AI services from Nutanix now also for the public cloud

With GPT-in-a-Box, hybrid cloud specialist Nutanix has been offering an integrated AI stack for local use on hyperconverged infrastructure for just over a year. Now the service is being expanded for cloud-native environments based on Kubernetes.

Thanks to the Nutanix Enterprise AI (NAI) service, customers will now be able to use the manufacturer's AI infrastructure platform not only in the Kubernetes clusters on the Nutanix infrastructure in their data center or in edge environments. But now also in managed Kubernetes environments of major cloud providers such as AWS EKS, Azure AKS and Google GKE. According to the manufacturer, customers benefit from a consistent operating model across providers, freedom of choice of environment to use models and data, and rapid deployment within minutes instead of days or weeks.

Technically, Nutanix Enterprise AI is based on in-house services such as the HCI platform "Nutanix Cloud Infrastructure" (NCI), the Nutanix Kubernetes platform NKP as well as the block, object and file storage services "Nutanix Unified Storage" (NUS). It is also based on the Inference Microservices from Nvidia (NIM) and the open source base models from Hugging Face. The basic GPT-in-a-Box offering has already received Nvidia certification.

Provider promises secure use and a great deal of flexibility

Nutanix sees the advantage for users in the simplified setup of a GenAI infrastructure for their own company with the necessary security features and great flexibility in the operating model. So, customers can choose between classic on-premises operation and hosting in the public cloud and also switch if necessary. This should also eliminate data protection and security concerns, as customers can also operate the models and data in environments that they can fully control and manage.

At the same time, an easy-to-use dashboard is available for error detection and elimination as well as optimization and control, and Nutanix Enterprise AI also offers role-based access controls via an RBAC model for controlled access to the LLMs.

(nie)