Cloud native is no longer a bet. It is the operating baseline for modern software and it is increasingly the substrate for AI.
The CNCF Annual Cloud Native Survey report, based on a September 2025 survey of 628 practitioners, maps how teams are building, shipping and governing platforms at scale. The data shows Kubernetes consolidating its role in production, containers spreading across portfolios and GitOps moving from nice-to-have to maturity marker. It also surfaces the friction points that slow progress, especially culture, skills and security, plus the new pressures created by generative AI adoption and open source sustainability. Priorities are shifting fast now.
Below are the ten key takeaways from the report:
1) Cloud native has effectively become the default
Cloud native adoption has reached 98% of organizations, so the conversation is shifting from whether teams will adopt to how they will mature and extract value.
2) Kubernetes is now the standard inside container environments
Among container users, 82% are using Kubernetes in production in 2025, up from 66% in 2023, a strong signal that Kubernetes is no longer a “maybe” platform for most teams.
3) Container usage keeps moving from pockets to production
The share of organizations using containers for most or all production applications rose from 41% in 2023 to 56% in 2025, while container pilots shrank from 11% to 6%.
4) The hardest part of cloud native is people, not primitives
The top container challenge in 2025 is cultural change with the development team (47%), followed by lack of training (36%) and security (36%), showing that maturity bottlenecks increasingly live in operating models and skills, not YAML.
5) Kubernetes is taking on a new job: hosting generative AI
Two-thirds of organizations (66%) are already using Kubernetes to host generative AI workloads, pushing Kubernetes from an orchestration layer to an AI infrastructure platform.
6) Most enterprises are consuming AI, not training it
A majority (52%) do not build or train AI models, so the center of gravity shifts toward inference, cost controls and deployment patterns. The hosting split reinforces that story: 37% use managed APIs, 25% self-host models and 13% are deploying at the edge.
7) AI deployment is cautious and that is a tooling problem as much as a model problem
Deployment frequency data shows 47% deploy generative AI models only occasionally and just 7% deploy daily, reflecting the additional validation and governance steps that production AI demands.
8) GitOps and CI/CD have reached maturity
GitOps is absent among explorers (0%) but shows up in 58% of innovators. CI/CD is already used by 42% of explorers and climbs to 91% among innovators, making it the closest thing to a universal foundation for cloud native progress.
9) Advanced platform capabilities show up only after the basics are in place
As organizations move from explorers to innovators, production use of more demanding patterns rises sharply, including stateful containers (79%), serverless (64%) and service mesh (39%). The same maturity curve shows up in delivery cadence, with innovators reporting far higher automation and release velocity.
10) The “next big things” are still waiting, while open source sustainability is becoming a board-level risk
WebAssembly remains early, with about 65% reporting no experience and only 5% reporting full deployment in 2025. At the same time, the report flags sustainability concerns as usage rises, especially as automated, machine-driven workloads increase pressure on shared infrastructure across a CNCF ecosystem that now spans 234 projects and more than 270,000 contributors.
