Running Databases on Kubernetes Is Becoming the New Normal

Running databases on Kubernetes has moved from experimentation to standard practice, driven by platform maturity, cost pressures, and AI/ML demands. According to the 2025 Data on Kubernetes survey, organizations are now focused on operational excellence, with cost optimization, storage performance, and AI workloads shaping the next phase.

For years, the conventional wisdom was simple: run applications on Kubernetes, but leave databases to managed cloud services. That assumption is now cracking.

Industry reports and technical analyses show a clear trend: more organizations are running database workloads directly on Kubernetes, driven by platform maturity, cost control, and architectural consolidation.

Kubernetes has evolved beyond stateless orchestration. Improvements in StatefulSets, persistent volumes, operators, and cloud-native storage have made it increasingly viable to run production databases inside clusters. According to Solutions Review, databases have become one of the most common production workloads on Kubernetes, especially in data-intensive and AI-driven environments.

This shift is also reflected in broader platform engineering discussions. The New Stack notes that Kubernetes is finally addressing its long-standing weakness around persistent state, making database operations more predictable and automatable using operators and declarative workflows.

Cost pressure is another major driver. Organizations are reassessing the long-term economics of managed cloud databases, which often carry opaque pricing and scaling penalties. CIO.com reports that cloud repatriation is back on the agenda, particularly for data-heavy workloads where cost predictability and control matter more than convenience.

Rather than abandoning the cloud, many teams are converging on a hybrid model: Kubernetes as the universal control plane, hosting both applications and databases, across cloud and on-prem environments. This allows tighter integration, reduced vendor lock-in, and a single operational model for stateful and stateless systems alike.

The trend is reinforced by data from the 2025 Data on Kubernetes Community (DoKC) survey, which reports that running data workloads on Kubernetes is now a standard practice rather than an emerging one. The survey shows that organizations have largely moved past initial adoption and are now focused on operational optimization, particularly around cost management, performance, and AI/ML readiness.

Cost optimization emerged as the top priority for 2025, overtaking security and scaling initiatives. This focus is especially pronounced among organizations running AI and machine learning workloads, where storage costs have become a primary concern due to the size of training datasets, model checkpoints, and inference artifacts. The survey also highlights the growing role of vector databases, which a large majority of respondents now consider critical infrastructure for AI workloads.

The data further points to a shift in architecture. A majority of respondents identified edge computing and real-time data processing as essential to their future data strategy, reflecting growing demand for low-latency processing and compliance with data sovereignty requirements. These needs are pushing organizations toward more distributed Kubernetes-based architectures rather than centralized, batch-oriented systems.

At the same time, the survey identifies persistent challenges. Performance optimization, security and compliance, and skills shortages remain the most commonly cited obstacles. The skills gap is particularly notable, with organizations reporting difficulty finding practitioners who combine Kubernetes operational expertise with deep knowledge of data and storage optimization.

Overall, the DoKC findings suggest that Kubernetes has become a foundational platform for data workloads. The focus has shifted from whether databases can run on Kubernetes to how organizations can operate them efficiently, securely, and at scale in AI-driven and distributed environments.

In short, the question is no longer “Should databases run on Kubernetes?” but “Which databases, and where does Kubernetes make the most sense?”

• https://dok.community/data-on-kubernetes-report-20...