Sundar Pichai, CEO of Alphabet | The All-In Interview

The episode opens at the Google plaque with a welcome to the CEO of Alphabet, followed by a short preview of questions covered later: whether Google is at risk of being disrupted by AI, the AI Mode experience being tested in labs for search, a request for views on the leaders of OpenAI, xAI, Meta and Microsoft, and a question about the company's biggest regret. In the preview, the regret answer mentions acquisitions that were "debated hard" and came close, and offers "Maybe Netflix" when pressed for a name.

Follow the user. All else will follow.

After a brief show intro, the interview proper begins with a personal note: the interviewer and guest started at Google on the same day in 2004, in the same "Noogler class." The interviewer then recites the guest's track record as stated in the conversation: running Chrome, Chrome OS, Drive and Google Maps; ten years as CEO of Google, now Alphabet; the stock rising four and a half times to a $2 trillion market cap; and quarterly revenue growing from $20 billion to nearly $100 billion.

Asked how the job has been, the response centres on product building. It describes Google as set up by its founders around a "deep computer science approach" applied to things that affect people day to day, calls that product and technical culture the essence of the company, and describes running a company of this scale as a privilege.

The central challenge is posed directly. The interviewer describes the view on Wall Street and in Silicon Valley that AI creates a new paradigm for human-computer interaction: users ask chat interfaces questions and get complete answers. The interviewer puts search ad revenue at about a $200 billion run rate out of $360 billion in total revenue, and most of the profits. The dilemma as framed is that disrupting yourself too quickly could put that revenue at risk. The guest had not read The Innovator's Dilemma but said they were familiar with the concept.

The answer begins with history. It cites Google Brain underway in 2012, the DeepMind acquisition in 2014, and a declaration in 2015 that the company would be "AI first," on the belief that AI would drive the biggest progress in search. The argument is that access to information is "very far from a zero sum construct." Concrete evidence follows: transformer-based BERT and MUM improved search quality, and AI Overviews, launched about a year earlier, is described as used by over one and a half billion users in over 50 countries.

The empirical claims are about query behaviour. For queries that trigger AI Overviews, the answer reports query growth that continues over time and a wider range of query types. AI Mode, being tested in labs and set to be discussed further at Google I/O, is described as a full AI experience in search with follow-on conversational queries, where models use search "as a real native tool." In that setting, the average query is said to be two to three times as long as in search two years earlier, with people typing in what the response calls "literally long paragraphs."

The craft of search is very hard.

The response leans on Google's long practice of measuring user intent. It calls the company "very metrics driven" and treats user engagement as the deciding signal. It then reframes the dilemma itself:

I think the dilemma only exists if you treat it as a dilemma, right?

Two precedents support this. The first is the shift to mobile, when people doubted whether ads would work with limited screen space and the transition "ended up working great." The second is YouTube's response to TikTok. Shorts initially "absolutely didn't monetize anywhere near long form," but the company leaned into the user experience and worked out monetization later. The interviewer ties this to the old Google principle of following the user, and the guest agrees.

The interviewer then raises standalone apps and cites figures said to come from recent court testimony with March data, while noting uncertainty about where the data came from. Those figures put the Gemini app at 350 million monthly users, ChatGPT at 600 million and Meta AI at 500 million. The question is whether the standalone app is the wrong way to measure Google's AI bet, compared with AI built into search.