For all the keynotes about the future of AI infrastructure, the most reassuring thing at KubeCon + CloudNativeCon Japan 2026 was seeing cloud-native technology doing unglamorous, real work. A short passage in the opening keynote made the case better than any roadmap slide: manufacturers in Japan are already running AI on the Kubernetes foundation, not experimenting with it.
An optics company, faster and smarter with AI
The keynote speaker described an optics manufacturer that has been improving its product development and manufacturing processes with AI. The mechanism is the part worth noting: they took the investments they had already made in cloud-native infrastructure and pointed new workloads at them — training models, serving those models, and making the business faster and smarter in ways they couldn’t before.
The quiet insight is in the last sentence of that story: underneath all of it, there is still infrastructure fundamentally enabling the shift to AI. The model is the headline, but the platform is the reason it shipped. AI didn’t replace the cloud-native foundation; it rode on top of it.
A slide from the opening keynote, “Cloud Native AI Adoption Surges”: Uber runs 30M+ predictions per second on a Kubernetes-based control plane, Wayve uses cloud native technologies for self-driving inference, and ZEISS Vision Care used Dapr Agents to go from concept to production in two months.
Subaru is on the same path — and won
The same thread connected directly to Subaru, which is following the identical arc: take cloud-native systems, and use them to enable AI workloads. The punchline was a celebration — Subaru was announced as the winner of that year’s KubeCon Japan Case Study Contest.
The opening keynote congratulating Subaru, the 2026 KubeCon + CloudNativeCon Japan Case Study Contest winner.
That is the signal I look for at these events. A case-study contest win means a real, reviewed production deployment — not a demo, not a PoC that died at the end of the quarter. Automotive ML on Kubernetes, judged by the community as the standout.
From Subaru’s keynote: large container image pulls went from about 3 hours to about 3 minutes.
The end of Subaru’s keynote in the Main Hall at PACIFICO Yokohama.
A walk through the Uber booth
And then there was the show floor itself. A quick walk through the Uber booth — a name everyone recognizes, running real production at a scale most of us will never see. No deep technical reveal in the clip, just the reminder that the sponsors on the floor are the teams operating the systems the talks describe. When you want to know what “AI on Kubernetes” means in practice, the booth is where the practitioners stand.
The real-world takeaway
Strip away the futurism and the pattern is consistent:
- Established manufacturers (optics, automotive) are already on cloud-native AI.
- The infrastructure they built for cloud-native apps is the same infrastructure now serving models.
- Recognition like the Case Study Contest is how the community separates production from theater.
KubeCon Japan 2026 wasn’t a conference about whether this works. It was a conference full of people who already shipped it.
More of my Japan 2026 coverage is in the KubeCon Japan 2026 media-partner preview.
