Cloud Native Development 2025: CNCF Report Key Trends

Cloud Native Development Landscape in 2025

The cloud native ecosystem has reached a critical inflection point. The CNCF State of Cloud Native Development Q1 2025 report, produced in partnership with SlashData, provides the most comprehensive snapshot of how developers are building, deploying, and scaling modern software infrastructure. Drawing on data from the 29th edition of SlashData’s Developer Nation survey — fielded between December 2024 and February 2025 and reaching more than 10,500 developers across 126 countries — the report reveals both remarkable stability and transformative shifts in cloud native adoption.

Cloud native development, defined by the combined use of containers, container orchestration tools, service meshes, Kubernetes, and cloud functions or serverless computing, has become the de facto standard for modern backend services. The findings confirm that the ecosystem is not just growing — it is maturing, with clear patterns emerging around AI/ML integration, deployment architecture preferences, and the evolving role of orchestration platforms.

For technology leaders, platform engineers, and development teams evaluating their infrastructure strategies, understanding these trends is essential. This analysis breaks down every major finding from the CNCF report, offering actionable insights on where cloud native development is heading and what it means for organizations investing in digital transformation and cloud infrastructure modernization.

9.2 Million Cloud Native Developers Worldwide

The headline figure from the CNCF report is striking: approximately 9.2 million cloud native backend developers are active worldwide as of Q1 2025, representing 49% of all backend service developers. This figure comes with a 95% confidence interval of 8.8 million to 9.6 million, underscoring the statistical rigor behind the estimate.

To put this growth in perspective, cloud native backend developers numbered 6.9 million in Q1 2021, meaning the ecosystem has added 2.3 million developers in just four years. The proportion of backend developers adopting cloud native technologies has risen steadily from 46% in Q1 2021 through a peak of 47% in Q3 2023, reaching the current high of 49%.

What makes this growth particularly significant is its consistency. Unlike many technology trends that experience rapid spikes followed by plateaus, cloud native adoption has maintained a steady upward trajectory. Each quarterly survey since Q1 2021 has shown either stability or incremental improvement, suggesting that cloud native development is not a hype cycle but a structural shift in how software is built.

The machine learning and data science (MLDS) segment tells an even more dynamic story. Cloud native MLDS developers reached 4.2 million by Q3 2024, with the proportion of MLDS developers using cloud native technologies jumping from 19% in Q1 2024 to 24% in Q3 2024 — a five-percentage-point increase in just two quarters. This acceleration signals that AI/ML practitioners are increasingly adopting cloud native infrastructure patterns that were originally developed for traditional backend services.

Cloud Deployment Trends: Hybrid and Multicloud Growth

One of the most revealing sections of the CNCF report examines how developers are deploying their applications across different cloud environments. The data shows a clear trend: developers are diversifying their deployment strategies, moving away from single-cloud approaches toward more distributed architectures.

Among all developers surveyed, 93% now deploy to cloud for some part of their development process, up significantly from 85% in Q1 2021 and 87% in Q1 2023. For backend service developers specifically, the figure is even higher at 96%, up from 86% in Q1 2023. Cloud has effectively become universal infrastructure.

The most notable shift is in hybrid cloud adoption. Among all developers, hybrid cloud usage grew from 22% in Q1 2021 to 29% in Q1 2025 — the largest increase of any deployment type. Multicloud also saw strong growth, rising from 17% to 22% over the same period. These trends align with research from organizations like Gartner, which has consistently predicted that hybrid and multicloud strategies would become the dominant deployment paradigm.

Interestingly, public cloud usage among all developers actually decreased slightly from 38% to 36%, while private cloud grew from 35% to 37%. Among backend developers specifically, both public and private cloud usage declined to 33% each — their lowest levels since Q1 2021. This does not mean cloud is declining; rather, developers are increasingly choosing hybrid and multicloud configurations that combine multiple deployment models to optimize for cost, performance, compliance, and resilience.

A new category in the survey — distributed cloud (geographically distributed compute managed centrally) — captured 12% of backend developers in its first measurement. This emerging architecture reflects the growing importance of edge computing and low-latency requirements in modern applications.

Container Adoption and Orchestration Shifts

Container technology remains the foundation of cloud native development, and the CNCF report reveals a fascinating pattern of stability. Container usage among backend service developers has held remarkably steady at approximately 61% for over four and a half years, from Q3 2020 through Q1 2025. This plateau suggests that containers have reached market saturation among their target audience — developers who need them are already using them.

