Nutanix Inc., a hybrid cloud converged infrastructure computing leader, today announced a software solution aimed to help enterprise clients accelerate their transformation toward adopting agentic artificial intelligence across their business processes.

Today’s release, Nutanix Agentic AI, integrates with Nvidia Corp.’s agent builder to address the needs of enterprise firms as they build AI factory infrastructure and scale to meet new demands.

A multitude of enterprise outfits have already brought chatbots into their workflows, providing conversational interfaces that can give contextual guidance to employees and customers through text and voice.

However, these one-shot question-and-answer bots have started to give way to a new paradigm of AI agents that can reason over complex tasks, work autonomously and complete long-term goals with little or no human oversight. This transformation requires a new type of software infrastructure and platform design that delivers low latency while maximizing performance and still maintaining security and compliance.

“Contrary to AI infrastructure for model training that was optimized to run ‘one big job,’ production Agentic AI infrastructure needs to handle scale and high rates of change for thousands of AI services, agents, and concurrent users and developers,” said Executive Vice President of Product Management Thomas Cornely.

Cornely added that the new solution extends its already existing solutions, AHV hypervisor, part of the Nutanix platform, which provides cost-effective decoupling of software from physical hardware to run isolated virtual machines. Users can also incorporate Flow Virtual Networking, Kubernetes and enterprise AI solutions to strategize on-time cloud operating models to build, orchestrate and govern AI agent fleets.

The company’s backend infrastructure has been enhanced with BlueField data processing units for high-performance networking to reduce host compute and memory consumption costs.

Using the Kurbernetes platform, the system provides AI platform-as-a-service, model-as-a-service and Kubernetes-native software layers to drive predictable token costs when deploying fleets of agents.

It is not an uncommon experience for AI engineers and data scientists to build an agentic AI experience for clients or customers to discover that AI agents’ “thinking” can get messy. The one-shot reasoning by chatbots is easy to follow, it’s comparatively clean and happens only once. However, agentic AI is a multi-turn process that steps through tasks and iterates over time, generating complex reasoning trees that can spiral wildly out of control without careful testing and regulation.

When it comes to AI processing, tokens are the currency of AI reasoning; they measure the inherent economy of agentic AI. The more tokens that get used, the more power is consumed and costs go up. The company said enterprise users will benefit from virtual machines with greater performance and enhanced infrastructure with better resource optimization to achieve a lower cost per token.

Nutanix added that it also simplified the path to delivering control by extending its Kubernetes platform with a catalog of prebuilt developer tools. These include Notebooks, vector databases, MLOps workflow engines and agentic frameworks. By fully integrating with Nvidia’s enterprise software, developers can test, build and iterate with NIMs, including Nemotron, to develop in sandboxes safely before going live.

Image: SiliconANGLE/Microsoft Designer