Cloud-Native Developers Lead AI Adoption as Open-Source Maintainers Face More Work
- Cargo
- Machine Learning
- Shipped
- Shipper
- Sloane Kali Faye
- Readers
- 348
A report counts 21 million cloud-native developers and higher AI use, while maintainers contend with low-quality contributions and teams face closer token budgets.

Cloud-native developers are adopting AI faster than their peers, but the tools spreading through their workflows also create more work for open-source maintainers. A new report counts more than 21 million cloud-native practitioners worldwide and finds that about three-quarters use AI in some form. The same shift is bringing pressure to review a growing flow of machine-generated contributions and to manage model spending.
The State of Cloud Native Development report, produced by the Cloud Native Computing Foundation (CNCF) and SlashData Ltd., estimates that cloud-native developers make up 41% of developers globally. That headcount may still be incomplete: people who use applications built on cloud-native infrastructure may not identify themselves as cloud-native developers.
Bob Killen, CNCF’s director of technical program management, raised that distinction in a conversation with SlashData principal market research consultant Liam Bollmann-Dodd at KubeCon + CloudNativeCon NA. The discussion focused on the report’s findings, AI use, the load on open-source projects and the growing attention to costs.
Cloud-native developers report higher AI use
About 75% of cloud-native developers use AI in some form, compared with about 55% of other developers, Bollmann-Dodd said. The 20 percentage point difference suggests that familiarity with cloud-native tools and infrastructure may go along with earlier or broader AI adoption. The reported figures don’t establish why the gap exists, or how much use involves coding assistants, automated agents or other AI products.
For organizations, the finding points to a practical connection: teams already building and deploying cloud-native software may be well placed to experiment with AI in their development work. But adoption alone says little about the quality of the output, its reliability in production or the costs of operating it. Those questions become more pressing as AI agents take on work that can generate code and open requests for changes in shared projects.
The report’s figures describe developer use, rather than the performance of any particular model or tool. They also don’t show whether AI use has improved productivity. That leaves a gap between adoption and impact: teams can track who is using AI, but they still need to assess whether it improves review quality, speeds delivery or shifts work onto other people.
Generated contributions add review pressure
For maintainers, more AI-generated output can mean more incoming work even when the software itself is open and available for contributions. Killen said sustainability was a major topic at the Maintainer Summit ahead of KubeCon, with maintainers grappling with a flood of pull requests and issues. Some submissions, he said, are “just good enough” to demand attention without being useful or ready to merge.
That creates an uneven cost. Generating a patch or issue can take little time for the person or agent producing it, while a maintainer still has to understand the change, check it against project requirements and decide what to do with it. A contribution that looks plausible may take substantial effort to validate. If its author hasn’t done that work, the project absorbs the review burden.
The challenge is not simply counting submissions. Projects need to distinguish useful changes from work that is incomplete, inaccurate or detached from the project’s needs. AI assistance may help contributors prepare clearer reports or patches, but the number of submissions alone won’t show whether maintainers are receiving better contributions. The useful measure is whether the added activity helps projects make progress without crowding out the people responsible for review and upkeep.
Token budgets bring cost discipline
Bollmann-Dodd expects teams to pay closer attention to spending on models and tokens. He compared that shift with the cost discipline that followed rising cloud bills, when organizations adopted financial operations practices to understand and control infrastructure spending. As AI use expands, teams will have to weigh the cost of model calls against the work they help complete.
That makes unbounded experimentation harder to sustain. Developers and organizations will need to decide where AI assistance is useful, which tasks justify the expense and how much output a workflow can absorb. A lower token bill doesn’t guarantee a better result, just as frequent model use doesn’t prove value. Teams will need to connect spending with outcomes such as completed work and reduced effort, while accounting for the human review that AI-generated output can require.
The report offers a snapshot of adoption, while the discussion highlights the operational questions that adoption leaves open. Cloud-native developers are using AI at higher rates than other developers, but the next test is whether teams can turn that use into dependable work, keep project maintenance manageable and control the cost of running AI workflows. Those measures will show more than usage rates alone about how deeply AI is changing software development.