Amazon Web Services’ annual technology conference AWS re:Invent wrapped up another day with a deluge of product news and keynotes — plus the obligatory customer success stories.
The unsurprising topic is artificial intelligence for business. This year, it’s all about upgrades that give customers more control to customize AI agents, including one that AWS claims can learn from you and then work independently for days.
AWS re:Invent 2025, which runs through December 5, kicked off with a keynote from AWS CEO Matt Garman, who leaned into the idea that AI agents can unlock the “true value” of AI.
“AI assistants are starting to give way to AI agents that can perform tasks and automate on your behalf,” he said during the Dec. 2 keynote. “This is where we start to see tangible returns to your business from your AI investments.”
On December 3rd, the conference continued with AI agent messaging as well as deeper dives into customer stories. Swami Sivasubramanian, vice president of Agentic AI at AWS, gave one of the keynote speeches. To say it was sleazy might be understating the atmosphere.
“We live in times of great change,” Sivasubramanian said during the speech. “For the first time in history, we can describe what we want to achieve in natural language, and agents create the blueprint. They write the code, call the necessary tools, and execute the complete solution. Agents give you the freedom to build without boundaries, greatly accelerating how quickly you can go from idea to impact.”
While agent AI news promises to be a persistent presence throughout AWS re:Invent 2025, there have been other announcements. Here’s a rundown of the ones that caught our eye. TechCrunch will update this article, with the newest information at the top, until the end of AWS re:Invent. Be sure to check back.
Doubling of LLMs
AWS announced more tools for enterprise customers to build their own models. Specifically, AWS said it is adding new capabilities to both Amazon Bedrock and Amazon SageMaker AI to make it easier to build custom LLMs.
For example, AWS brings serverless model customization to SageMaker, which allows developers to start building a model without having to think about computing resources or infrastructure. Customization of the serverless model can be accessed either through a self-guided path or by prompting an AI agent.
AWS also announced Reinforcement Fine Tuning in Bedrock, which allows developers to select a predefined workflow or reward system and have Bedrock perform the tuning process automatically from start to finish.
Techcrunch event
San Francisco
|
13-15 October 2026
Andy Jassy shares some numbers
Amazon CEO Andy Jassy took to social networking platform X to explain AWS chief Matt Garman’s keynote. The message: The current generation of Nvidia’s rival Trainium2 AI chip is already bringing in a lot of cash.
His comments were tied to the unveiling of the next-generation chip, Trainium3, and were intended to predict a promising revenue future for the product.
Database saving is coming
Among the dozens of announcements is one element that is already excited: The discounts.
Specifically, AWS said it is launching Database Savings Plans, which help customers reduce database costs by up to 35% when they commit to a fixed amount of usage ($/hour) over a year. The company said the savings will automatically apply hourly for eligible usage on supported database services, and any excess usage beyond the commitment is billed with on-demand charges.
Corey Quinn, chief cloud economist at Duckbill, summed it up well in his blog post“Six years of complaining is finally paying off.”
There’s no better deal than free, Amazon hopes
Is there any way for another AI coding tool to win the hearts of startup founders? Amazon is hoping that a year’s worth of credits, free of charge, will do the trick for its Kiro offering. The company will give Kiro Pro+ credits to niche startups that apply for the deal before the end of the month. However, only early-stage start-ups in some countries are eligible.
An AI training chip and Nvidia compatibility
AWS introduced a new version of its AI training chip called Trainium3 along with an AI system called UltraServer that runs it. The TL;DR: This upgraded chip comes with some impressive specs, including the promise of up to 4x performance gains for both AI training and inference, while reducing power usage by 40%.
AWS also gave a teaser. The company already has Trainium4 in development, which will be able to work with Nvidia chips.
Expanded AgentCore capabilities
AWS announced new features in its AgentCore AI agent creation platform. A key feature is Policy in AgentCore, which enables developers to more easily set limits for AI agents.
AWS also announced that agents will now be able to record and remember things about their users. In addition, it announced that it will help its customers rate agents through 13 pre-built rating systems.
An unstoppable worker bee AI agent
AWS announced three new AI agents (there’s that term again) called “Frontier agents,” including one called the “Kiro autonomous agent” that writes code designed to learn how a team wants to operate so it can largely operate on its own for hours or days.
Another of these new agents handles security processes, such as code reviews, and the third does DevOps tasks, such as preventing incidents when pushing new code live. Preview versions of the agents are now available.
New Nova models and services
AWS is releasing four new AI models in the Nova AI family of models — three of which generate text and one that can generate text and images.
The company also announced a new service called Nova Forge that allows AWS cloud customers to access pre-trained, mid-trained, or post-trained models that they can then finish training on their own proprietary data. AWS’ big step is flexibility and customization.
Lyft’s argument for AI agents
The routing company was among several AWS customers that released during the event to share their success stories and evidence of how the products have impacted their business. Lyft uses Anthropic’s Claude model through Amazon Bedrock to build an AI agent that handles driver and rider questions and issues.
The company said this AI agent has reduced the average resolution time by 87%. Lyft also said it has seen a 70% increase in driver use of its AI agent this year.
An AI factory for the private data center
Amazon also announced “AI Factories” that allow large corporations and governments to run AWS AI systems in their own data centers.
The system was designed in collaboration with Nvidia and includes both Nvidia and AWS technology. While companies using it can equip it with Nvidia GPUs, they can also opt for Amazon’s newest homegrown AI chip, the Trainium3. The system is Amazon’s way of dealing with data sovereignty, or the need for governments and many companies to control their data and not share it, even using artificial intelligence.
See the latest revelations on everything from agent AI and cloud infrastructure to security and more from the flagship Amazon Web Services event in Las Vegas. This video is brought to you in partnership with AWS.
