The cloud-native ecosystem gathered in London from April 1-4, 2025, for KubeCon + CloudNativeCon Europe, bringing together over 12,000 attendees from across the globe. This year's event marked a significant milestone—the 10th anniversary of cloud native computing—and showcased how the ecosystem has matured while continuing to evolve at a rapid pace.
As DevOps leaders and platform engineers navigate an increasingly complex landscape, several key themes emerged that will shape Kubernetes and cloud-native technologies in the coming year. From the rise of platform engineering to AI workloads on Kubernetes, here's what you need to know from this landmark event.
Platform Engineering and Internal Developer Platforms Take Center Stage
Platform engineering has firmly established itself as more than just a buzzword, evolving into a critical discipline for organizations seeking to balance developer autonomy with operational excellence. At KubeCon Europe 2025, several trends in platform engineering became apparent:
The Maturation of Internal Developer Platforms (IDPs)
Spotify's journey with Backstage, now the de facto standard for building internal developer platforms, demonstrated the power of open-source collaboration. As one Spotify engineer noted during a panel discussion, "If we hadn't open-sourced Backstage, we'd likely still be living with our previous, more complicated backend system. In five years, we've gone from basically just a thin framework for building an IDP, to a better Backstage than we could have built on our own."
This sentiment was echoed across multiple sessions, with organizations reporting significant improvements in developer productivity after implementing IDPs. The focus has shifted from simply providing self-service infrastructure to creating comprehensive developer experiences that abstract away complexity while maintaining necessary guardrails.
Platform Teams as Product Teams
A recurring theme was the importance of treating platform teams as product teams, with users (developers) at the center of their design thinking. Hans Kristian Flaatten and Audun Fauchald Strand from NAV shared their "Adventures of building platform as a service for the government," highlighting how their public PaaS now serves 81 Norwegian government organizations by focusing on developer needs first.
The most successful platform teams are now measuring their success not just by uptime or cost savings, but by developer satisfaction metrics and reduced cognitive load. As one speaker put it, "Your platform isn't successful if developers aren't choosing to use it."
Golden Paths and Standardization
Organizations are increasingly implementing "golden paths"—opinionated, wellsupported workflows that make it easy for developers to follow best practices. These paths aren't mandated but are designed to be the path of least resistance, allowing for exceptions when necessary.
Michelin's container-as-a-service team shared how their platform transformation led to a 44% reduction in platform costs and an 85% reduction in upgrade lead time while doubling their Kubernetes footprint. Their success came from standardizing on Kubernetes while providing clear, well-documented paths for different application patterns.
AI/ML Workloads on Kubernetes: From Experimentation to Production
The integration of AI and ML workloads with Kubernetes has accelerated dramatically, with several key developments highlighted at KubeCon Europe 2025:
Kubernetes Evolving for AI Workloads
Jago Macleod, engineering director of GKE and Kubernetes at Google, outlined three key goals to evolve Kubernetes for "the next trillion core hours": 1. Improve reliability at scale, across upgrades 2. Redefine Kubernetes' relationship with hardware 3. Move from container and workloads to framework orchestration
Google's vision of making Kubernetes "GenAI aware" resonated with attendees,
particularly as organizations struggle with the unique demands of AI workloads. The strategy of maintaining, extending, and expanding Kubernetes specifically for AI/ML workloads demonstrates the platform's continued relevance in the AI era.
Multi-Cloud AI Infrastructure
Mirantis unveiled k0rdent, an open-source, declarative, and composable Kubernetes-native Distributed Container Management Environment (DCME) designed specifically for AI/ML applications. With 19+ validated integrations in its catalog, k0rdent aims to simplify multi-cloud, multi-cluster operations with a particular focus on accelerating AI application buildout, especially for inference workloads.
This solution addresses a growing challenge: how to efficiently manage AI workloads across heterogeneous infrastructure spanning cloud, datacenter, and edge environments.