Kubernetes, the leading container orchestration platform maintained by the Cloud Native Computing Foundation, showed a small but consistent positive trend, reaching 31% adoption among backend developers in Q1 2025, up from 28% in Q3 2020. While this growth appears modest in percentage terms, it translates to millions of additional developers incorporating Kubernetes into their workflows.

Container orchestration tools and management platforms, however, tell a different story. Usage decreased from 29% in Q3 2023 to 22% in Q1 2025 — a notable seven-percentage-point decline. The report suggests this may be driven by cloud providers increasingly abstracting orchestration complexity into managed services. As platforms like Amazon EKS, Google GKE, and Azure AKS mature, developers can leverage orchestration capabilities without directly managing orchestration tools.

Service mesh adoption has remained remarkably stable at 14% since Q3 2020, suggesting that while service meshes provide value for specific use cases (traffic management, security, observability in microservices architectures), they have not achieved broad adoption beyond their core user base. The complexity of service mesh implementations may be a barrier, and the emergence of simpler alternatives and platform-integrated solutions may be satisfying some of the same needs.

Kubernetes Adoption for AI and ML Workloads

Perhaps the most forward-looking section of the CNCF report examines the intersection of Kubernetes and artificial intelligence. Among professional developers working in backend services, ML/AI, or data science (n=4,456), 36% are already running ML/AI workloads on Kubernetes, with an additional 18% planning to adopt this approach. Combined, more than half of this professional developer population is either using or intending to use Kubernetes for AI/ML.

The adoption rates vary significantly by development specialization. Among dedicated ML/AI developers, 52% are using Kubernetes for ML/AI workloads — a majority. Data scientists follow closely at 50%. DevOps professionals sit at 38%, while general backend and web backend developers trail at 28% and 27% respectively, though notably 26-27% of these groups plan to adopt Kubernetes for ML/AI.

The specific ML/AI workloads running on Kubernetes reveal important patterns about how organizations are operationalizing AI. Data pre-processing leads at 11% of the broader developer population, followed by model experimentation, real-time model inference, and training large-scale models at 9% each. Model monitoring and drift detection, batch processing for model inferencing, automated retraining, and batch jobs for ML/AI pipelines all register at 8%.

Kubernetes Adoption by ML/AI Activity Type

The report’s most granular analysis examines Kubernetes usage across different ML/AI activities. Developers working on model infrastructure and pipeline engineering show the highest adoption at 71%, followed by those developing ML frameworks and libraries at 70%, and ML/AI educators at 69%. These figures underscore that Kubernetes has become the default platform for production-grade ML infrastructure.

In contrast, developers using third-party models through APIs show the lowest Kubernetes adoption at 52% — still a majority, but reflecting that API-based model consumption requires less infrastructure management. Developers focused on model training (53%) and data analysis (54%) also show lower adoption rates, though the majority still leverage Kubernetes in some capacity.

Serverless Computing Decline and Market Dynamics

One of the most surprising findings from the CNCF report is the continued decline in serverless computing adoption. Cloud functions and serverless architecture usage among backend developers dropped from 26% in Q3 2020 to 20% in Q1 2025 — a six-percentage-point decrease that challenges the narrative of serverless as the inevitable future of cloud computing.

The report attributes this decline partly to the shift toward AI/ML workloads. Serverless computing excels for event-driven, short-duration tasks with unpredictable traffic patterns. However, ML model inference — especially real-time inference — typically involves sustained, high-volume compute that makes serverless pricing models less economical than reserved or on-demand instances. As more developers integrate AI into their applications, the economics of serverless become less favorable for a growing share of their workloads.

This does not mean serverless is disappearing. A 20% adoption rate among backend developers still represents millions of practitioners. Rather, serverless appears to be finding its equilibrium as a specialized tool for specific use cases rather than a general-purpose compute paradigm. The technology’s sweet spot remains in API backends, event processing, scheduled tasks, and other workloads where the pay-per-execution model aligns with usage patterns.

For organizations evaluating their infrastructure strategy, the serverless decline underscores the importance of matching technology choices to workload characteristics. The AWS Well-Architected Framework and similar resources from major cloud providers increasingly emphasize this workload-driven approach to architecture decisions.

DevOps and Cloud Native Infrastructure Patterns

The CNCF report’s findings on DevOps practitioners provide valuable insights into how cloud native infrastructure is being managed and operated at scale. DevOps professionals show a distinctive pattern of Kubernetes adoption for AI/ML workloads: 38% are already using Kubernetes for these workloads, with 22% planning to adopt — the highest planning rate among all development specializations.

