Earlier this year, I outlined our strategic commitment to strengthening the cloud-native ecosystem through upstream collaboration. Enterprise technology is fundamentally stronger when developed openly, shaped by diverse communities, and engineered for cross-environment interoperability rather than single-platform lock-in.

Heading into KubeCon + CloudNativeCon North America 2026, we are demonstrating how that commitment translates into operational practice.

A central focus across these initiatives is driving ecosystem-wide consistency. As Kubernetes adoption expands, organizations require unified patterns not only for infrastructure management, but also for application runtime, data protection, and core service consumption.

From advancing a provider-neutral VM API and establishing workload conformance standards to refining data protection and storage interfaces, our engineering focus targets challenges that transcend individual vendor ecosystems.

While these projects span different maturity stages, they align under a single architectural goal: making Kubernetes more consistent, scalable, and interoperable through open collaboration.

Creating a Provider-Neutral Kubernetes API for Virtual Machines

As Kubernetes expands beyond containers, virtual machines need a first-class, Kubernetes-native API that preserves their distinct lifecycle, performance, and I/O characteristics.

VM API builds on Broadcom’s VM Operator, which introduced a Kubernetes-native approach to provisioning and managing virtual machines on the vSphere component of VMware Cloud Foundation. The project extends that foundation towards a common, provider-neutral API for managing VMs across different infrastructure.

Similar to Cluster API’s approach to managing Kubernetes clusters, VM API separates common VM lifecycle definitions from provider-specific implementations, enabling declarative Kubernetes workflows while preserving underlying infrastructure capabilities.

VM API’s acceptance into the CNCF Sandbox marks an important milestone toward vendor-neutral governance and broader community collaboration.

As emerging workloads, including AI agents executing untrusted code, drive demand for stronger isolation, VM API could provide a consistent way to provision and manage virtual machines while relying on the underlying virtualization platforms for security and isolation capabilities.

Extending Kubernetes Conformance Beyond the Cluster

Standard Kubernetes conformance establishes a common baseline for Kubernetes API compatibility, but it does not independently validate application-level portability or consistent workload behavior. Consequently, enterprise platform engineering teams and independent software vendors (ISVs) can face repeated validation efforts across heterogeneous Kubernetes environments.

To help address this gap, Broadcom is co-chairing the new Kubernetes Workload Conformance Working Group. The working group aims to develop common, verifiable expectations for workload behavior across Kubernetes environments, helping address the validation challenges faced by enterprises and ISVs.

By advancing this work upstream, the community has an opportunity to reduce duplicated validation efforts, improve consistency, and establish a stronger foundation for application portability.

Over time, this approach could help ISVs validate workloads against common requirements while giving platform teams greater confidence in how applications behave across different Kubernetes distributions.

Making Data Protection More Efficient for Stateful Kubernetes Workloads

Our continued engineering contributions to Velero address evolving data protection requirements as Kubernetes supports increasingly large, enterprise-grade stateful workloads.

Velero 1.19 introduces block-level data movement capabilities designed to improve the efficiency of backup and restore operations for large, stateful Kubernetes workloads.

Data protection for these applications presents distinct performance challenges. As persistent volumes grow, file-system-level backup approaches can introduce significant resource and latency overhead.

By operating at the block level rather than relying exclusively on file-system traversal, these capabilities can reduce data movement overhead for supported workloads and storage environments, helping improve backup and restore efficiency.

Velero’s transition into the CNCF Sandbox established a neutral governance foundation. Our ongoing contributions demonstrate how participation in CNCF projects supports continued engineering investment, technical evolution, and community collaboration.

Building a Consistent Kubernetes Model for Object Storage

While CSI standardized persistent block and file storage for Kubernetes, object storage historically lacked a unified, native consumption specification across cloud and on-premises environments.

The Container Object Storage Interface (COSI) project addresses this gap. Broadcom engineers are contributing upstream to advance COSI toward Beta, providing a Kubernetes-native approach to provisioning object storage and managing access across infrastructure environments.

In parallel, we are developing the vSphere COSI Driver, an open-source driver implementing the COSI specification for enterprise vSphere infrastructure.

As cloud-native architectures, data platforms, and AI systems increasingly rely on object stores, standardized interfaces deliver operational consistency while preserving underlying storage innovation.

Building a Secure Runtime for Agentic Workloads

As autonomous AI agents execute code dynamically and interface with infrastructure, Kubernetes must adapt to meet strict isolation and high-density performance requirements. Unlike traditional microservices, agentic workloads exhibit bursty execution patterns, maintain state across invocations, and require strong runtime sandboxing.

Agent Substrate, a CNCF Sandbox initiative, addresses these demands using lightweight virtualization and runtime sandboxing (such as gVisor) to enforce isolation boundaries, enable fast suspend/resume state management, and maximize hardware utilization efficiency.

Broadcom is actively collaborating with the Agent Substrate community and evaluating integration paths with Kubernetes control planes via the VKS add-on framework. Our engineering goal is to deliver secure, enterprise-grade runtime environments for next-generation AI workloads.

To illustrate how these technologies can work together, the following demonstration explores Agent Substrate in VMware vSphere Kubernetes Service (VKS) environment: https://youtu.be/mDfZmFEvCdw. It offers a closer look at how Kubernetes-native workflows can support emerging agentic AI workloads.

Contribution Is a Continuing Process

Upstream collaboration requires adaptable engagement models across project lifecycles. Whether contributing core engineering to established specifications, transitioning internal projects to neutral foundations, or publishing open-source components, effective open collaboration centers on shared industry challenges.

Broadcom remains committed to dedicating engineering capacity to upstream projects, participating actively in community feedback loops, and incorporating open standards back into our enterprise platforms.

We look forward to engaging with the community at KubeCon + CloudNativeCon North America 2026 to discuss these technical developments and future ecosystem requirements.

To connect with our engineering team during the conference, review our technical sessions and schedule in the KubeCon North America 2026 event preview.

We look forward to collaborating with you.