The new CNCF cloud native report shows a continual spread of Kubernetes, and container usage, in organizations. The chart above shows a shift from experimenting with Kubernetes to trying it out in production, to running in production.
At the same time, containers, as you’d expect, are also moving steadily into production. Application containers in production have risen from 41% in 2023 to 56% in 2025. Simultaneously, pilot-only container deployments are down to a mere 6%. People no longer play with containers; they move them straight to deployment. From Steven J. Vaughan-Nichols' coverage.
Large organizations (5,000+ employees) are still the slowest to use Kubernetes. There’s a break out of company size by amount of usage in the charts. I don’t really understand how to read it, but the point is that larger companies are the slowest to move to using Kubernetes in production. The report authors theorize on why:
As shown in Figure 11, The heavy representation of large enterprises (45% with 5,000+ employees) suggests that size creates complexity. The large enterprise skew reveals that thousands of applications across diverse technology stacks, decades of accumulated technical debt, risk-averse cultures requiring extensive evaluation, and distributed decision-making slow adoption for these organizations. These organizations are in learning mode, with minimal revenue tied to cloud native technologies.
When it comes to barriers to adoption - things that are difficult with Kubernetes - “culture” has moved to the top of the list:
“Culture,” here, means:
Cultural resistance may manifest in multiple ways across organizations. Developers can express skepticism that containers add unnecessary complexity for simple applications or question whether Kubernetes is production-ready. Operations teams resist what they perceive as developer toys and express concern about troubleshooting containerized systems. Management worries about costly distractions from feature delivery and dependencies on specialized knowledge.
A simplistic reading would be that culture hangups are things like “I don’t trust” and/or “I’m not gettin a priori ROI vibes.” In surveys like these, displacing security from a top two spot is rare, let along for something as catch-all as “culture.”
Claude Opus 4.5 would like to add:
The “culture” point deserves more unpacking. You’re right that displacing security from the top is notable. But I’d push back slightly on your “I don’t trust / no ROI vibes” framing—the report’s description is more specific and structural. It’s departmental friction: devs think it’s overkill, ops thinks it’s a dev toy they’ll get paged for, management sees feature delivery risk. That’s not just skepticism; it’s misaligned incentives across organizational boundaries. Platform engineering exists precisely to solve this coordination problem, which the report nods to.
I’ve found it difficult to estimate how many applications run on Kubernetes over the years.
There’s of course some AI coverage. Of note:
- 52% of organizations don’t build or train AI models at all, they just consume, them.
- Only 7% deploy AI models daily, 47% deploy AI models only “occasionally” (a few times a year).
Check out the full report. Good stuff as always.