Nutanix launches NAI 2.8 and NKP 2.19 for enterprise agentic AI
- Nutanix announced general availability of NAI 2.8, featuring a generally available MCP Gateway, scalable multiGPU inference via tensor parallelism, and speculative decoding that accelerates LLM token generation by up to 2.5x
- NKP 2.19, expected soon, introduces NKP Metal for bare-metal Kubernetes, an AI Applications Catalog with one-click deployment for Kubeflow, Milvus, and Slurm, and CNCF certification
- Nutanix Unified Storage achieved NVIDIA-Certified Storage validation, establishing a low-latency, high-throughput data path to GPUs for large-scale AI workloads
- The Powered by Nutanix: Verified Services program and SP Central are now available to help partners build recurring revenue streams and deliver multitenant cloud and AI services
- The dual-native architecture is designed to let enterprises run VMs and containers side by side, bringing AI to existing applications and data without costly rearchitecting

*this image is generated using AI for illustrative purposes only.
Nutanix has announced the general availability of Nutanix Enterprise AI (NAI) 2.8 and the upcoming release of Nutanix Kubernetes Platform (NKP) 2.19, expanding its dual-native architecture to help enterprises deploy governed agentic AI across virtual machines and containers.
Many enterprises face a structural challenge when deploying AI: critical applications and data remain spread across both virtualised and containerised environments, forcing infrastructure silos or costly rearchitecting. Nutanix addresses this with a governed, dual-native architecture that runs traditional applications and modern AI side by side, integrated with silicon partners to provide choice and flexibility. By bringing AI to where enterprise data already lives, Nutanix aims to help customers reduce silos and accelerate return on investment without added networking and data layer complexity.
NAI 2.8: centralised control for AI inference and agentic workloads
NAI 2.8 delivers a unified and secure platform to deploy, manage, and scale AI workloads across hybrid environments. It enables enterprises to enforce governance over agents and models, gain visibility over token usage, and streamline AI development with built-in observability metrics and model-as-a-service capabilities.
Key updates in NAI 2.8 include:
| Feature | Description |
|---|---|
| Agent Gateway | Generally available MCP Gateway serving as a secure, unified front door for AI agents to access tools and data without custom engineering; includes MCP Server for NCP |
| Private Inference | Scalable multiGPU inference for LLMs via tensor parallelism; supports batch inference and speculative decoding |
| Parameter-Efficient Fine-Tuning | Supports Low-Rank Adaptation (LoRA) fine-tuning for models under 8B parameters using single-GPU compute |
| Speculative decoding | Accelerates LLM inference token generation by up to 2.5x using lightweight draft models |
| Enhanced security | Fine-grained Identity and Access Management, custom roles, model sharing, and support for air-gapped NVIDIA NIM deployment |
NKP 2.19: AI-optimised container management coming soon
NKP 2.19, expected to be available soon, is designed to simplify container operations across bare metal and virtualised environments without requiring complex custom stacks. It will provide an AI-optimised platform for building and running agentic applications at scale.
Upcoming features in NKP 2.19 include:
- NKP Metal: Brings HCI-grade simplicity to bare-metal Kubernetes with automated OS, firmware, and container deployment, and persistent enterprise-grade storage natively.
- NKP Full Stack: NKP on AHV combined with Nutanix Flow is designed to deliver stronger network-level sandboxing for AI agents, providing isolation to mitigate rogue attacks and lateral movement.
- AI Applications Catalog: Offers a one-click deployment path for curated, validated AI/ML software including Kubeflow, Milvus, and Slurm.
- Hardware and compliance: Planned expansion of ecosystem support with validated GPU integrations and dynamic resource allocation for AI workloads.
NKP has also attained formal Cloud Native Computing Foundation (CNCF) certification, validating that it provides the standardised APIs and capabilities required to operate enterprise AI workloads reliably.
Storage performance and partner ecosystem
Nutanix Unified Storage (NUS) has achieved NVIDIA-Certified Storage validation at the enterprise level, establishing a low-latency, high-throughput data path directly to GPUs to maximise GPU utilisation and support linear scalability for large-scale production AI workloads.
For partners, Nutanix launched the Powered by Nutanix: Verified Services program, designed to help partners build validated, high-margin services practices spanning the full customer lifecycle across hybrid cloud infrastructure, Kubernetes, and VM migration. Service Provider (SP) Central, also now generally available, provides a unified multitenant control plane giving service providers a consistent foundation to build and monetise infrastructure, application, cloud-native, and AI services.
Industry and customer perspectives
Thomas Cornely, Executive Vice President of Product Management at Nutanix, stated: "Enterprise AI should not require customers to rebuild the systems that already run their business. With our dual-native architecture, customers can bring AI to their existing applications and data while running each workload on the infrastructure best suited to it, with consistent operations and governance across VMs, containers and AI."
Matt Flug of IDC noted that Nutanix's dual-native model, which unifies services such as networking, security, and data protection across both VMs and containers, positions consistent policy enforcement and lifecycle management as key differentiators as application portfolios span legacy, refactored, and cloud-native designs.
NAI 2.8 and SP Central are generally available. NKP 2.19 will be available soon.
How might Nutanix's dual-native architecture impact the total cost of ownership for enterprises currently maintaining separate infrastructure silos for VMs and containers?
What competitive advantages does the NVIDIA-Certified Storage validation provide Nutanix against other hyperconverged infrastructure vendors in the high-performance AI inference market?
How could the introduction of NKP Metal and automated bare-metal Kubernetes management influence adoption rates among service providers looking to monetize AI workloads?
















