Google's agent bet

Google is turning AI agents into the center of its cloud sales pitch, betting businesses will pay to deploy digital assistants that can plan and act. (reuters.com) At Google Cloud Next in Las Vegas on April 22, chief executive Sundar Pichai and Google Cloud chief Thomas Kurian said the company’s enterprise push is now built around “Gemini Enterprise” and a new Gemini Enterprise Agent Platform. Google said the three-day conference marked a shift from experiments to production deployments for large corporate customers. (google.com; reuters.com) Google described agents as software that can connect data, people, and goals, then carry out longer-running tasks with tools such as Agent Designer, agent inboxes, skills, and project management controls. Pichai said the platform is meant to help companies “manage thousands” of agents rather than just build one. (google.com; google.com) The money behind the pitch is large. Google Cloud announced a $750 million fund on April 22 to help consulting firms, systems integrators, software partners, and channel partners prototype, build, deploy, and govern agentic artificial intelligence projects. (googlecloudpresscorner.com) Google said the fund will back training, proofs of concept, usage incentives, and embedded engineering teams working with firms including Accenture, Capgemini, Cognizant, Deloitte, Devoteam, HCLTech, and Tata Consultancy Services. Partners including Accenture, Bain, Boston Consulting Group, Deloitte, and McKinsey will also get early access to Gemini models. (googlecloudpresscorner.com; bloomberg.com) Google is leaning on consultants because they often decide which artificial intelligence systems large companies actually buy and deploy. Reuters reported that OpenAI and Anthropic have also shifted resources toward enterprise customers in recent months, where recurring software spending is easier to measure than consumer hype. (reuters.com; bloomberg.com) Google paired the agent push with new infrastructure claims meant to show scale. The company said its first-party models now process more than 16 billion tokens per minute through direct customer application programming interface use, up from 10 billion last quarter, and that nearly 75% of Google Cloud customers now use its AI products. (google.com; google.com) Pichai also said Google will spend $175 billion to $185 billion in 2026 and that just over half of its machine-learning computing investment will go to the cloud business. Reuters reported that Google is trying to prove returns on years of spending on data centers, custom chips, and networking gear as it competes with Amazon, Microsoft, OpenAI, and Anthropic. (google.com; reuters.com) Inside Google, the company said three-quarters of new code is now generated by AI and then reviewed by engineers. Business Insider reported that figure was about 25% in October 2024 and 50% by late 2025, while Pichai said one recent code migration was completed six times faster by agents and engineers working together than by engineers alone a year earlier. (businessinsider.com; google.com) Google’s pitch lands as agent software raises questions about reliability, safety, and oversight when systems are allowed to make decisions and take actions on their own. Kurian told Reuters that the main use of Vertex AI has shifted from older machine-learning work to a surge in customers building custom agents, which helps explain why Google is adding more governance and security controls around them. (reuters.com) The bet is that companies will not just buy models, but whole operating systems for agent work. Google spent Cloud Next arguing it wants to sell that full stack — chips, cloud, models, security, and the agents themselves. (google.com; reuters.com)