What Happened at KubeCon India 2026? A Complete Recap

A Quick History, For Anyone Who Missed the Earlier Editions

KubeCon India is younger than it feels.

The first one happened in Delhi in December 2024. It was small by global KubeCon standards, but it mattered. It was the first time CNCF brought its main conference to India, even though India already had one of the biggest open source communities in the world.

The second one moved to Hyderabad in August 2025 and grew fast. Around 4,000 people came for 12 keynotes, 10 lightning talks, 56 sessions, and 30 booths. People who were there talk about it more like a reunion than a conference. Friends from social media meeting in person, engineers arguing about eBPF over biryani.

This year it moved again, to the Jio World Convention Centre in Mumbai, running June 18-19 with 55 sessions and 8 lightning talks, the biggest the India edition has been so far.

It was not the only thing happening in the city that week either. Several other open source events ran in Mumbai around the same days, and you would see the same conference badges at every coffee shop nearby. It felt less like one event and more like Mumbai had quietly become open source's capital for a week.

This growth is not an accident. India has one of the largest and fastest-growing open source and cloud native communities in the world, and a startup ecosystem that leans heavily on open source to keep infrastructure costs down. KubeCon India did not create that momentum. It just finally caught up to where it was already happening.

The Stat Everyone Kept Repeating

Every KubeCon has one number that ends up in every recap and every hallway conversation. This year it traced back to CNCF's data and the opening keynote.

The framing that stuck: a large majority of organizations now run AI workloads on Kubernetes. CNCF's 2025 Annual Cloud Native Survey puts inference workloads on Kubernetes at 66%. Yet only 7% of organizations are deploying AI models on a daily basis. The infrastructure is in place. The daily habit, and the trust behind it, is not.

CNCF leadership also noted from the stage that Indian contributors now rank third in the world for CNCF commits, which puts the community here in a strong position to help close that gap between having the infrastructure and trusting it enough to use it every single day.

One keynote made that scale feel concrete. A session from JioStar walked through how JioHotstar streams cricket to tens of millions of concurrent viewers, built on Kubernetes and OpenTelemetry. For an Indian crowd, that is not a hypothetical scale problem. Half the room has felt that exact app slow down during a big match. It grounded the rest of the conference in something real.

Platform Engineering Has Grown Up

If you only read the AI headlines from this event, you would miss one of its steadiest, most useful tracks. The platform engineering sessions were not about whether to build an internal developer platform. That question is settled now. They were about the real, unglamorous problems teams hit once they already have one. A developer hitting a wall the golden path never accounted for. A platform team realizing their setup made one workflow easy and three others quietly painful.

This is the part of cloud native that never trends online, but it is exactly what separates a team that scales smoothly from one that scales painfully.

Security Is Trying to Catch Up to AI Agents

Security talks this year had a specific new angle: how do you secure an AI agent, not just the cluster it runs in.

Rahul Jadhav from AccuKnox showed how to use SPIFFE and OpenFGA to give AI agents an actual identity with scoped permissions, instead of letting them run with whatever access happened to be sitting around. A lightning talk from Dhruv Puri on why cluster-wide policies are a hidden risk made a similar point from a different angle.

There is a useful nuance in the data underneath these sessions. CNCF's survey shows security is still a top adoption challenge, cited by around 36% of respondents, but its relative ranking has actually eased over recent years. For the first time, cultural change within teams has overtaken every technical concern as the number one barrier. The on-the-ground takeaway from hallway conversations was consistent: most teams that already have an AI agent running in their cluster have not gone back and updated their security setup to match.

The agents showed up faster than the rules did. Nobody seemed surprised by that. Everyone seemed a little worried about it.

What Was Happening on the Show Floor

Walk into the exhibition hall and you basically had two completely different vibes running side by side.

One side was the Solutions Showcase, companies with booths, demos, and a pitch ready to go. Cast AI, Chainguard, Microsoft Azure, and VMware by Broadcom were the ones you kept running into. VMware was showing their Kubernetes runtime for private cloud and AI workloads. Chainguard had a demo on supply chain risk, basically what happens when your team starts shipping faster with AI and nobody checks what just got pulled in. Cost was also everywhere, which felt new. There was a packed session on making cost a platform responsibility, and another one on the hidden cost of telemetry data conversions from the observability side. People are clearly starting to care.

A few steps away was the Project Pavilion, which had a completely different energy. CNCF open source projects staff their own kiosks to talk directly to the people actually using their software. CNCF supports maintainer participation through pass allocations and travel funding via the Dan Kohn Scholarship, so the folks at these kiosks are the real deal. Lightning talks covered KubeEdge at the edge, Harbor's GitOps updates, Kyverno's next policy features, KubeVela, and Longhorn's storage roadmap. Nothing trended on social media. That is kind of the point.

