Today’s applications, tomorrow’s AI workloads: Nutanix is building the platform for both, says CEO

2026-04-08 · Source: AI – SiliconANGLE · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Cloud Computing & IT Infrastructure, Robotics & Autonomous Systems · Depth: Intermediate, short

Summary

Nutanix Inc. is evolving its enterprise computing platform to address the shift towards agentic infrastructure, which is becoming the organizing logic for new applications. According to CEO Rajiv Ramaswami, this transformation requires infrastructure to not only host workloads but also orchestrate intelligent agents, govern data pipelines, and optimize inference economics. Nutanix aims to be the platform of choice for all applications, including current and future AI workloads. The company's platform expansion, announced at Nutanix .NEXT 2026, focuses on maximizing GPU utilization to reduce the cost per token, mirroring its past success with CPU virtualization. Nutanix is also building a robust ecosystem, evidenced by over 100 partners at .NEXT and a $250 million strategic investment from Advanced Micro Devices Inc. for co-developing an open agentic AI platform. Additionally, Nutanix is capitalizing on the growing demand for sovereign AI clouds, positioning its hybrid platform as a solution for governments seeking to control their data and infrastructure within national borders.

Key takeaway

For AI Architects and Directors of AI/ML evaluating infrastructure for agentic workloads, your focus must shift to platforms that demonstrate clear GPU optimization and robust data governance. Nutanix's emphasis on maximizing GPU utilization and supporting sovereign cloud deployments suggests a strategic alignment with future enterprise AI needs. You should investigate how their platform's architecture directly addresses cost-per-token economics and integrates with a broad partner ecosystem to ensure long-term scalability and control over your AI initiatives.

Key insights

Agentic infrastructure demands platforms that optimize GPU utilization and govern data pipelines to reduce cost per token.

Principles

• Maximize GPU utilization to reduce cost per token.
• Platform value is tied to its surrounding ecosystem.

Method

Nutanix applies optimization logic from CPU virtualization to GPU workloads, focusing on eliminating idle compute to improve efficiency and reduce cost per token for agentic infrastructure.

In practice

• Prioritize GPU efficiency in AI infrastructure design.
• Consider hybrid cloud for sovereign AI deployments.

Topics

• Agentic Infrastructure
• GPU Optimization
• Sovereign AI Clouds
• Nutanix Platform
• Cost per Token

Best for: AI Architect, Director of AI/ML, MLOps Engineer

Related on AIssential

• How Agentic AI Platforms Organize Their Hardware Infrastructure — On-premises GPU clusters provide optimal control, performance, and cost predictability for agentic AI pipelines.
• NetApp and Nutanix say storage has become the last line of defense in the AI era — NetApp and Nutanix partner to integrate data management and cloud platforms for AI-ready, secure, and modernized infrastructure.
• The Future Live | 04.03.26 | Guests from BEP, Ornn, and MOTS Podcast! — The AI landscape is rapidly evolving with powerful open-source models, specialized agentic tools, and new financial infrastructure for compute.
• Google Cloud sees its marketplace as the launchpad for the agentic enterprise — The agentic enterprise, driven by autonomous AI, represents a fundamental economic and organizational rewiring.
• Building for the Rising Complexity of Agentic Systems with Extreme Co-Design — Agentic AI systems demand specialized infrastructure to manage high token consumption and complex, probabilistic workloads efficiently.
• Building Token‑Metered AI Services on Telco AI Factories — Telcos can significantly boost AI infrastructure revenue by shifting from GPU-hour billing to token-metered AI services.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by AI – SiliconANGLE.