Google's Enterprise AI Push

Google used its Cloud Next conference in Las Vegas on April 22 to tell customers that AI agents, not search ads, are the next big business inside the company. (reuters.com) An AI agent is software that can take a goal, use tools, and complete multi-step work with limited human input. Google’s new Gemini Enterprise Agent Platform packages that into one system for building, running, governing, and tuning those agents for companies. (cloud.google.com) Google said the platform folds together Vertex AI’s model and agent tools with new features for integration, DevOps, orchestration, and security. The company also used Next ’26 to pitch Gemini Enterprise as an end-to-end product for customer service, Google Workspace, and other business tasks. (cloud.google.com, cloud.google.com) The company’s sales pitch is a full stack: custom chips at the bottom, Gemini models in the middle, and agent software on top. Google said nearly 75% of Google Cloud customers now use its artificial intelligence products, and 330 customers processed more than 1 trillion tokens each over the past 12 months. (cloud.google.com) That framing marks a sharper turn in how Google talks about making money from artificial intelligence. Reuters reported that executives presented agents as the clearest way to monetize AI inside enterprise software, where companies pay for workflow automation rather than consumer queries. (reuters.com) Google paired that software push with new chip design. Its eighth-generation Tensor Processing Units now come in two versions: TPU 8t for training large models and TPU 8i for inference, the stage where a trained model answers real requests. (cloud.google.com) Google said the split reflects a practical change in AI economics, because training and serving models now need different hardware. TPU 8t is built for large pre-training runs, while TPU 8i is tuned for high-volume inference and reinforcement learning. (cloud.google.com) The company is also trying to make agents work across vendors instead of only inside Google tools. Its Agent2Agent protocol, first introduced in April 2025, is an open standard for agents to communicate with one another across platforms and cloud environments. (developers.googleblog.com, a2a-protocol.org) That open-standard message helps Google answer a basic enterprise concern: companies already run software from Microsoft, Salesforce, SAP, and internal systems. Google has said Gemini Enterprise connects to Workspace, Microsoft 365, Salesforce, SAP, and other business data sources under a central governance layer. (cloud.google.com, cloud.google.com) The backdrop is a crowded market. Reuters said Google is chasing enterprise spending while competing with Microsoft, Amazon, OpenAI, and Anthropic, all of which are trying to turn large language models into paid workplace software. (reuters.com) Google’s message in Las Vegas was that it wants to sell the whole machine at once: chips, models, security, and agents that do office work. The next test is whether customers buy enough of that stack to turn AI enthusiasm into durable cloud revenue. (reuters.com, cloud.google.com)