The Solutions Showcase tells you what companies want to sell next year. The Project Pavilion tells you what is holding everything together right now.

The Community Side of This Event Is the Part That Sticks With You

The technical sessions are why people fly in. The community is why people come back.

CNCF runs the Dan Kohn Scholarship Program at every India edition, covering registration and travel so cost is never the reason a student or first-time speaker misses out. Pair that with a Cloud Native Novice track, where beginner sessions give complete newcomers a real way in instead of a watered-down advanced talk, and you get a very specific feeling on the floor.

Walk the hallways and you will see a college student who just gave their first ever lightning talk standing two feet away from a principal engineer who works on systems running at real production scale. Nobody is pretending those two people are at the same level. But nobody is treating one of them as less important either. That feeling, more than any single keynote, is what makes KubeCon India feel like a real community event with great technical content, not a vendor expo with a community section bolted on.

The AI SRE Boom on the Show Floor

This is the part that genuinely stood out if you had also been to the 2024 or 2025 editions.

Walking the Solutions Showcase, AI SRE was clearly one of the biggest categories on the floor this year, right up there with observability and platform tooling. Tools that bring AI into incident detection, triage, and root cause analysis were one of the categories everyone was watching closely.

A few names kept coming up, each making a different architectural bet on the same problem. Dr Droid connects to your cloud, code, and observability tools with no agents and no code changes, and builds a live map of how everything connects before it ever investigates an alert. The company describes getting from alert to root cause in under nine minutes. NudgeBee offers pre-built AI assistants that triage alerts, correlate logs, metrics, traces, and recent deployments, then surface a root cause with a suggested fix via PRs and runbooks. On its own site, NudgeBee cites teams cutting MTTR by around 50% and L3 escalations by up to 68%. Middleware takes an observability-platform angle, with an AI agent that traces errors from logs and Kubernetes signals toward the underlying cause, aiming to shorten the path from symptom to source.

The figures above are each vendor's own cited numbers, drawn from their materials and websites, not independent measurements.

The common thread across all three: nobody is struggling with detection anymore. Every serious tool on that floor can tell you something broke. The real competition has moved one step deeper, into the gap between an alert firing and someone actually understanding why it happened, which is still where most of the time in any incident goes.

What This Means If You Were Not There

If you take one thing from this year's event, take this: KubeCon India is no longer the smaller cousin of the global conference. It is becoming its own real signal for where cloud native, and AI infrastructure especially, is actually headed.

Platform engineering matured. Security started catching up, a bit late, with what it means to let an agent take action on its own. OpenTelemetry settled into being the obvious default. AI SRE went from a handful of small startups to a genuinely funded, competitive category with real disagreements between the players on how to do it right. And underneath all of it sat the same gap from the data: broad adoption of AI on Kubernetes, but only 7% daily use.

That gap is the real story of where this industry is right now. Not whether the tools exist. Whether anyone trusts them enough yet to use them every day.

Key Takeaways

• •KubeCon India started small in Delhi in 2024, grew fast in Hyderabad in 2025 with around 4,000 attendees, and hit its biggest scale yet in Mumbai in 2026 with 55 sessions and 8 lightning talks.
• •The framing that defined the event: a large majority of organizations now run AI on Kubernetes, with CNCF's 2025 survey putting inference workloads at 66%, but only 7% deploy AI daily.
• •Indian contributors now rank third in the world for CNCF commits, as noted from the keynote stage.
• •Platform engineering talks have moved past whether to build a platform into the real, specific problems teams hit once they already have one.
• •Security sessions made clear that most teams running AI agents have not updated their access controls to match. Cultural change has overtaken security as the top overall adoption barrier per CNCF's survey.
• •OpenTelemetry is no longer debated. It is just the starting point now.
• •The show floor split into two worlds: the Solutions Showcase where Cast AI, Chainguard, Microsoft Azure, and VMware led, and the Project Pavilion where open source projects like KubeEdge, Harbor, Kyverno, and Longhorn show up with CNCF-supported maintainer participation.
• •The Dan Kohn Scholarship and Cloud Native Novice track are a big part of why this event feels like a real community, not a vendor showcase.
• •AI SRE was one of the biggest categories on the show floor, with Dr Droid, NudgeBee, and Middleware each making a different bet, though it was one part of a much bigger event, not the whole story.

Frequently Asked Questions

Further Reading