Industry executives and experts share their predictions for 2026. Read them in this 18th annual VMblog.com series exclusive.
By Ant Newman, Director of Content, Spectro Cloud
Kubernetes definitely ain’t new any more (65% of organizations we surveyed in our latest research have been using it for over five years)… it would be fair to say that the experimentation phase is well and truly over.
In fact, I’d argue that we’ve entered the everywhere era: where Kubernetes seems to be central to every hot topic in enterprise IT, from AI to legacy modernization.
When there’s this much change going on, it’s no easy task to make bold predictions. But here’s what the data tells us about where Kubernetes is really heading.
1. Kubernetes gets serious about scale – again
After a cautious 2024 caused a dip in adoption stats, confidence came roaring back in 2025. More than half of organizations now run over 20 clusters, and one in five has broken the 100-cluster mark.
The median enterprise runs production Kubernetes across five or more environments – spanning clouds, data centers, sovereign and GPU environments, and increasingly, the edge.
Looking ahead, we can expect this scale and diversity to become the norm, as trends like multicloud, on-prem repatriation, and data sovereignty push more clusters into more places. By the end of 2026, we predict at least 30% of enterprises will operate more than 100 clusters.
Scale and diversity, unfortunately, breed cost. We already found in 2025 that cost is rising sharply: 88% saw their Kubernetes TCO grow in the past year, and two-thirds said they’re under more pressure than ever to reduce spend.
Expect tension in the boardroom as ROI comes under scrutiny, and investigation of new tooling, and new practices, to address the cost of expansion.
2. The edge goes mainstream – thanks to AI
Here’s a stat that’ll make you blink: half of enterprise Kubernetes users are already running clusters at the edge, up from 38% last year.
Why? Because AI inference workloads demand it. Real-time decisioning – from industrial automation to healthcare imaging – can’t tolerate a round-trip to the cloud. As one respondent said, “AI inference workloads that demand real-time decisions belong at the edge – latency is non-negotiable.”
We’re not talking about a handful of IoT boxes anymore. Many adopters already run 50 or more edge clusters, and a few are well into four digits.
And because 90% of organizations expect their AI workloads on Kubernetes to grow, the next 12 months will see the edge cluster explosion. Would it be crazy to think that 75% of K8s-using enterprises will have deployments at the edge?
3. AI isn’t just the workload – it’s the optimizer
Every survey under the sun (including a few of ours) will confirm that managing Kubernetes at scale is complex. No surprises there.
Configuration, patching, rightsizing, troubleshooting – these are the actions that consume platform teams’ time, and block them from delivering the responsiveness and service levels their users demand.
With operating budgets constrained, and a shortage of skills and headcount continuing unabated, what’s the answer?
Well, AI of course.
In our 2025 research, 92% said their businesses are investing in AI-driven optimization tools to keep costs in check. For years, the industry has talked about autoscaling and FinOps; this is the year AI starts taking the wheel.
Interviewees we spoke to shared some concerns – about job loss as much as about technical feasibility – but the momentum around AI tooling for K8s operations seems inexorable.
So we’ll make the prediction: one in three enterprises will be using AI-assisted operations by the end of 2026, from copilots that tune resources, to bots that generate manifests and fix misconfigurations.
4. Virtual machines refuse to die – they just move in
68% of Kubernetes adopters say they today run more than half their workloads in Kubernetes.
That’s cool for containers, but what about all those VM workloads you’re still running on vSphere?
Our prediction: a good chunk of them will end up in Kubernetes too, thanks to KubeVirt. The reason is simple: economics and control. After the VMware licensing saga, CIOs are keen to consolidate on open platforms. The challenge isn’t tech; it’s culture – convincing veteran VM admins to share turf with platform engineers. But that’s solvable, with time.
26% already run KubeVirt in production (over 50% among large enterprises) – and a third of organizations say they plan to migrate VMs into Kubernetes this year. By 2026, we reckon KubeVirt adoption will pass 50%. It will become the de facto bridge between the old IT world and the cloud-native one.
5. Platform engineering grows up
Last year, 70% said they had adopted platform engineering. This year, 80% say they have a mature platform function, and 90% of developers can self-serve via an internal developer platform (IDP).
That’s progress – but we’re still in the awkward teenage years. Over half admit their clusters are “snowflakes,” and operations are still “highly manual.”
The good news? Our data shows that when central platform teams manage standardized pipelines via Helm or GitOps, they deliver the best outcomes across all five DORA-style metrics – deployment frequency, change failure rate, recovery time, and so on.
Our call: by the end of 2026, the “golden path” pattern becomes standard practice. Platform engineering shifts from aspirational to expected – and the next big debate will be about governance automation, not whether to centralize.
The bottom line
If there’s one theme running through all of this, it’s that Kubernetes has hit a real degree of maturity – but the AI era is reshaping it faster than ever.
As AI boosts K8s growth and diversity, the real question is whether platform teams can keep control… not just of governance, security and availability, but cost. We predict a few sleepless nights in store.
Want the data behind these predictions? Read the full 2025 State of Production Kubernetes report at bit.ly/sopk2025
##
ABOUT THE AUTHOR
Ant Newman, Director of Content, Spectro Cloud
Ant has spent 20 years working across the enterprise technology landscape, covering trends from IoT to cybersecurity. After stints agency-side, and at Gartner and Cisco, he joined Spectro Cloud to lead its content creation. His passion is data-driven thought leadership, and for the past four years he has driven Spectro Cloud�s State of Production Kubernetes research. Ant has a degree in English from Oxford University and lives in the UK with his partner and two children.
