Cloud Next previews Agentic Data Cloud and split TPU 8t/8i chips as Google commits $185bn to AI

Google used Cloud Next in Las Vegas to show how it wants companies to build, run and control large fleets of artificial intelligence agents. (blog.google) The centerpiece was the Gemini Enterprise Agent Platform, which Google described as a full-stack system to build, scale, govern and optimize autonomous agents for work. Sundar Pichai said customers are moving from asking whether they can build one agent to asking how to manage thousands. (blog.google 1) (blog.google 2) Google paired that software pitch with new hardware. At Cloud Next on April 22, it introduced eighth-generation Tensor Processing Units in two versions: TPU 8t for training large models and TPU 8i for low-latency inference, with general availability planned later in 2026. (blog.google 1) (blog.google 2) The split matters because training and inference are different jobs. One chip is tuned to teach a model on huge data sets, while the other is tuned to answer quickly when an agent is actually doing work for a user. (blog.google) Google also introduced what it calls the Agentic Data Cloud, a data layer meant to let agents use business context and act on it. In its recap, Google said that package includes Cross-Cloud Lakehouse, built on Apache Iceberg, so customers can query data where it already sits, including in Amazon Web Services. (blog.google) That is the practical problem Google is trying to solve. An agent is only useful if it can reach company data, understand permissions, and complete multi-step tasks without forcing a business to move every database into one vendor’s stack. (blog.google 1) (blog.google 2) The spending behind the pitch is unusually large. Alphabet told investors on February 4 that 2026 capital expenditures should total $175 billion to $185 billion, and Pichai said just over half of the company’s machine-learning compute investment this year is expected to go to the Cloud business. (cnbc.com) (blog.google) Google also used customer metrics to show demand. The company said nearly 75% of Google Cloud customers now use its artificial intelligence products, 330 customers processed more than 1 trillion tokens each over the past 12 months, and direct customer API traffic rose to more than 16 billion tokens per minute from 10 billion last quarter. (blog.google) The competitive backdrop is the cloud industry’s push from copilots toward agents that can take actions on their own. CNBC reported Alphabet’s 2026 spending forecast topped the expectations of hyperscale peers, while Google’s own Cloud Next materials framed this year’s event around the “agentic enterprise.” (cnbc.com) (blog.google) Google’s message in Las Vegas was that enterprise AI is becoming an infrastructure business as much as a model business. The company is selling chips, networking, data access and governance as one package for customers that want agents to do real work inside the enterprise. (blog.google) (blog.google)