Google unveils Agentic Cloud push
- Google used Cloud Next ’26 to launch Gemini Enterprise Agent Platform, folding Vertex AI into a new system for building, running, and governing AI agents. - The platform adds Agent Studio, Agent Runtime, Memory Bank, Registry, Gateway and evaluation tools, while Google said 75% of Cloud customers use its AI. - The push ties agents to Google’s cloud, data and TPU stack as rivals race for enterprise AI control. (blog.google)
Google used Cloud Next ’26 in Las Vegas to pitch a new idea to business customers: build AI agents on Google’s cloud, not just chatbots. (cloud.google.com) (blog.google) The centerpiece is Gemini Enterprise Agent Platform, announced April 22 as a new system to build, scale, govern and optimize agents. Google said it is the evolution of Vertex AI, and future Vertex AI services and roadmap updates will ship through this platform instead of as a standalone service. (cloud.google.com) In plain terms, Google is packaging the tools for multi-step software workers under one roof: models to reason, software to connect to business systems, and controls so information-technology teams can see what agents are doing. Google said the package also includes access to more than 200 models through Model Garden, including Gemini, Gemma and Anthropic’s Claude family. (cloud.google.com) Google’s new low-code Agent Studio is aimed at teams that want to design agents visually, while the upgraded Agent Development Kit is for code-first builders. The reworked Agent Runtime is built for long-running agents that can keep state for days, backed by a Memory Bank for persistent context. (cloud.google.com 1) (cloud.google.com 2) The governance layer is a big part of the pitch. Google said Agent Identity, Agent Registry and Agent Gateway are meant to give each agent a trackable identity, central oversight and enterprise guardrails, while Agent Simulation, Evaluation and Observability are designed to test quality and trace behavior. (cloud.google.com) Google tied the software push to its broader infrastructure business. Thomas Kurian said the “unified stack” for the agentic enterprise combines chips, models, data, agents and security, and Google paired the agent announcements with new eighth-generation Tensor Processing Units and other cloud infrastructure updates. (cloud.google.com) (blog.google) The company also put numbers behind the sales pitch. Google said nearly 75% of Google Cloud customers now use its artificial-intelligence products, 330 customers processed more than 1 trillion tokens each over the past 12 months, and direct customer application programming interface traffic is running above 16 billion tokens per minute, up from 10 billion last quarter. (blog.google) This is also a consolidation move inside Google’s product lineup. Gemini Enterprise is now being presented as the end-to-end system for agent development, orchestration and governance, with a Gemini Enterprise app as the employee-facing front door and a partner marketplace that includes agents from Oracle, Salesforce and ServiceNow. (cloud.google.com) Google spread that message across a big stage. The company said more than 32,000 attendees came to Cloud Next ’26, where it made 260 announcements and repeatedly framed the week around the “agentic enterprise” rather than one-off generative artificial intelligence tools. (cloud.google.com 1) (cloud.google.com 2) The bet is straightforward: if companies move from asking models questions to assigning software agents real business tasks, Google wants those workflows running on Gemini, connected to Google data services, and paid for on Google Cloud. (cloud.google.com 1) (cloud.google.com 2)