What's New In Kubernetes 2026
Kubernetes continues to evolve as organizations adopt cloud-native architectures, containers, AI workloads, and distributed applications. In 2026, Kubernetes introduced several improvements focused on resource management, workload scheduling, security, observability, storage, and developer experience.
The latest stable release, Kubernetes 1.37, was released on August 26, 2026, and includes 67 enhancements: 16 Stable, 23 Beta, 27 Alpha, and one deprecation/removal.
1. Kubernetes 1.37 Is the Latest Release
Kubernetes 1.37, named Garhwal, is the latest stable Kubernetes release as of September 2026.
The release continues Kubernetes' gradual approach of moving important capabilities from Alpha to Beta and Stable stages, allowing organizations to adopt mature features while continuing to experiment with newer capabilities.
2. KYAML Reaches Stable
KYAML, a safer and less ambiguous subset of YAML designed specifically for Kubernetes, has reached Stable status in Kubernetes 1.37.
Existing YAML remains compatible, so teams do not need to rewrite their existing manifests. Kubernetes also supports kubectl get -o kyaml as a Stable capability.
This can help developers work with Kubernetes configuration files more consistently and reduce some of the ambiguity associated with traditional YAML syntax.
3. Metrics API Becomes Stable
The metrics.k8s.io API has graduated to Stable after spending several years in Beta.
The API provides standardized CPU and memory usage information for pods and nodes and supports commonly used functionality such as:
kubectl top
Horizontal Pod Autoscaler
Resource monitoring
Capacity analysis
This gives Kubernetes users a more mature API foundation for basic resource usage monitoring.
4. Better Resource Management for AI and Specialized Workloads
Modern AI and machine-learning applications often require specialized hardware and complex resource allocation.
Kubernetes 1.37 continues to expand Dynamic Resource Allocation (DRA) capabilities. Improvements include better resource accounting and additional support for devices and specialized workloads.
These developments are particularly relevant for organizations running workloads that require GPUs, accelerators, or other specialized resources.
5. Workload-Aware Scheduling
Kubernetes is continuing to improve its scheduling capabilities for complex workloads.
The new Alpha CompositePodGroup API allows workloads to be represented as hierarchical groups rather than simply as a flat collection of Pods.
This can support advanced scheduling scenarios such as:
Multi-level gang scheduling
Workload-aware preemption
Topology-aware scheduling
Complex AI/ML workloads
These capabilities are particularly relevant as Kubernetes is increasingly used for sophisticated machine-learning and batch-processing workloads.
6. StatefulSet Recreate Strategy
Kubernetes 1.37 introduces a new Alpha Recreate strategy for StatefulSet rollouts.
Previously, StatefulSets primarily used OnDelete and RollingUpdate strategies. The new approach can delete the existing StatefulSet Pods before creating new Pods based on the updated specification.
This provides another deployment strategy for applications where a rolling update may not be appropriate.
7. Improved Node Lifecycle Management
Kubernetes 1.37 introduces Node Declared Features as a Stable capability.
Nodes can declare supported features so control-plane components can make better decisions when clusters contain nodes with different capabilities.
Kubernetes 1.37 also introduces standardized node lifecycle conditions for situations such as:
Drain in progress
Node drained
Planned maintenance
Maintenance in progress
Graceful node shutdown
These improvements can provide more consistent information about node state.
8. Improved Container and Storage Security
Kubernetes 1.37 includes storage-related security improvements, including enhancements around SELinux mounting and storage permissions.
The Kubernetes project also highlighted new storage security capabilities for emptyDir permission modes and bind mount options in September 2026.
These improvements can help administrators implement more restrictive storage policies for containerized applications.
9. Memory QoS Improvements
Memory Quality of Service (QoS) has graduated to Beta in Kubernetes 1.37 and is enabled by default.
On Linux nodes using cgroup v2, Memory QoS uses the Linux memory controller to provide better guidance for managing container memory.
Better memory management can be important for applications running many workloads on shared Kubernetes nodes.
10. Native Histograms for Better Observability
Native histogram support for Kubernetes metrics has graduated to Beta in Kubernetes 1.37 and is enabled by default.
Histograms can provide more detailed information about distributions such as request latency, rather than relying only on averages or individual measurements.
This can improve observability for applications where understanding performance distributions is important.
11. Rootless Kubernetes Improvements
The KubeletInUserNamespace, also known as rootless mode, has graduated to Beta in Kubernetes 1.37.
With this capability, Kubernetes node components can run as a non-root user on the host when the required configuration and feature support are available.
Reducing the need for root privileges can contribute to stronger security boundaries for Kubernetes infrastructure.
12. Kubernetes Continues Moving Away from kube-dns
Kubernetes continues to move toward CoreDNS for cluster DNS.
The Kubernetes project has deprecated kube-dns, noting that it has not kept pace with capabilities such as EndpointSlices and dual-stack Services. Kubernetes recommends that clusters still using kube-dns plan migration to CoreDNS.
Organizations maintaining older Kubernetes environments should therefore review their DNS configuration as part of modernization planning.
What These Kubernetes Updates Mean for Businesses
The 2026 Kubernetes improvements reflect several broader trends in cloud-native development.
Better AI Infrastructure
Improved resource allocation and workload-aware scheduling help Kubernetes support increasingly complex AI and machine-learning workloads.
Stronger Security
Updates around SELinux, rootless operation, storage permissions, and node management provide additional tools for improving container security.
Better Observability
Stable metrics APIs and improved histogram support can help engineering teams understand resource consumption and application performance.
More Flexible Workload Management
Scheduling and StatefulSet improvements provide additional options for managing different workload types and deployment requirements.
Improved Developer Experience
Features such as KYAML and more mature APIs can make Kubernetes configuration and administration easier to manage.
Should Businesses Upgrade to Kubernetes 1.37?
Organizations should evaluate upgrades based on their current Kubernetes version, application compatibility, infrastructure, dependencies, and operational requirements.
Kubernetes currently maintains the most recent three minor releases. As of September 2026, these are versions 1.37, 1.36, and 1.35. Kubernetes 1.37 is supported until October 28, 2027.
Before upgrading, teams should:
Review the Kubernetes release notes.
Check deprecated APIs and features.
Test workloads in a non-production environment.
Verify third-party integrations.
Review storage and networking components.
Validate monitoring and security tools.
Prepare a rollback or recovery strategy.
Upgrade production clusters in a controlled manner.
Conclusion
Kubernetes in 2026 continues to expand beyond basic container orchestration. The latest 1.37 release introduces improvements across resource management, AI/ML scheduling, observability, security, storage, node management, and developer experience.
For organizations running Kubernetes in production, keeping clusters on supported versions and regularly reviewing new capabilities can help maintain security, reliability, and compatibility as cloud-native environments evolve.
