From cloud native to agentic AI: Building what is next on Arm at KubeCon 2026

As cloud native technologies evolve to support AI and agentic workloads, Arm continues to play a growing role in the infrastructure and software ecosystems powering these applications.

With the majority of CNCF Graduated projects optimized for Arm, the architecture has become a mature, first-class platform for cloud-native computing.

This mature cloud native ecosystem also provides a foundation for AI applications. Technologies such as Kubernetes, containers and WebAssembly help organizations deploy, manage and isolate AI workloads at cloud scale.

As AI evolves toward autonomous agents that execute code, invoke tools, and coordinate tasks, demands on compute infrastructure are expanding.

CPUs play a critical role in orchestrating agent workflows, executing tools, scheduling workloads, and supporting efficient utilization of GPUs and AI accelerators. Arm-based CPUs provide scalable, energy-efficient compute to support these workloads across the broader agentic AI software stack, including inference and model serving.

Arm’s growing role in cloud infrastructure is reflected in adoption, with about 50% of compute shipped to leading hyperscalers in 2025 is Arm-based.

At KubeCon + CloudNativeCon North America 2026, we are highlighting how this foundation is enabling the next generation of AI and agentic applications.

Cloud native project enablement on Arm

Arm works closely with the cloud-native open-source community to expand Arm64 support across CNCF projects. Through upstream contributions, testing, and collaboration with project maintainers, Arm helps ensure that cloud-native technologies can be deployed and operated across Arm-based infrastructure.

Harbor and KubeVirt are two current examples of Arm’s ongoing collaboration with the CNCF community.

Harbor

A CNCF Graduated project, Harbor provides an open-source registry for managing and distributing container images, AI models, and other artifacts. As organizations deploy AI workloads at scale, Harbor helps secure the software supply chain through vulnerability scanning, access controls, and artifact signing.

Arm has collaborated with the Harbor community to expand Arm64 and multi-architecture support, helping bring these capabilities to Arm-based environments. Arm will share insights from this collaboration at the Harbor Project Maintainers Session, highlighting community contributions and what is next for Harbor on Arm.

KubeVirt

KubeVirt enables organizations to run virtual machines alongside containers on Kubernetes, allowing existing VM-based applications and cloud-native workloads to share a common orchestration platform. As an open, vendor-neutral virtualization solution, KubeVirt gives organizations greater flexibility in managing hybrid workloads.

Arm actively collaborates with the KubeVirt community on Arm64 enablement and optimization, helping expand multi-architecture support and infrastructure choices for cloud-native workloads.

Powering AI and agentic workloads on Arm

Building on this mature cloud-native foundation, Arm is enabling the next generation of AI and agentic applications. From secure agent execution to scalable infrastructure orchestration and automation, cloud-native technologies running on Arm provide the flexibility and efficiency needed to deploy and manage AI workloads.

The following examples highlight how Arm and its ecosystem partners are supporting this evolution.

Google Kubernetes Engine (GKE) - Agent Sandbox: Coding agents on Arm-based Google Axion Processors

Built on the open-source Agent Sandbox project within Kubernetes SIG Apps, GKE Agent Sandbox provides isolated execution environments for AI agents running potentially untrusted, AI-generated code. It helps organizations securely execute and scale agentic workloads on Kubernetes.

At KubeCon, Arm will demonstrate From Reasoning to Action, showing how AI agents turn a business request into a tested application update. Using Google Axion C4A and N4A processors, the Axion Agent Studio combines hosted Gemini planning with three specialized coding agents running local Gemma models inside GKE Agent Sandboxes. The agents work together to update an online shop, with independent validation of the results. A live dashboard shows agent activity, CPU utilization, and test results throughout the process.

Agent workflow showing how Gemini, Google Axion, and GKE Agent Sandbox coordinate coding, testing, and application updates

Cosmonic is working with the CNCF wasmCloud project to enable lightweight, scalable execution of AI agents and tools. Its open-source AgentSDK enables developers to isolate and sandbox individual components of AI agents, with support for building and scaling across Windows, Linux, macOS, and Kubernetes, including Arm-based environments.

