Google bets on agents

Google used its Cloud Next conference on April 22 to tell corporate buyers that AI agents are now the main product it wants to sell. (usnews.com) At the three-day event in Las Vegas, Chief Executive Sundar Pichai and Google Cloud chief Thomas Kurian said the company’s tools had moved past experimentation and into production for large businesses. Reuters reported that Google is folding products under the “Gemini Enterprise” name and adding new governance and security controls for agents. (usnews.com) An AI agent is software that can plan, decide and carry out multi-step work with limited human input, rather than just answer a prompt once. Kurian said the main use of Vertex AI, Google’s cloud AI platform, has shifted from older machine-learning projects to companies building custom agents. (usnews.com) Google’s pitch is not just models but a full package: chips, cloud infrastructure, software to build agents, and tools to manage large fleets of them. In his April 22 post, Pichai said Google launched a new Gemini Enterprise Agent Platform because customers are now asking how to manage “thousands” of agents. (blog.google) The scale numbers were part of the sales case. Google said nearly 75% of Google Cloud customers are already using its artificial intelligence products, 330 customers processed more than 1 trillion tokens over the past 12 months, and its first-party models now handle more than 16 billion tokens per minute through direct application programming interface use, up from 10 billion last quarter. (cloud.google.com) Google paired that software message with new hardware for the same workload. On April 22, it introduced eighth-generation Tensor Processing Units, split into TPU 8t for training models and TPU 8i for inference, the step where a trained model generates answers or actions for users. (blog.google) Google said TPU 8t can scale to 9,600 chips in one superpod, while TPU 8i is designed for low-latency inference, the fast response time needed when agents are taking actions across multiple steps. The company said both chips will be available later in 2026. (blog.google; blog.google) To show that this is not just a keynote demo, Google pointed to customers already deploying agent-style systems. In a customer roundup published April 22, Google highlighted companies including Capcom, Citi Wealth, Home Depot and Mars as examples of businesses using agents in operations, service and research. (blog.google) Some of those deployments predate this week’s event. Deutsche Bank said in September 2025 that its DB Lumina research agent had been adopted by hundreds of analysts to automate research tasks while meeting privacy and compliance requirements. (cloud.google.com) Google is making this push as rivals chase the same budget. Reuters said OpenAI and Anthropic have also shifted aggressively toward enterprise customers in recent months, while Pichai reaffirmed Alphabet’s 2026 capital-spending plan of $175 billion to $185 billion and said just over half of machine-learning compute investment would go to the cloud business. (usnews.com) The bet in Las Vegas was that companies will buy the whole stack, not just a chatbot. Google’s message was that agents are becoming the reason to rent its cloud, use its models and wait for its next chips. (cloud.google.com; usnews.com)