Open source is the foundation for AI and, as AI workloads scale, developers need that foundation to be more secure, more predictable, and easier to build apps and agents.
At Open Source Summit North America 2026, we’re announcing two updates that strengthen exactly that: the upcoming public preview of Azure Linux 4.0 on Azure Virtual Machines and the general availability of Azure Container Linux, our immutable container-optimized operating system (OS), with the broader rollout at Microsoft Build on June 2. Together, they give developers and organizations a hardened Linux distribution purpose-built for cloud native and AI workloads.
That foundation is also what makes the next chapter possible. The move from cloud native to AI native is the next evolution of open source, and it’s the focus of my keynote this week, From Open Source to Agentic Systems: Building the AI Native Era. Open source is already at the core of AI today, and AI in turn is reshaping how open source itself gets built, from how we collaborate, to how we test, to how the developer experience comes together. We’ve done this before. We know what it takes to build an open ecosystem at scale, and we know that openness is what makes it work.
Linux, Kubernetes, and containers have made the modern cloud possible. Every hyperscaler, every AI training cluster, every inference endpoint serving millions of tokens a second is built on open source. Open standards, shared governance, and community innovation have been the way an ecosystem of this scale comes together, allowing the best ideas, from anywhere, to compound.
For Microsoft, this story started in earnest in 2009, when we contributed more than 20,000 lines of Hyper-V driver code to the Linux kernel. A small patch then, but a clear signal of where we were headed.
Today, more than two-thirds of customer cores in Azure run Linux, and the platforms running Microsoft 365, GitHub, and OpenAI’s ChatGPT all sit on Linux foundations. When ChatGPT scales across more than 10 million compute cores worldwide and serves a billion queries a day, Linux and Kubernetes are what make that possible.
For developers running modern workloads on Azure, the OS layer should be invisible: secure by default, consistent across hosts and containers, and out of your way. That’s what Azure Linux and Azure Container Linux are designed to do.
Both are hardened, with a reduced package footprint, transparent supply chain, and consistent performance characteristics from the host all the way up to the container. Teams running regulated or security-sensitive workloads get a smaller attack surface and a Linux distribution maintained by the same team that operates the cloud it runs on. And because we develop in the open and contribute upstream first, the work that hardens Azure Linux benefits the broader ecosystem too.
AI isn’t just a new workload sitting on top of open source; it’s changing how open source itself gets built.
That’s a good thing for the ecosystem, but it raises the bar on the fundamentals: provenance, review discipline, supply chain integrity, and clear standards. The communities that figure out how to fold AI into their workflows while keeping the trust model intact are the ones who will define the next decade of open source.
Delivering agentic systems at global scale takes collaboration across the open source ecosystem. Agents need to work everywhere developers build—across frameworks, clouds, languages, and runtimes. That kind of portability only happens when the foundations are open.
That’s why we are working alongside the open source community on the building blocks of an open agentic stack:
Those building blocks need a shared standards body to keep them interoperable. That is where the Agentic AI Foundation comes in.
The Agentic AI Foundation (AAIF) is already the fastest-growing project in Linux Foundation history. Microsoft is a founding member, and we believe deeply in its mission: establishing open standards for agent-to-agent communication, agent runtimes, and agent orchestration.
The AAIF builds on and complements what the Cloud Native Computing Foundation (CNCF) has done for cloud native—the two are designed to work together. The reason this is happening so quickly is straightforward: customers and the broader community are asking for interoperability. They don’t want to bet their agentic future on a single vendor’s stack and open standards are how we make sure they do not have to.
The early momentum across industry and academia tells you how much the ecosystem wants this to be open. As customers scale multi-agent systems composed of custom built and third party agents, interoperability becomes essential to truly deliver on the business transformation goals. The agentic future cannot be proprietary, and the AAIF is how we make sure it isn’t.
None of this works if the underlying ecosystem isn’t trustworthy. The same projects that power the cloud and AI also power critical infrastructure, and the people maintaining them are often a handful of volunteers in their spare time. As agents become more autonomous, every dependency they touch becomes part of their trust boundary. Securing open source isn’t just hygiene anymore. It’s a prerequisite for letting AI agents do real work.
That’s why Microsoft has made a sustained, multi-phase investment in OpenSSF and Alpha-Omega:
We’re also a founding partner in the GitHub Secure Open Source Fund, which pairs direct financial support ($10,000 per project) with a three-week program of security education, mentorship from GitHub Security Lab, tooling, and ongoing check-ins. The model is designed to scale; invest in maintainers as people, not just packages, and the security improvements compound across the dependency graph.
For three years running, Microsoft Azure has been the largest public cloud contributor and the second-largest contributor overall to CNCF projects.
That includes core upstream work in Kubernetes, Helm, containerd, Istio, and Envoy—the runtime layer the entire cloud-native ecosystem depends on, alongside community projects we contribute to heavily, like OpenTelemetry, ArgoCD, HolmesGPT, OPA Gatekeeper, and Cilium as well as projects we’ve launched and donated:
Every one of these projects started with a problem we hit running Kubernetes on Azure at scale. When we do work in the open, we get better solutions, and the broader community strengthens and benefits from the work too.
The takeaway from a decade of cloud native is that the principles still apply:
Kubernetes and Linux fueled the cloud era as the foundational layers. We believe they will be foundational for the agentic era too, alongside the new open standards the community is building right now.
If you’re at the summit this week, please come say hello. The Microsoft team is at the booth with live demos across:
We have engineers, maintainers, and product managers ready to dig into whatever’s on your mind—whether that’s a thorny Kubernetes question, an idea for a new CNCF sandbox project, an AAIF contribution, or how to get your first agent into production.
The cloud era was built by this community. The AI native era will be too. I can’t wait to see what we build together.
See you in Minneapolis.
—Brendan
Sign up to express your interest in the upcoming public preview and be among the first to get access.
Conductor is an open-source CLI (MIT license, Microsoft org) that takes a different approach: you define your multi-agent workflows in YAML, and the routing between agents is deterministic. Jinja2 templates and expression evaluation handle conditions and branching. The orchestration layer consumes zero tokens. The structure is fixed at definition time—and that's the point.
How Microsoft uses AI agents and Drasi to keep open‑source documentation accurate and working.
Microsoft unveils the Agent Governance Toolkit to bring OS‑level security, trust, and compliance to autonomous AI agents.
At KubeCon + CloudNativeCon Europe 2026 in Amsterdam, we're making announcements that reflect the goal of bringing the operational maturity of Kubernetes to today's workloads and demands.
