Google control‑plane push

Google heads into Cloud Next 2026 with a different pitch: the main event is the software layer that runs, governs, and monitors cloud systems. (googlecloudevents.com) (siliconangle.com) Google Cloud Next runs April 22-24 at Mandalay Bay in Las Vegas, with opening and developer keynotes on April 22 and April 23. The featured speaker list includes Chief Executive Officer Thomas Kurian, Chief Technologist for AI and Infrastructure Amin Vahdat, and President of Google Cloud Platform and Site Reliability Engineering Brad Calder. (googlecloudevents.com) Google’s own event materials still promise advances in artificial intelligence, infrastructure, and security, but the conference page also points attendees to infrastructure and Google Kubernetes Engine sessions before the show opens. The company’s Google Cloud Next hub is already highlighting “Top Infrastructure and GKE Sessions at Cloud Next ’26” alongside launch posts for storage, networking, and compute. (googlecloudevents.com) (cloud.google.com) A control plane is the part of a cloud platform that decides what runs where, who can do what, and how operators see failures. Google has described that layer in plain product terms for years: Workflows coordinates services through a central orchestrator, and Cloud Composer uses Apache Airflow as a “universal control plane” for data and machine learning pipelines. (cloud.google.com 1) (cloud.google.com 2) That framing has become more concrete in 2026 as Google adds more operator controls around automated systems. In January, Google said Airflow 3.1 on Cloud Composer adds human-in-the-loop approvals, deadline alerts, and notifier integrations so a workflow can pause for a person before it keeps running. (cloud.google.com) Google has been making the same case in Kubernetes, the cluster software that many large companies use as a control room for applications. At KubeCon Europe in March, Google said it was making Autopilot available inside Standard Google Kubernetes Engine clusters and pointed to tools for agents to manage, analyze, and monitor workloads through its open-source Model Context Protocol server. (cloud.google.com) The contrast with last year is sharp. Google’s official wrap-up for Next ’25 said the company made 229 announcements, ran more than 10 keynotes and spotlights and 700 sessions, and centered much of the show on Gemini models, video, audio, and multi-agent systems. (cloud.google.com) This year’s preview coverage says the enterprise story is moving from model demos to the machinery that keeps those systems under policy and under watch. SiliconANGLE wrote on April 20 that Google is trying to make orchestration, observability, and governance the real center of gravity at Next 2026. (siliconangle.com) That puts the spotlight on the people who run production systems, not just the teams testing new models on stage. If Google follows through in Las Vegas, the biggest announcements may be about who approves an action, how a workflow is traced, and where operators regain control when automation goes wrong. (siliconangle.com)