Google launches agentic cloud stack

Google used Cloud Next on April 22 to launch Gemini Enterprise Agent Platform, a new Google Cloud system for building, deploying, governing, and optimizing AI agents. (cloud.google.com) The platform is the next version of Vertex AI, and Google said future Vertex AI services and roadmap updates will ship through Agent Platform instead of as a separate product. (cloud.google.com) Google said the platform adds agent integration, DevOps, orchestration, and security tools on top of model selection, tuning, and agent-building features. It also gives customers access to more than 200 models through Model Garden, including Google and third-party systems. (cloud.google.com) Agent Studio is the low-code workspace inside the platform. Google says it lets teams compare prompts, edit code in a visual canvas, connect to search or company data, and move prototypes into production. (docs.cloud.google.com) Google also folded the developer tools into a broader Gemini Enterprise lineup. The Gemini Enterprise app is designed for employees to discover, run, share, and manage agents in one secured environment tied to company identity and data controls. (cloud.google.com) That app also supports an open partner ecosystem with agents from companies including Oracle, Salesforce, and ServiceNow, according to Google. The company said IT teams will get a single control plane to manage permissions, oversight, and auditability as organizations deploy thousands of agents. (cloud.google.com) Google’s pitch is that agent software needs its own hardware stack. At the same event, it introduced TPU 8i for low-latency inference and TPU 8t for training larger models, saying both chips are built for the “agentic era.” (blog.google) The company paired that product launch with usage numbers meant to show demand is already here. Google said nearly 75% of Google Cloud customers use its artificial intelligence products, 330 customers processed more than 1 trillion tokens in the past year, and direct customer API traffic now exceeds 16 billion tokens per minute, up from 10 billion last quarter. (blog.google) Google is trying to sell enterprises a full stack at once: models, agent tools, employee apps, governance, and chips. Cloud Next made that argument in one line of products rather than a single chatbot announcement. (cloud.google.com) (blog.google)