KubeCon EU landed in Amsterdam this year. It also landed in Amsterdam in 2024. Same city, same conference, two years apart — which gave us something most surveys don't get: a genuine like-for-like.
We've been working with platform teams for over a decade now. That builds up a lot of pattern recognition — but patterns aren't signal until something external confirms them. So we surveyed hundreds of practitioners on the floor both times, with the same core questions, to understand what's actually changing — and what isn't.
Here's what stood out.
In 2024, 47.6% of respondents named increasing stability and reliability as their main platform objective. In 2026, reliability and security topped the list again. Same answer, two years running, from hundreds of practitioners.
Here's the part that hasn't shifted either: when those same teams were asked what was getting in the way, stability didn't register as a challenge at all. Not because it's solved — because it's assumed. Nobody writes "keep the platform running" on a list of problems. It's just supposed to work. Instead, the challenges were everything that makes that assumption hard to keep.
Hiring the right people, too many tools for the team size, operational overload, no time for automation — four responses, essentially tied, each around 29–32%. Four different answers describing the same condition. Teams are trying to stabilize something they don't have enough people to run, with more tools than any reasonably-sized team can support, and no slack in the system to fix any of it.
Two years later, that picture looks almost identical.
The cluster distribution isn't a bell curve — it never was. In both 2024 and 2026, the largest cohorts were at opposite ends: teams running 1–5 clusters, and teams running 50 or more. In 2026 those two groups were almost exactly matched.
A team running 3 clusters and a team running 80 are not solving the same problem. They're not even in the same conversation. Worth knowing which one you're in.
AWS, Azure, bare metal, Google — in that order, both years. Nobody consolidated. Most teams are managing several providers simultaneously not because they planned to, but because that's what the organization handed them. Every additional provider is another upgrade cycle, another security surface, another claim on platform team attention that wasn't in the original headcount plan.
This question was new in 2026, and it's the one I keep thinking about. When asked how teams plan to make improvements, 99 out of 143 said: internal staff only.
Teams are planning to build their way out of a staffing and tooling problem using the same staff who named staffing and tooling as their biggest obstacle. I understand the instinct — autonomy matters to platform teams, and outside help is rarely plug-and-play. But internal-only delivery works when you have capacity. When you don't, the fix and the problem are the same thing.
Some of this I saw coming. In my 2024 predictions I called the IDP trend going full force — the survey confirms teams are still in the middle of that transition. I predicted AIOps would become a real operational tool — it's showing up in the 2026 focus areas, though more slowly than I expected. And I've been writing for a while about open source winning over proprietary stacks, which the multi-cloud picture here supports indirectly.
What I underestimated was how persistent the structural constraints would be. The direction was right. The pace of change was slower and the gap between ambition and capacity stubbornly wider than I anticipated. That gap has a name. We've been calling it the platform assembly tax — the compounding cost of keeping a platform alive that was assembled from too many parts, with not enough people, and never quite enough time to address the root cause. The practitioners who answered our survey didn't use that term. But across two years and hundreds of responses, they described it with uncomfortable precision.
If any of this sounds like your week, I'd love to talk.
Survey conducted at KubeCon EU 2024 and 2026. Self-selected respondents on the conference floor. Treat year-on-year comparisons as directional.