Google Cloud Unveils Agentic AI and 8th-Gen TPUs at Next '26
- •Google Cloud launched the Gemini Enterprise Agent Platform and eighth-generation TPUs for training and inference.
- •Nearly 75% of Google Cloud customers use AI products, with 330 processing over 1 trillion tokens annually.
- •API usage surged to 16 billion tokens per minute, up from 10 billion in the previous quarter.
Google Cloud introduced new infrastructure and platform updates for agentic enterprise AI at the Cloud Next '26 conference held on April 22, 2026. The company announced the Gemini Enterprise Agent Platform to facilitate the scaling and governance of autonomous agents, alongside the new Virgo Network, a data center fabric designed to support future machine learning workloads.
Adoption metrics show 75% of Google Cloud customers now utilize AI products. Usage volume has increased, with 330 customers processing more than 1 trillion tokens each over the past 12 months. Current aggregate API traffic reaches 16 billion tokens per minute, a 6 billion increase from the 10 billion tokens per minute recorded last quarter. There are 1,302 generative AI use cases currently deployed by organizations.
Google released two distinct eighth-generation Tensor Processing Units (custom chips for ML workloads), one optimized for training and one for inference. Additional announcements included the Agentic Data Cloud, which connects AI agents to business data, and Workspace Intelligence to integrate agentic workflows into Google Workspace. CEO Sundar Pichai and Google Cloud CEO Thomas Kurian emphasized that the transition to agentic enterprises is now a mainstream organizational priority.