Amazon Web Services (AWS) today at its re:Invent 2025 conference added a bevy of artificial intelligence (AI) capabilities to its portfolio, including a set of agents capable of autonomously performing a range of IT tasks.
AWS CEO Matt Garman told conference attendees he eventually expects there will be billions of AI agents running in IT environments, including a new generation of so-called long running frontier agents that will run continuously.
For example, AWS is previewing an AI agent within its Kiro AI coding tool that autonomously works in the background to optimize tasks such as pull requests and provide feedback to developers. There is also now an AWS Security Agent to review and test code and an AWS DevOps Agent that will always be on call to help manage IT incidents.
At the same time, AWS is expanding its existing Amazon Bedrock AgentCore for building AI agents to add tools for implementing guardrails as code and evaluating over time the behavior of an AI agent.
Additionally, AWS also announced it is making available AWS AI Factories, a set of integrated tools and platforms for building AI applications that it provides as a managed service using a combination of graphics processing units (GPUs) from NVIDIA and its own Trainium series of AI accelerators.
The cloud service provider is also making available the latest generation of NVIDIA GB300 NVL72 GPUs while also revealing a plan to build an alternative Trainium 4 processor that will provide six times the performance of Trainium 3 processors that it is now making available to provide up to 362 FP8 PFLOPs in an AWS UltraServer. Organizations that are using Trainium processors to reduce AI inference costs by up to 50% include Anthropic, Decart, Karakuri, Metagenomics, Neto.ai, Ricoh, and Splashmusic.
AWS also revealed it has added an additional 18 open weight models from providers such as Mistral in addition to a next-generation series of Nova 2 models that include lite, pro, sonic and omni editions that can be used to embed a wide range of AI functionality into applications that vary depending on cost and capability. For example, the Omni model makes it possible to embed multimodal functionality that enables text or speech inputs to generate video output.
Finally, AWS announced that it is making available Amazon Nova Forge, a service that enables organizations to build their own foundational model rather than customizing one that has already been created.
Overall, AWS claims that more than 100,000 organizations are now using Amazon Bedrock services to build AI applications.
There is, of course, no shortage of options for building AI applications. AWS is clearly counting on the fact that hundreds of thousands of organizations already rely on its cloud services to build and deploy applications, all of which will soon be augmented to include AI capabilities. The challenge and the opportunity now is for those organizations to find ways to accelerate the pace at which applications are developed without compromising the quality of the code generated to build them.
