Google Cloud bets on agentic AI

Google Cloud used Next ’26 to argue that enterprise AI agents need a full platform, not just a model. It packaged that pitch around a new Gemini Enterprise Agent Platform. (cloud.google.com) The conference took place in Las Vegas from April 22 to April 24, and Google said it made more than 250 announcements across agents, data, security and infrastructure. Google Cloud said more than 32,000 attendees came to the event. (cloud.google.com 1) (cloud.google.com 2) Google described the new platform in operational terms: build, scale, govern and optimize agents. It paired that with an “Agentic Data Cloud,” “Agentic Defense,” and new Gemini Enterprise features for customer service and workplace tasks. (cloud.google.com) (blog.google) An AI agent is software that can take steps on a user’s behalf, not just answer a prompt. Google’s pitch is that companies need routing, policy controls, data access and monitoring around those systems before they can trust them in production. (cloud.google.com 1) (cloud.google.com 2) That framing also explains why Google spent so much time on hardware. The company introduced two eighth-generation Tensor Processing Unit systems: TPU 8t for frontier-model training, and TPU 8i for large-scale inference and reinforcement learning. (cloud.google.com) Google said the split design reflects a change in AI workloads, with pre-training, post-training and real-time serving now needing different machines. It positioned those chips as part of its broader AI Hypercomputer stack of hardware, software and networking. (cloud.google.com 1) (cloud.google.com 2) The company also tied the agent push to rising demand on its cloud. Sundar Pichai said Google’s first-party models now process more than 16 billion tokens per minute through direct customer API use, up from 10 billion last quarter. (blog.google) Pichai said “just over half” of Google’s overall machine-learning compute investment in 2026 is expected to go to the Cloud business for customers and partners. Forbes reported Alphabet’s broader 2026 capital spending plan at about $185 billion. (blog.google) (forbes.com) Google also used Next to stress cross-cloud operations rather than a single-vendor stack. Its infrastructure team said the “agentic enterprise” needs secure systems that can run across compute, Kubernetes, networking and multiple environments. (cloud.google.com) That leaves Google making two linked claims at once: agents will become normal enterprise software, and the companies that profit most will be the ones selling the control plane and the silicon underneath. Next ’26 was Google Cloud’s clearest attempt yet to sell both. (cloud.google.com) (forbes.com)