Google makes agents central

Google used its Cloud Next conference on April 22 to tell customers and investors that AI agents are now the center of its enterprise business. (usnews.com) At the three-day event in Las Vegas, Chief Executive Sundar Pichai and Google Cloud Chief Executive Thomas Kurian said the company’s AI tools are ready for production use by large businesses, not just experiments. Kurian said in the keynote that “the experimental phase is behind us.” (usnews.com; cloud.google.com) Google also folded a group of products under the “Gemini Enterprise” name, including a bigger version of Vertex AI, its platform for building business AI systems, and added new governance and security controls for agents. Google said nearly 75% of Google Cloud customers now use its AI products, and 330 customers processed more than 1 trillion tokens in the past 12 months. (usnews.com; blog.google) An AI agent is software that can plan, decide and take actions with less human input than a standard chatbot. Google’s pitch is that companies want one vendor to supply the model, the cloud infrastructure, the data tools and the controls needed to run large numbers of those systems. (usnews.com; cloud.google.com) That sales pitch comes as OpenAI and Anthropic have also shifted harder toward enterprise customers, which Reuters described as the industry’s most reliable revenue stream. Pichai said Alphabet still expects to spend $175 billion to $185 billion on capital expenditures this year, with “just over half” of its machine-learning computing investment going to the cloud business. (usnews.com) Google paired the software message with a large infrastructure deal. TechCrunch reported on April 22 that Mira Murati’s Thinking Machines Lab signed a new Google Cloud agreement valued in the single-digit billions for AI infrastructure built on Nvidia’s GB300 chips. (techcrunch.com) Google confirmed the agreement in a same-day announcement, saying Thinking Machines will expand its use of Google Cloud AI Hypercomputer and become one of the first customers to use Google Cloud systems with Nvidia GB300 NVL72. Google said early testing showed a 2X increase in training and serving speed versus prior-generation graphics processors. (googlecloudpresscorner.com) The companies said Thinking Machines is using Google Kubernetes Engine, Spanner, Cloud Storage and other services to support frontier-model training and its Tinker product, which TechCrunch described as a tool for automating custom frontier AI models. That gives Google a way to sell not only raw computing power, but also databases, orchestration software and storage in the same contract. (googlecloudpresscorner.com; techcrunch.com) Google’s argument in Las Vegas was that agents are not a single product category but a stack of products that starts with chips and ends with software workers inside a company. The test for that strategy now is whether enterprise customers buy enough of that stack to justify Google’s rising AI spending. (cloud.google.com; usnews.com)