The Cloud Native Computing Foundation ( CNCF ) stated on Tuesday that it has made significant progress over the past year in terms of community growth globally, industry investment, and building a new generation of AI systems for production environments. CNCF mentioned that the newly added Sandbox projects include OpenShell and Agent Substrate, expanding its portfolio to 231 projects. At the same time, the community has seen an increase of over 5.5 million cloud-native developers, with 8 platinum members joining, and global events have attracted a total of 82,000 participants.

The organization stated that over the past year, cloud-native and CNCF activities have been held in more than 50 countries, attracting a total of 82,000 participants, setting a record; at the same time, 32 new projects have expanded the foundation's technical portfolio, including the new Sandbox projects Agent Substrate, llm-d, as well as OpenShell which provides a secure runtime for AI agents. New platinum members include Adyen, Broadcom, CoreWeave, CVS, HAProxy, NVIDIA, OVHcloud, and Vultr, covering chip manufacturers, neocloud, and enterprise users. These participants are supporting the infrastructure of the next generation of AI.

Executive Director Jonathan Bryce stated: "The AI team needs to control the location where the model runs, what the agents can do, and how their systems behave in a production environment."

He added, "A year ago, we were still discussing the possibilities of inference. Today, this community is building the infrastructure to train models, run inference, and get agents to work under the requirements of safe and scaled-up production. We are putting these efforts into an open environment, allowing developers to maintain their choice between models, hardware, and service providers. Bring your most challenging problems, share the experiences you've gained, and help us build the next step forward."

OpenShell Incorporates Runtime Security into Ecosystem Governance

OpenShell is incorporated into the CNCF project as Sandbox, aiming to bring the secure runtime of autonomous AI agents under a neutral, ecosystem-led governance framework.

OpenShell is led by NVIDIA and is responsible for managing how proxies execute tasks, which resources and data they can access, and where inference requests are routed to. It provides policy execution capabilities for the infrastructure surrounding the proxies, helping organizations to control access to files, networks, credentials, and model endpoints as proxies undertake increasingly complex tasks and operate on a larger scale.

This contribution enables developers, platform teams, and security professionals to directly participate in shaping the foundation of the AI runtime, and helps to address the challenges of security, control, and governance between the experimental and production phases.

Continuously expanding infrastructure: training, inference, and execution

CNCF The project contributions announced on the same day expanded the range of technologies available when building the infrastructure for AI.

• Agent Substrate provides a zero-trust runtime for executing the AI proxy on Kubernetes, while isolating untrusted code.
• gVisor enhances workload isolation through user-mode kernel, reducing exposure to the host operating system.
• KubeRay supports running and managing Ray workloads on Kubernetes, which is used for distributed computing, including AI training and inference.
• OpenRL introduces native orchestration of Kubernetes into reinforcement learning to help teams coordinate the computational resources required for post-training model development.
• OpenShell adds runtime governance and policy execution for autonomous proxies.

These projects are built upon significant progress made in the area of AI over the past 12 months:

• Distributed Reasoning: llm-d has joined with CNCF to release version 1.0, promoting an open framework for the deployment and expansion of distributed large language model reasoning.
• AI Lifecycle Maturity: Kubeflow graduation indicates that its ecosystem in data processing, model training, fine-tuning, and services has become mature.
• Accelerator Efficiency: HAMi has entered the incubation phase, supporting the sharing and scheduling of heterogeneous accelerators on Kubernetes.
• Infrastructure capabilities: Kubernetes has advanced the scheduling of workload perception and dynamic resource allocation, improving how complex AI workloads coordinate, compute, and access dedicated hardware.
• Interoperability: Kubernetes AI Conformance Program has reached 40 certified platforms, promoting community-led unified AI infrastructure requirements.

In a broader portfolio of projects, Cloud Native Buildpacks, Karmada, Kubeflow, KubeVirt, NATS, and OpenTelemetry have also been completed, while CloudNativePG, K8gb, Meshery, and Open Cluster Management have entered the incubation phase. CNCF indicates that these milestones have collectively strengthened the application delivery, orchestration, messaging, data, and operations infrastructure upon which the production system relies.

The global community continues to expand.

According to CNCF, in 2026, the global cloud-native developer community reached 21.2 million people, and cloud-native and CNCF events attracted 82,000 participants in over 50 countries. KubeCon + CloudNativeCon Europe set a new CNCF attendance record, with nearly 13,500 people participating. In addition, there were Kubernetes Community Days, offline gatherings, and thematic events that connected practitioners around the world.

The organization stated that this growth is expanding the community's influence and sharing of experiences. CNCF has added 58 new ambassadors and 28 cloud-native community groups, creating more opportunities for local participation. Community members have also released 40 new cloud-native case studies and 12 reference architectures, transforming production experience into practical guides that others can use as references.

Building the Future of Cloud and AI

CNCF Chief Technology Officer Chris Aniszczyk stated: "The people who built Kubernetes are now working side by side with a group of new contributors who focus on inference, training, and agents, which is evident from the growth of our projects and contributors."

He said, "The workload of AI is moving towards cloud-native infrastructure, and the patterns for scaling these workloads are being defined within the CNCF project. If you are a developer who wants to shape how AI will operate in a production environment, now is the time to get involved."

Looking to the future, CNCF also plans to launch two new AI-related certifications in 2027: Kubernetes AI Associate ( KAIA ) and Certified Kubernetes Inference Engineer ( CKIE ). CNCF indicates that these upcoming certifications will further expand its training and certification portfolio, as AI continues to integrate with cloud-native technologies.

CNCF invites developers, end-users, and technology providers to jointly build the next phase of cloud and AI: contribute code, share production experience, join community groups, and invest in projects that global organizations rely on.

The challenges of running AI in a production environment have arrived. Bring your most difficult problems – as well as your professional experience – to this community that is working on building solutions.

Supplementary resources

• Browse CNCF project: https :// www.cncf.io / projects /
• Contribute code to the CNCF project: https :// contribute.cncf.io /
• Join CNCF : https :// www.cncf.io / about / join /

About Cloud Native Computing Foundation

Cloud-native computing enables organizations to build and run scalable applications using open-source software stacks in public clouds, private clouds, and hybrid clouds. CNCF hosts key components of the global technology infrastructure, including Kubernetes, Prometheus, and Envoy. CNCF brings together developers, end-users, and suppliers to build a sustainable ecosystem for cloud-native software and is affiliated with the non-profit organization Linux Foundation. For more information, please visit www.cncf.io.

Media Contact

Linux Foundation