Nutanix announced a new software solution designed to help organizations deploy agent AI at scale. With Nutanix Agentic AI, the company is targeting companies looking to go beyond traditional chatbots and integrate AI into more complex business processes.
SiliconANGLE reported. Many organizations use chatbots to support employees and customers via text or voice. These systems typically answer individual questions and perform simple tasks. But Nutanix says the focus is shifting to AI agents that independently navigate multiple steps, make complex decisions, and perform long-running tasks with minimal human intervention.
This development places different demands on the infrastructure. While AI systems for model training often focus on a single large-scale computational task, Nutanix points out that agent AI requires an environment that can support thousands of AI services and concurrent users. Additionally, the infrastructure must be flexible enough to continually accommodate new agents, applications, and developers.
Nvidia integration for AI agents
Nutanix Agentic AI is designed as part of a broader platform for AI workloads in enterprise environments. The software works with Nvidia technology to build and manage AI agents. This makes it easier for companies launching AI factories for large-scale AI development to deploy and scale new services.
This solution is built on existing Nutanix components, including the AHV hypervisor. This virtualization layer separates software from physical hardware, allowing virtual machines to run in an isolated and cost-effective manner. Organizations can also use Flow Virtual Networking and Kubernetes to develop, orchestrate, and manage AI agents within a cloud-like operating model.
On the infrastructure side, Nutanix has enhanced the platform with BlueField data processing units for high-performance network processing. These processors are designed to handle network traffic more efficiently while reducing host CPU and memory load.
This software also integrates with Kubernetes. This allows organizations to offer AI services through a platform-as-a-service model and offer models as separate services. According to Nutanix, this will allow enterprises to more accurately predict the cost of deploying large numbers of AI agents.
Lower cost per token
Tokens are an important cost element in AI applications. Tokens form the basis of how AI models process information and perform inferences. The more tokens a system uses, the more computing power it requires and the more it costs. Nutanix says more efficient virtual machines and better resource optimization should lower the price per token.
In addition to infrastructure, the platform also focuses on developers. Nutanix has extended its Kubernetes environment with a catalog of integration tools. Developers will have access to notebooks, vector databases, MLOps workflow engines, frameworks for agent AI, and more. Combined with Nvidia software, teams can build, test, and refine AI agents in a sandbox environment before going into production.
Nutanix says this approach facilitates the secure development and management of complex AI agents. Agentic AI works iteratively and takes multiple steps to complete a task. As a result, the inference process can quickly become complex. The company aims to use a comprehensive testing and development environment to ensure that systems do not exhibit unexpected behavior when deployed into production.