At KubeCon, Cosmonic plans to demonstrate large-scale agent execution on AWS Graviton, highlighting the potential to scale to millions of sandboxed agents with scale-to-zero capabilities and tens of thousands running concurrently.

Cosmonic Session: Sandboxing on Kubernetes with WebAssembly: A Hands-on Tutorial - Wednesday, November 11, 4:10–5:30 PM MST.

Infrastructure orchestration and automation for AI on Arm

As AI and agentic workloads scale, infrastructure orchestration and automation become increasingly important for deploying and managing applications.

Rafay Systems provides Kubernetes cluster management and orchestration capabilities, helping organizations manage containerized workloads across cloud environments. Pulumi enables infrastructure provisioning and automation through infrastructure as code.

These solutions help simplify cloud-native deployments and infrastructure management across Arm-based environments.

What is next for cloud native AI?

As CNCF projects evolve to support AI, demand for efficient compute, secure execution, and scalable orchestration will continue to grow.

Arm Neoverse-based platforms continue to evolve alongside these requirements, with AWS Graviton5, Microsoft Azure Cobalt 200, Google Cloud Axion, and NVIDIA Vera expanding the range of Arm-based infrastructure for cloud and AI workloads.

Building on this momentum, Arm introduced the Arm AGI CPU earlier this year, extending the Neoverse platform beyond IP and compute subsystems with Arm-designed silicon built for cloud and agentic AI workloads. Looking ahead to 2027, Oracle Cloud Infrastructure (OCI) plans to offer AGI CPU-based compute, while AI cloud providers such as Verda are developing agentic-first infrastructure.

Together, these developments expand the options available to developers deploying cloud-native and AI workloads across Arm-based infrastructure.

Arm at KubeCon 2026

Join Arm at KubeCon + CloudNativeCon North America, November 9–12 in Salt Lake City, to explore the latest developments in cloud-native computing and AI.

Lightning Talk - Agent Sandbox in Action: Secure, Scalable Agent Execution on GKE and Axion

• Date & Time: Tuesday, November 10 | 1:00–1:15 PM MST.
• Location: Google Cloud Booth #122, Salt Palace Convention Center

Harbor Project - The Maintainers Session

Harbor Project Maintainers Session

• Date & Time: Wednesday, November 11 | 11:00–11:30 AM MST.
• Location: Salt Palace Convention Center | Room 251 A–C

Join Arm and other Harbor project maintainers to learn about the project's evolution, multi-architecture enablement, and community contributions.

Arm Demo - From Reasoning to Action: Arm-Powered Coding Agents with GKE Agent Sandbox and Axion

• Date & Time: Thursday, November 12 | 10:30 AM–1:30 PM MST
• Location: Google Cloud Booth #122, Salt Palace Convention Center

See Arm-powered coding agents turn a business request into a tested application update using Google Axion processors and GKE Agent Sandbox.

Ready to decompress after day one?

Join Arm and Datum to kick off KubeCon week with an informal evening of drinks, food, and conversation with the cloud native community.

Datum x Arm KubeCon Social

• Date & Time: Tuesday, November 10 | 6:00 PM MST
• Location: Registration required; location details provided upon registration.

Explore the KubeCon 2026 event program for additional sessions.

Resources for building cloud-native applications on Arm

Explore these resources to build and deploy cloud-native workloads on Arm:

• Arm Software Ecosystem Dashboard: Check Arm64 support across cloud-native, AI/ML, and agentic AI software.
• Arm Learning Paths: Hands-on tutorials for cloud-native development:
  • WebAssembly with wasmCloud on Arm: Learn to deploy and run WebAssembly workloads with wasmCloud on Arm-based infrastructure.
  • Kubernetes management with Rafay on AWS Graviton: Provision and manage Amazon EKS clusters.
  • Infrastructure as Code with Pulumi on Azure : Automate cloud resource provisioning.
• Arm Developer: Development tools, documentation, and community resources.