Convergence of AI/ML and HPC
Several sessions highlighted the growing convergence between AI/ML and High
Performance Computing (HPC) requirements. The "Slinky: Slurm in Kubernetes"
presentation by Marlow Warnicke and Tim Wickberg from SchedMD demonstrated how traditional HPC workload managers are being integrated with Kubernetes to provide performant AI and HPC workload management.
This convergence is driving innovations in how scarce and expensive hardware
accelerators are allocated and utilized, with Dynamic Resource Allocation (DRA)
emerging as a critical capability for AI workloads on Kubernetes.
Evolving Observability Practices and Tooling
Observability continues to evolve rapidly, with several key trends emerging at KubeCon Europe 2025:
Observability in the Age of LLMs
Christine Yen, CEO and cofounder of Honeycomb, delivered a compelling keynote on "Observability in the age of LLMs." She explained that while LLMs are transforming software development, they also complicate traditional testing, mocking, and debugging approaches. The solution, according to Yen, is enhanced observability that embraces the unpredictability LLMs introduce.
Her recommendation to bundle observability with evals to capture both successes and failures as they happen resonated with attendees working on AI-powered systems. Importantly, she emphasized that this shift builds on existing observability skillsets rather than requiring entirely new approaches.
AI-Enabled Observability
eBay's Vijay Samuel shared how his team is handling 4600 microservices and processing 15PB of logs daily by using LLMs as observability building blocks. Their approach of starting with small experiments to automate simple flows demonstrated a practical path to AI-enhanced observability that balances LLM capabilities with traditional code approaches.
This pragmatic approach—using AI to enhance rather than replace existing observability practices—was a common theme across sessions.
Platform Engineering for Observability
Kasper Borg Nissen from Dash0 made a compelling case for applying platform
engineering principles to observability in his talk, "The observability platform
engineering advantage: from zero code to monitoring as code."
"We don't have a metrics problem, or a tracing problem. We have systems problems," Nissen argued. "And yet, many of us still treat these as separate entities. We have one browser tab for logs, another for metrics, and a third for tracing. We are relying on humans to correlate signals across them. It's inefficient, it's error-prone, and frankly, it's not how modern observability should work."
Open Source Dominance in Observability
Grafana Labs' 2025 Observability Survey, presented at KubeCon, revealed that 75% of respondents are now using open source licensing for observability, with 70% reporting that their organizations use both Prometheus and OpenTelemetry. Half of all organizations increased their investments in both technologies for the second year in a row.
The survey also found that complexity is the number one observability concern, while alert fatigue is cited as the biggest obstacle to faster incident response. This explains why training-based alerts and faster root cause analysis topped respondents' AI/ML wishlist for observability.
Cloud-Native Security, Compliance, and Supply Chain Integrity
Security remained a top concern at KubeCon Europe 2025, with supply chain security receiving particular attention:
Supply Chain Security Innovations
Kusari, a software supply chain security startup, introduced a new book, "Securing the Software Supply Chain," and showcased next-generation approaches centered around GUAC (Graph for Understanding Artifact Composition), OpenSSF's new Open Source Project Security Baseline, and CNCF's Security Technical Advisory Group Software Supply Chain Best Practices v2 white paper.
Michael Lieberman from Kusari delivered a keynote on "Cutting Through the Fog:
Clarifying CRA Compliance in Cloud Native," addressing the growing importance of compliance with the EU's Cyber Resilience Act for cloud-native organizations.
Interactive Security Experiences
The event featured several hands-on security experiences, including Security Slam
sessions focused on projects like Flux and Meshery, and a Capture The Flag security experience designed to help participants "delve deeper into the dark and mysterious world of cloud native security."
These interactive sessions reflect a growing recognition that security must be practiced, not just discussed, to be effective.
Security Tooling Advancements
Microsoft highlighted several security-focused contributions to the cloud-native
ecosystem, including Istio's ambient mode (now generally available), Hyperlight (a Rust library for executing small, embedded functions using hypervisor-based protection), and Ratify (a verification engine for artifact security metadata).
These tools address the growing complexity of securing modern cloud-native
environments, particularly as organizations adopt multi-cloud strategies and deploy increasingly sophisticated workloads.
