en.Wedoany.com Reported - Nutanix has partnered with AMD to launch enterprise AI infrastructure, combining AMD processors and accelerators to help enterprises run agentic AI workloads on private and hybrid infrastructure.
As autonomous AI agents enter production environments, this solution aims to give enterprises greater control over model selection, data governance, and computing costs. Nutanix stated that organizations can adopt a two-tier model strategy, using cutting-edge AI systems for complex reasoning tasks while high-volume workloads are processed by optimized models running on private infrastructure.
Autonomous agents can retrieve information, interact with enterprise applications, and execute long-running workflows, activities that may generate a large number of model requests. When organizations primarily rely on leased infrastructure or external model services, the resulting token usage can increase operational costs.
Nutanix proposes using an agent gateway to route requests between different models and infrastructure environments. The gateway can apply organizational policies, manage access controls, and determine which workloads require external frontier models. Routine requests can be routed to models running within the organization's own infrastructure, while more complex tasks can be sent to larger external models. This structure aims to reduce the proportion of workloads processed through higher-cost models and provide organizations with a central control point for managing model access and usage policies.
Debo Dutta, Chief AI Officer at Nutanix, stated that owning one's intelligence does not mean completely abandoning frontier models, but rather mastering control over routing, capacity, and costs. Expensive frontier models can be used only for a small portion of tasks requiring complex edge-case reasoning, while most high-volume tasks are routed to highly optimized models running securely on private infrastructure.
The architecture combines the Nutanix Cloud Platform and Nutanix Enterprise AI with AMD EPYC processors and AMD Instinct MI355X accelerators. Nutanix Enterprise AI is used to deploy and run AI models and applications in a governed infrastructure environment, while the Cloud Platform provides the underlying compute, storage, and management layer for workloads running across private and hybrid environments. Keeping selected AI workloads within private infrastructure helps organizations maintain control over proprietary information and supports data sovereignty requirements. The platform is designed to support long-running agents and can apply policies to determine which models agents can use, what information they can access, and where workloads are processed.
AMD EPYC processors and AMD Instinct MI355X accelerators provide computing resources for model inference and enterprise workloads. Nutanix stated that the combined infrastructure can support a shared inference environment without requiring separate systems for each AI application or business unit. Shared capacity allows organizations to allocate computing resources across multiple workloads, improving infrastructure utilization. The design targets enterprises that want to scale AI deployments but do not want to dedicate hardware to each model or agent. Centralized infrastructure can serve multiple applications while managing governance and access policies through the platform.
Enterprise AI projects are moving from isolated experiments to systems that run continuously within business processes. Agent systems have different infrastructure requirements than standalone chatbots or occasional model queries. Agents may remain active for extended periods, consulting multiple sources and issuing repeated requests before completing tasks. As deployments scale, these operating patterns can increase compute and token costs, as well as add complexity to model governance. Nutanix's two-tier structure separates high-volume processing from tasks requiring larger frontier models, allowing organizations to retain the option of using external systems while handling more predictable workloads through privately operated models. The platform also allows AI workloads and proprietary data to remain in a governed hybrid multi-cloud environment, with access, routing, and policy controls managed through the agent gateway and Nutanix software stack.
Nutanix stated that the shared inference infrastructure is designed to maximize utilization of AMD EPYC CPUs and AMD Instinct MI355X GPUs, while supporting long-running agents and controlling token-related costs.
