The new offering adds an Opinionated AI stack and related services to Nutanix’s infrastructure and storage stacks, and the company believes it will find a good reception from the SMB to the enterprise.
Manosiz Bhattacharyya, Nutanix’s CTO
Today, Nutanix is announcing Nutanix GPT-in-a-Box, a full-stack software-defined AI-ready platform designed for customers who are looking to jump-start artificial intelligence [AI] and machine learning [ML] innovation, but who don’t know quite where to start.
“Nutanix is primarily an infrastructure player, with a parallel storage stack,” said Manosiz Bhattacharyya, Nutanix’s CTO. “The infrastructure stack and the storage stack are both the same as we use here. What is new is the addition of an Opinionated AI stack on top, and services that in turn go on top of it.”
This makes the solution a stack of Nutanix Cloud Infrastructure, Nutanix Files and Objects storage, and Nutanix AHV hypervisor and Kubernetes platform with NVIDIA GPU acceleration, all on the Nutanix Cloud Platform. 78% of Nutanix customers indicated that they were likely to run their AI and ML workloads on the Nutanix Cloud Platform.
“The AI stack lets you deploy, fine tune and do inference,” Bhattacharyya said. There are foundational base models for LLM, which include all the standard open source models and a curated set of large language models like Llama2, Falcon GPT and MosaicML.
“Customers want to use AI but don’t know where to start,” Bhattacharyya indicated. “There are tons of tools today, although some, like PyTorch, have become more popular than others, like TensorFlow. And we also bundle in services, so it’s not just the infrastructure stack or the AI stack, but the services as well. What we are doing is making sure we give you the right tools with the services, and this combination of services and the right tools makes it easy and lets the customer hit the ground running.”
Nutanix sees the potential market for their GPT-in-a-Box solution as fairly large.
“We are trying to target the higher end of SMB through to larger customers,” Bhattacharyya stated. “We want to take it downstream as well. This kind of AI has been typically seen in larger markets, but there are areas like fraud detection where the use can be much broader because it makes a lot of sense in smaller markets.”
Bhattacharyya said that many prospects for this typically have both data scentists and IT, but they are siloed separately.
“The data scientists don’t understand the infrastructure side of things and IT doesn’t know what the data scientists want,” he commented. “We give them a solution where the data scientists can do their models without bothering IT.”
At the same time, Bhattacharyya cautioned that Nutanix is well aware that Generative AI today can’t have the final decision on anything important.
“In terms of where we think Generative AI is going, we see it as more of a helping tool for specialists right now,” he said. “For example, in pharmacy manufacturer, you can use it to check regulations, but a person is charge has to have the authority to implement it. What has improved is the fewer number of steps that a human has to do. But there is no way at this juncture that AI can be given the authority on its own to make decisions that affect things that matter.”
There are no immediate plans to follow up on GPT-in-a-Box with near-term roadmap solutions.
“As of now, there is no extension planned to what we are offering here,” Bhattacharyya said. “We think that will suffice for a majority of customers – those who want a simple infrastructure and a stack. We may announce additional things later on, but not now.”
Service integrator partners have been particularly excited about the new solution.
“I see a lot of traction with them because services are a big part of this,” Bhattacharyya indicated. “Most companies and most industries do not have people who can create a fine-tuned AI model by themselves. We believe that service integrators can play a big part with this, and they are excited.”
