Nutanix Inc. today is breaking out the inferencing components of its GPT-in-a-Box toolkit in the form of a cloud-native artificial intelligence infrastructure platform that runs on any Kubernetes installation.

That can be at the edge, in private data centers and on public cloud services like Amazon Web Services Inc.’s Elastic Kubernetes Service, Microsoft Corp.’s Azure Kubernetes Service and Google LLC’s Google Kubernetes Engine.

Nutanix Enterprise AI provides a consistent multi-cloud operating model that can cut the deployment of generative AI applications from days to minutes, the company said.

“It gives you automation to stand up an inference endpoint, download a model and deploy the systems in a secure and controlled way,” said Thomas Cornely, senior vice president of product management at Nutanix.

Introduced last summer, GPT-in-a-Box is positioned as a full-stack, software-defined platform that includes all the elements needed to build AI-ready infrastructure. It’s intended for companies building models from scratch, “but a lot of people are starting in the cloud,” Cornley said. “We can meet them where they are.”

Nutanix Enterprise AI does away with the infrastructure, Nutanix Kubernetes Platform, and Nutanix Unified Storage components of the integrated offering, leaving it to customers to provide those elements.

The release is a break from tradition for Nutanix. “It’s the first time we’ve released something that’s cloud-native on day one,” Cornley said. “We more typically work on-prem and extend in the cloud.”

The offering is aimed at data scientists, who often have to provision infrastructure for generative AI models. “That’s not something data scientists are good at,” Cornley said. “They want to focus on the application and consuming the model, not standing up the infrastructure.”

Generative AI workloads are often inherently hybrid, with applications built in the public cloud, fine-tuning occurring on-premises and inferencing determined by business need. That presents issues of complexity, data privacy, security and cost that Nutanix is addressing.

Simple and consistent

Simplicity and consistency are “true to the Nutanix story,” Cornley said. “It’s about how to make it simple for customers who don’t want the risk of things not doing what they’re supposed to do. This is about deploying in a predictable, cost-effective fashion and keeping in control.”

Nutanix Enterprise AI comes with built in support for Nvidia Corp.’s AI Microservices, a set of services that help deploy AI models consistently across different platforms. It also supports open-source foundation models from Hugging Face Inc. However, customers can deploy whatever LLMS they choose.

Nutanix Enterprise is being offered under what the company calls a “transparent and predictable pricing model” based on the infrastructure customers use, such as the number of CPU cores and graphic processing units. Provisioning is done in a point-and-click workflow similar to the way cloud servers are provisioned

Resource-based pricing contrasts with most cloud services, which price based on usage. “We see a lot of customers reacting to AI the same way they first reacted to the cloud; they don’t what they are getting and how much it’s going to cost,” Cornley said. “This is a subscription term license paid per year based on resources. That’s common to entire portfolio.” Pricing specifics weren’t provided.

Photo: Flickr CC