Google's agent push
- Google made agents central to its enterprise strategy and announced a new Agent Platform at Cloud Next. - The company also unveiled two new TPUs it says are faster and cheaper than prior versions. - This full‑stack push aims to make multi‑agent management and secure orchestration easier for enterprises ( )
Google used its Cloud Next conference in Las Vegas on April 22 to put artificial intelligence agents at the center of its enterprise pitch. (reuters.com) Chief Executive Sundar Pichai said agents are a core part of how Google plans to make money from artificial intelligence through cloud software for businesses. Google announced a new Gemini Enterprise Agent Platform to build, deploy and manage those systems. (reuters.com) (cloud.google.com) In plain terms, an agent is software that can take a goal, break it into steps and use tools or company data to finish the job. Google said its new platform is meant to give companies one place to handle that work, including security controls and orchestration across multiple agents. (zdnet.com) (cloud.google.com) Google also introduced two eighth-generation tensor processing units, or TPUs, its in-house artificial intelligence chips. The TPU 8t is designed for training models, while the TPU 8i is aimed at inference, the step where a trained model answers requests in production. (techcrunch.com) (blog.google) Google said TPU 8t delivers 2.8 times the performance of its prior-generation training chip at the same price. It said TPU 8i improves inference performance by 80% over the previous version and is built for fast response times for agent workloads. (cnbc.com) (blog.google) The push comes as Google tries to sell more cloud services to large companies that want OpenAI-style assistants without handing over control of internal data or workflows. Reuters reported that Google’s cloud business has become a key venue for turning its artificial intelligence research into revenue. (reuters.com) The hardware announcement also shows how Google is trying to compete on the full stack: chips, models, cloud infrastructure and business software. TechCrunch reported that Google is still offering Nvidia chips in its cloud even as it expands its own TPU line. (techcrunch.com) Google framed the new software around a problem many companies are just starting to hit: one agent can be useful, but many agents create a management problem. ZDNet reported that the new platform is designed to streamline automated work while tightening governance and security. (zdnet.com) That gives Google a clearer enterprise message than a simple chatbot pitch. At Cloud Next, the company presented agents as workers to supervise, and the new platform and new chips as the tools to run them at scale. (reuters.com) (blog.google)