Google CEO Sundar Pichai on the future Google is building

Google and Alphabet CEO Sundar Pichai has defended his company’s AI-driven transformation of Search while acknowledging growing publisher concerns about vanishing web traffic. Nilay Patel posts for The Verge’s Decoder podcast an in-depth conversation with Pichai after Google’s annual I/O developer conference, covering everything from internal restructuring to the timeline for artificial general intelligence (AGI).

A company rebuilt for the AI era

Pichai described how the launch of ChatGPT prompted him to fundamentally reorganize Google. He merged the company’s two leading AI research groups, Google Brain and DeepMind, into a single unit called Google DeepMind. He also centralized AI infrastructure under a dedicated senior vice president and appointed a chief AI architect. Search, previously split across multiple leaders, was consolidated under a single executive.

Pichai also introduced weekly AI product reviews, which he attended personally, to ensure every consumer-facing AI feature received direct scrutiny before launch.

“I pivoted the company to be AI-first,” Pichai said. “I realized we need a core model and a core infrastructure team to power everything we are doing across Google.”

Search is changing, and publishers are worried

The conversation grew pointed when Patel raised the concept he calls “Google Zero,” the prospect that Google will eventually send no traffic at all to external websites as it answers more queries directly on its own results page. Pichai has previously pushed back on the idea. This time, Patel confronted him with a direct quote from Roger Lynch, CEO of Condé Nast, who told another podcast: “Last year I told our teams, assume there is no search. You have to have your businesses planned as if search is zero.”

Pichai declined to tell publishers how to run their businesses, but said Google remains “committed to connecting people to what’s out on the web.” He added that lower-quality clicks are declining naturally as search improves, and that Google has added more links within AI-generated overviews over the past year.

When Patel showed Pichai a live search result for “best Chromebook” on his phone, pointing out that the AI overview gave one recommendation while the first organic result and a New York Times link offered different answers, Pichai conceded: “It’s probably more opinionated than it should be for the particular query you showed me.”

Agents, YouTube, and the next phase of search

Pichai outlined three overlapping directions for Google’s AI products:

• An intelligent search box that delivers answers and may trigger tasks rather than just listing links
• Gemini Spark, a cloud-based agent platform that can carry out multi-step tasks such as booking tickets on a user’s behalf
• Antigravity, a coding agent platform Google uses internally as well

Pichai said these products will converge over time. He compared the underlying logic to folders or notebooks: a common primitive that should work consistently across different Google surfaces.

On YouTube, Pichai confirmed that Google is training its AI models on YouTube videos and changing YouTube search so that results drop users directly into the most relevant section of a video. Patel asked whether Google was prepared to face the same conflicts with YouTube creators that it currently faces with publishers in copyright disputes. Pichai pointed to Google’s existing opt-out tool for web publishers, called Google-Extended, and said the company continues to have conversations with rights holders, but stopped short of making specific commitments to creators.

Public anxiety and the case for AI

Patel pressed Pichai on polling data showing widespread public distrust of AI, citing research that young people actively dislike the technology and that seven in ten Americans oppose new data center construction. Pichai rejected the idea that this is simply a marketing problem.

“It makes sense to me why people would feel concerns about it,” he said. “People are standing and talking about how AI could make a lot of jobs go away. Why wouldn’t you feel a sense of anxiety about it?”

He called for greater industry responsibility on energy use, workforce adaptation, and public involvement in decisions about a technology he described as more profound than fire or electricity.

How close is AGI?

The interview ended with Pichai addressing remarks made by Google DeepMind CEO Demis Hassabis at the close of the I/O keynote. Hassabis told the audience that humanity is standing “at the foothills of the singularity.” Pichai said he and Hassabis define the singularity in this context as the arrival of AGI, meaning AI systems capable of performing a wide range of cognitive tasks at a level comparable to humans.

Pichai said there is broad consensus among frontier AI researchers that AGI is coming “sooner rather than later,” with some placing the timeline at around three years. He declined to commit to a specific date, arguing that the pace of progress already makes the label less important than the practical reality.

“Three years from now, whether you and I call it AGI or not doesn’t matter,” Pichai said, “because it’ll be very, very powerful, and we have to prepare for it.”