This week at Cloud Next ‘26, we announced a bunch of new tools to take businesses all over the world into the agentic era. We introduced the Gemini Enterprise Agent Platform, a single system for developers to build agents, along with our eighth-generation TPUs. They’re our most efficient hardware to date, delivering dramatically more power and performance to put AI agents to work. With nearly 75% of our customers already using Google Cloud AI — and 330 organizations processing over a trillion tokens each in just the past year — the agentic age is here and Cloud is helping businesses make the shift.

At Cloud Next ‘26, we announced our new custom-made TPUs, TPU 8t and TPU 8i. We built our TPUs from the ground up to power the Google products you use every day and help companies around the world use AI. But how do TPUs actually handle their increasingly demanding AI workloads? Here’s a quick, helpful video on the inner workings of these small but mighty processors.