What sets DevOps professionals apart is their emphasis on operational aspects of ML/AI. Among DevOps developers using Kubernetes for AI/ML, model experimentation leads at 32% — higher than any other specialization. Real-time model inference follows at 28%. This pattern suggests that DevOps teams are not just operating ML infrastructure but are actively involved in model lifecycle management, from experimentation through production deployment.

The on-premise server trend also merits attention. Among all developers, on-premise deployment actually increased from 39% to 41% between Q1 2021 and Q1 2025. Among backend developers, the trend reversed, declining from 46% to 40%. This divergence may reflect different dynamics: cloud-first backend development is well-established, but edge computing, data sovereignty requirements, and AI training on proprietary hardware are driving broader on-premise investment.

Mainframe usage showed surprising growth from 9% to 16% among all developers and from 10% to 11% among backend developers. While modernization is the primary driver, this growth also reflects the continued importance of mainframe systems in financial services, government, and healthcare — sectors that are increasingly integrating cloud native approaches with legacy infrastructure through hybrid strategies.

Cloud Native Machine Learning Pipeline Engineering

The CNCF report provides unprecedented detail on how different ML/AI activities correlate with Kubernetes adoption and specific workload patterns. This data is invaluable for organizations designing their ML platform strategy and deciding where to invest in Kubernetes-based infrastructure.

Developers working on model infrastructure and pipeline engineering — the practitioners who build the platforms that other ML teams use — demonstrate the highest Kubernetes adoption across virtually every workload type. Their real-time model inference deployment on Kubernetes reaches 24%, automated retraining at 20%, batch jobs for ML/AI pipelines at 20%, and model experimentation at 20%. These figures are consistently above average, confirming that platform engineering teams see Kubernetes as the standard substrate for production ML.

ML framework developers show similar patterns, with particularly high adoption for data pre-processing (25%), real-time inference (25%), and training large-scale models (21%). These developers are building the tools that the broader ML community relies on, and their infrastructure choices often set the standard for the ecosystem as documented in resources from The Linux Foundation.

An interesting finding involves ML/AI educators and teachers. This group shows the third-highest overall Kubernetes adoption (69%) and notably high usage for data labeling and augmentation (20%), model monitoring and drift detection (20%), and batch jobs for ML/AI pipelines (21%). This suggests that educational institutions and training programs are adopting production-grade Kubernetes workflows, preparing the next generation of practitioners with industry-standard tools.

The gap between model infrastructure engineers (71% Kubernetes adoption) and developers consuming third-party models via APIs (52% adoption) highlights a strategic divide. As AI becomes increasingly accessible through API services, many developers may not need direct Kubernetes expertise. However, organizations building proprietary models or requiring fine-grained control over their ML pipeline will continue to invest heavily in Kubernetes-based infrastructure, and the data from this analysis of AI technology adoption trends supports this trajectory.

Future of Cloud Native Development: What Comes Next

The CNCF State of Cloud Native Development Q1 2025 report paints a picture of an ecosystem that has transitioned from rapid growth to strategic maturation. Several clear trends will shape the next phase of cloud native development.

First, the convergence of cloud native and AI/ML infrastructure will accelerate. With 18% of professional developers planning to adopt Kubernetes for ML/AI workloads — representing millions of practitioners — the next two years will see significant investment in making Kubernetes more ML-friendly. Projects like Kubeflow, KServe, and the emerging GPU scheduling capabilities in Kubernetes will become increasingly critical.

Second, hybrid and multicloud will continue to gain ground as organizations balance cost optimization, data sovereignty, and vendor diversification. The emergence of distributed cloud as a measurable category (12% among backend developers in its first measurement) signals growing demand for geographically distributed compute that combines cloud economics with edge performance.

Third, the abstraction layer above Kubernetes will become the new battleground. The decline in container orchestration tools (from 29% to 22%) alongside stable Kubernetes adoption suggests that the platform layer is consolidating. Developers increasingly expect Kubernetes capabilities without Kubernetes complexity, and managed services that deliver this abstraction will capture market share.

Fourth, serverless will find its niche. Rather than becoming the dominant compute model, serverless at 20% adoption appears to be settling into a complementary role alongside containers and orchestrated workloads. Organizations will increasingly adopt polyglot compute strategies, using serverless for event-driven workloads while running sustained AI/ML and backend services on container platforms.

Finally, the 93% cloud deployment rate among all developers means that the remaining debate is not about whether to use cloud, but about how to use it optimally. Cloud native development has won the architecture debate. The next frontier is operational excellence — making cloud native infrastructure more reliable, efficient, secure, and cost-effective at scale.