future of Work

Vogels addressed the common concern about AI replacing jobs, acknowledging that while automation will change certain roles and render some skills outdated, adaptability is key. “Will AI take my job? Maybe,” he mused. “But will AI make me obsolete? Absolutely not, as long as you continue to evolve.”

Introducing the Graviton5 CPU

AWS unveiled the Graviton5, its most powerful and efficient CPU to date. Featuring 192 processor cores, the chip’s compact design reduces data travel distance between cores, cutting communication delays by up to 33% and boosting bandwidth.

Enhanced Tools for Large Language Models

To support enterprises in developing custom AI models, AWS expanded capabilities in Amazon Bedrock and SageMaker. New serverless model customization in SageMaker lets developers build models without managing infrastructure, accessible via guided workflows or AI agent prompts. Bedrock now offers Reinforcement Fine Tuning, automating the customization process with preset workflows or reward systems.

Strong Performance from Trainium2 and a Look Ahead

Amazon CEO Andy Jassy highlighted the financial success of AWS’s Trainium2 AI chip, positioning it as a strong competitor to Nvidia. His remarks coincided with the introduction of the next-generation Trainium3 chip, signaling a promising future for AWS’s AI hardware.

Database Savings Plans Introduced

Among the many announcements, AWS introduced Database Savings Plans, enabling customers to cut database expenses by up to 35% when committing to consistent usage over a year. Savings are automatically applied to eligible services, with additional usage billed at standard rates. Corey Quinn, Duckbill’s chief cloud economist, remarked, “Six years of complaints have finally paid off.”

Free Credits for Startups Using Kiro

To attract startup founders, Amazon is offering a year of complimentary credits for its Kiro Pro+ coding tool to qualifying early-stage startups in select countries that apply before the month’s end.

Trainium3 and UltraServer: AI Training Innovations

AWS introduced the Trainium3 AI training chip and the UltraServer system, promising up to four times the performance for AI training and inference while reducing energy consumption by 40%. The company also teased the upcoming Trainium4, which will be compatible with Nvidia chips.

AgentCore Platform Upgrades

New features in the AgentCore platform now allow developers to set policies for AI agents, making it easier to define operational boundaries. Agents can also remember user preferences and are supported by 13 built-in evaluation systems to help customers assess their performance.

Frontier Agents: Autonomous AI Solutions

AWS launched three new “Frontier agents,” including the Kiro autonomous agent, which can learn team workflows and operate independently for extended periods. Other agents focus on security tasks like code reviews and DevOps functions such as incident prevention during code deployment. Preview versions are now available.

Nova AI Models and Nova Forge Service

The Nova AI model family has expanded with four new models—three for text generation and one capable of producing both text and images. The new Nova Forge service lets AWS customers access and further train pre-existing models with their proprietary data, emphasizing flexibility and customization.

Lyft’s Success with AI Agents

Lyft shared its experience using Anthropic’s Claude model via Amazon Bedrock to create an AI agent that manages driver and rider inquiries. The company reported an 87% reduction in average resolution time and a 70% increase in driver engagement with the AI agent this year.

AI Factories for Private Data Centers

Amazon introduced “AI Factories,” enabling large organizations and government entities to deploy AWS AI systems within their own data centers. Developed in collaboration with Nvidia, these systems support both Nvidia GPUs and Amazon’s Trainium3 chip, addressing concerns around data sovereignty and privacy.

Watch Industry Highlights from re:Invent

Catch up on the latest developments in agentic AI, cloud infrastructure, security, and more from AWS’s flagship event in Las Vegas. This video content is presented in partnership with AWS.