[AWS re:Invent] Matt Garman CEO: "AI Agents Will Be a Turning Point in the AI Era"

Day Two Keynote Highlights AI Agent Services
"Set to Become Mainstream in Everyday Life and Work"
New Features Added to AI Agent Development Tools
Proprietary AI Chip 'Trainium' Performance Quadrupled

"Artificial intelligence (AI) agents are replacing traditional AI assistants. I believe the emergence of AI agents will bring about innovation as significant as the advent of early AI itself."

Matt Garman, Chief Executive Officer (CEO) of Amazon Web Services (AWS), made these remarks during his keynote speech at AWS's annual event, re:Invent 2025, held on December 2, 2025 (local time) in Las Vegas, Nevada, USA. He emphasized that while existing AI assistants have been limited to supporting some aspects of users' work, the arrival of AI agents capable of taking on tasks on behalf of users marks a turning point that will transform the trajectory of the AI era.

Matt Garman, Chief Executive Officer (CEO) of Amazon Web Services (AWS), is delivering the keynote speech at AWS's annual event, re:Invent 2025, held on the 2nd (local time) in Las Vegas, Nevada, USA. Provided by AWS

AWS highlighted that by putting AI agents at the forefront of its services, it can fundamentally change how everyone works through AI agents that can make decisions, plan, and even take action independently. The new services unveiled by AWS on this day also focused on integrating AI agents.

In particular, CEO Garman predicted, "AI agents will soon become mainstream in both our work and daily lives." He explained that the emergence of AI agents means that "(AI) is evolving from a technological marvel into something that delivers real value to us," and projected that "in the future, there will be billions of AI agents spanning all companies and functions."

CEO Garman emphasized that AWS has the optimal solutions for customers to adopt agentic AI. He stated, "Both hardware (HW) and software (SW) are required for the adoption of agentic AI," and added, "We have everything necessary for this." In addition to services like Bedrock AgentCore-a tool for developing and deploying AI agents-AWS operates infrastructure such as data centers and GPU-as-a-Service (GPUaaS). In the hardware sector, AWS designs its own AI semiconductors, such as Trainium and Inferentia.

Select Your Desired Model to Build AI Agents... Proprietary Chip Performance Quadrupled

Matt Garman, Chief Executive Officer (CEO) of Amazon Web Services (AWS), announced the company's proprietary artificial intelligence (AI) chip, "Trainium 3," at AWS's annual event "re:Invent 2025," held on the 2nd (local time) in Las Vegas, Nevada, USA. Provided by AWS

During his keynote, he introduced new features added to Bedrock AgentCore, a tool for developing AI agents. AgentCore is a fully managed platform that supports the entire process of building, deploying, and operating AI agents, and it was officially launched in October 2025. Companies can use AgentCore to select their preferred models and frameworks to build AI agents. Newly announced features include "Policy in AgentCore," which sets operational conditions for AI agents, and "AgentCore Evaluation," a tool for assessing the quality of AI agents.

Significant performance improvements have been made in AI semiconductors, which are considered essential infrastructure for developing AI agents. Notably, AWS unveiled the new model of its proprietary AI chip, Trainium 3, on this day. Trainium is a semiconductor designed by AWS specifically for AI model training. Compared to its predecessor, Trainium 3 utilizes a 3nm (nanometer; one billionth of a meter) process, increasing computing performance by more than four times while reducing power consumption by about 40%. CEO Garman emphasized, "Trainium 3 delivers industry-leading cost efficiency for large-scale AI training and inference."

The launch of the next-generation chip, Trainium 4, was also announced. This chip is expected to deliver six times the computing performance of Trainium 3. He stated, "AWS has long known that AI chips would be a game changer," adding, "That is why we began investing in custom chips ten years ago." AWS entered proprietary chip development in 2015 by acquiring the Israeli semiconductor company Annapurna Labs.

A range of AI agents to assist with work during the development process was also introduced. These include "CodeWhisperer Autonomous Agent," a tool that automates repetitive tasks for developers; "AWS Security Agent," which automates security verification tests; and "AWS DevOps Agent," which supports development and operations processes. An AI agent-based "custom" feature was also added to "AWS Transform," which rewrites legacy code into modern programming languages. This allows the AI agent to reflect each company's unique characteristics when modernizing old code.

Proprietary AI Models Updated... Latest AI Infrastructure Embedded in Customer Environments

Matt Garman, Chief Executive Officer (CEO) of Amazon Web Services (AWS), is delivering the keynote speech at AWS's annual event, re:Invent 2025, held on the 2nd (local time) in Las Vegas, Nevada, USA. Provided by AWS

AWS's proprietary AI model, Amazon Nova, has also been upgraded to the next-generation Nova 2. The Nova 2 lineup includes: "Lite," a low-cost model; "Pro," a high-performance inference model; "Sonic," a speech-to-speech model that receives and responds in voice; and "Omni," designed for multimodal inference and image generation.

The open training model "Nova Forge," which allows companies to tailor the training process to their needs, was also introduced. CEO Garman explained, "During the training phase of AI models, both preset data and a company's proprietary data can be used," adding, "This enables the model to retain its core knowledge while deeply understanding each company's unique information."

The "AI Factories" service, which converts customers' existing infrastructure into high-performance AI infrastructure, was also unveiled. This service enables companies or organizations to build AWS's AI infrastructure within their own data centers. AWS offers a variety of AI infrastructure options, including NVIDIA GPUs and Trainium, and AI Factories help alleviate concerns about data leakage during the training process.

New models such as "Mistral 3" and "Mistral Large 3" from France's Mistral AI have also been added to Amazon Bedrock, which allows users to access multiple AI models on a single platform.

Meanwhile, Matt Garman's keynote on this day featured Shantanu Narayen, CEO of Adobe; John Kodera, Chief Digital Officer (CDO) of Sony Group; and May Habib, CEO of Writer. They presented use cases of developing AI agent services using AWS offerings such as Bedrock and AgentCore.