Sundar Pichai on Google’s Agentic AI Shift

At Google I/O 2026, Sundar Pichai didn’t just announce new products. He painted a picture of where Google thinks computing is going next, and it’s a pretty clear one: AI is no longer something you “use” occasionally. It’s becoming the layer everything runs on.

The big theme this year was scale, but not just scale for the sake of it. It was about how deeply AI is now embedded into real products people already rely on every day.

Google says it has spent the last decade moving toward an AI-first approach, and now that shift is showing up everywhere. From custom-built AI chips to foundation models to consumer apps used by billions, the idea is simple: control the full stack, move faster, and keep improving continuously.

One of the more striking signals shared was usage growth. Google talked about “tokens,” basically the raw units of AI processing, and the numbers are hard to ignore. In just a couple of years, they’ve gone from trillions per month to over quadrillions. That’s not just technical growth, it reflects how quickly AI is being adopted across products and developers.

And it’s not just infrastructure. It’s people.

Search is changing into something more conversational. AI Overviews and AI Mode are now reaching billions of users, and the way people interact with Search is shifting from single queries to ongoing back-and-forth exploration.

The Gemini app tells a similar story. It has nearly doubled its user base in a year, and the usage patterns are becoming more personal and more creative. Features like image generation and “personal intelligence” are pushing it beyond a traditional assistant model.

What stood out most, though, was the direction everything is moving toward: less typing, more natural interaction.

Whether it’s voice-driven Docs where you can basically speak your thoughts into structured documents, or Ask YouTube that takes you directly to the most relevant part of a video, the goal is the same — reduce friction between intent and output.

Behind all of this is massive infrastructure investment. Google is scaling its custom TPUs and AI systems to support both training and real-time usage at a level that only a few companies globally can even attempt.

But the bigger narrative wasn’t just about speed or scale. It was about trust too. With AI-generated content becoming everywhere, Google is expanding watermarking systems like SynthID and content credentials so users can actually tell what’s AI-generated and what isn’t.

Then comes the real shift: agents.

Google is clearly betting that the next phase of AI won’t just answer questions, it will take action. From Gemini Spark, a 24/7 personal agent, to Search experiences that can run tasks in the background, the direction is toward systems that don’t just respond, but operate.

The takeaway from Pichai’s keynote wasn’t hype. It was momentum.

AI is no longer a feature being added. It’s becoming the foundation everything else is built on.