Open-Source Community Momentum and Project Updates
The 10-year anniversary of cloud-native computing provided an opportunity to reflect on the community's growth and look toward the future:
CNCF at 10
A special keynote session, "CNCF at 10: Navigating challenges, embracing
opportunities," featured Joseph Sandoval (Adobe), Liz Rice (Isovalent at Cisco), and Katie Gamanji (Apple) reflecting on the evolution of the ecosystem over the past decade.
The session highlighted how the CNCF has grown from a single project (Kubernetes) to a diverse ecosystem of graduated, incubating, and sandbox projects addressing every aspect of cloud-native computing.
Enterprise Adoption Stories
Several large enterprises shared their cloud-native journeys:
- HSBC reported servicing around 600 million discrete hits a day with over 7,000 services in production, all running on only about a dozen clusters. Their approach to stability during upgrades involves running clusters in blue/green pairs and rehydrating from etcd backups.
- Michelin shared how switching to Kubernetes and embracing open source brought platform costs down 44% and upgrade lead time down 85% while doubling their Kubernetes footprint.
- Apple's Katie Gamanji discussed how private cloud compute helps scale AI operations while maintaining privacy, emphasizing the role of Swift libraries working with the open source community.
Project Engagement Opportunities
The event featured numerous ways for attendees to engage with projects, including
Project Lightning Talks (5-minute presentations highlighting recent updates), the
Maintainer Track (deep dive sessions with project leaders), Contribfest (collaborative sessions for contributors), and the Project Pavilion (a central hub for connecting with maintainers).
These engagement opportunities reflect the CNCF's continued commitment to fostering a vibrant, participatory community around its projects.
Looking Ahead: What's Next for the Kubernetes Ecosystem
As KubeCon + CloudNativeCon Europe 2025 concludes, several trends are poised to shape the Kubernetes ecosystem in the coming year:
- AI-Native Infrastructure
The integration of AI workloads with Kubernetes will continue to deepen, with newpatterns, tools, and best practices emerging to address the unique requirements oftraining and serving AI models. Expect to see more specialized operators, enhanced resource management capabilities, and tighter integration between AI frameworks and Kubernetes. - Platform Engineering Standardization
As platform engineering matures, expect to see more standardization around best practices, metrics, and tooling. The Platform Engineering Working Group's efforts to define the discipline will likely accelerate, providing clearer guidance for organizations building internal platforms. - Observability Consolidation
The observability landscape will likely see consolidation as organizations seek to reduce tool sprawl and integrate signals across metrics, logs, and traces. OpenTelemetry's continued growth suggests it will become the standard instrumentation layer for cloudnative applications. - Supply Chain Security as a Core Discipline
Supply chain security will continue to evolve from a specialized concern to a core discipline for all cloud-native practitioners. Expect to see more automated tools, standardized practices, and regulatory requirements driving adoption of secure supply chain practices. - Kubernetes at the Edge
The extension of Kubernetes to edge environments will accelerate, driven by the need for consistent management across cloud, data center, and edge locations. Projects like KubeEdge and K3s will continue to evolve to address the unique constraints of edge computing.
Conclusion
KubeCon + CloudNativeCon Europe 2025 demonstrated that while Kubernetes has
reached a level of maturity and stability, the ecosystem around it continues to evolve rapidly. The event highlighted how organizations are moving beyond basic
implementation questions to focus on developer experience, operational excellence, security, and the integration of new workload types like AI/ML.
For DevOps leaders, platform engineers, and cloud-native architects, the key takeaway is clear: success in this ecosystem requires continuous learning, adaptation, and a willingness to embrace both the stability of core technologies and the innovation happening at the edges. By focusing on the human aspects of technology adoption—developer experience, team collaboration, and organizational change—as much as the technical aspects, organizations can fully realize the promise of cloud-native computing.
As we look ahead to the next decade of cloud-native innovation, the community's
commitment to open source, collaboration, and solving real-world problems ensures that Kubernetes and its ecosystem will continue to be at the forefront of modern infrastructure and application development.
