Sergey Brin, Google Co-Founder | All-In Live from Miami

Watch on YouTube ↗  |  May 20, 2025 at 15:08  |  33:24  |  All-In Podcast
Speakers
Sergey Brin — Co-founder, Google
Chamath Palihapitiya — CEO, Social Capital
David Friedberg — CEO, The Production Board
Jason Calacanis — Angel Investor / Founder, LAUNCH

Summary

The All-In hosts interview Google co-founder Sergey Brin live in Miami about his return to the company and the current state of AI. Brin says AI progress is compounding faster than the early web or smartphones, credits large-scale pre-training and newer thinking models plus deep research for the latest step up, and says nobody knows the ceiling. On infrastructure he says Gemini mostly runs on Google's own TPUs while Google remains one of the largest buyers of Nvidia chips, and that the hardware layer is not yet abstracted away. He is skeptical of humanoid robots, says past Google robotics efforts failed on software rather than hardware, thinks AI-era smart glasses finally make sense apart from battery life, and expects top Gemini models to stay paid while each generation pushes prior pro-tier capability into the free tier.

  • Brin says the pace of AI dwarfs the early web and smartphone eras, with no known ceiling.
  • Deep research reading a thousand results and following up is framed as AI's real superpower: volume.
  • Gemini mostly runs on Google's own TPUs, while Google is also one of the biggest Nvidia buyers, and chip, memory and communication design are not yet abstracted away.
  • Brin is not a fan of humanoid robots and doubts the human form factor is required.
  • Google's earlier robotics ventures failed on software, not hardware, and were later sold.
  • Smart glasses now look sensible to him after Google Glass; battery life remains the blocker.
  • Top Gemini tiers stay paid on compute cost, but each generation's free tier matches the previous pro tier.
  • Model architectures are converging toward one general model rather than many specialized ones.
Ideas
Sergey Brin Co-founder, Google 2:08
AI capability compounding faster than anything
Brin says AI progress is compounding faster than any technology wave in his career: the early web spread quickly but barely changed month to month, while today's systems change so much that a month away leaves you behind, and he calls it the most exciting technological moment of his life. He points to two step-ups he worked on personally, large-scale pre-training and more recently post-training with thinking models, and says nobody knows where the ceiling is. The practical superpower, in his framing, is volume: deep research can pull a thousand results, run follow-on searches for each and read them deeply, which would be a week of work for him, and he now defers to AI for his own math and coding.
Sergey Brin Co-founder, Google 13:02
Robotics gated by software, not hardware
Google acquired and later sold roughly five robotics companies, including Boston Dynamics, and separately built Everyday Robots internally before winding it down. Brin's read on all of those attempts is that the hardware was fine and the robots were cool, but the software was never good enough to make them truly useful, and that has been true every single time Google tried. He expects that to stop being true eventually, so he treats robotics as a software-gated category to monitor rather than a solved hardware opportunity.
Sergey Brin Co-founder, Google 13:23
Less bullish on humanoid robot form factor
Brin calls himself the one weirdo who is not a fan of humanoid robots, partly because Google acquired and later sold at least two humanoid robotics startups. The standard argument for humanoids is that the world is designed around the human form and that you can train on human video, but he thinks that does not give AI enough credit: AI can learn how to handle situations through simulation and real life, and he doubts a machine needs the same number of arms, legs and wheels as a person to work. He is explicitly less bullish on the form factor while conceding that a lot of smart people are building humanoids, so he would not discount it entirely.
Sergey Brin Co-founder, Google 20:51
AI glasses now sensible, battery still limits
Brin admits he got the timing totally wrong on Google Glass and that the technology simply was not ready then, but says glasses with a screen now make much more sense to him and that it is a cool form factor in the AI era. The blocker he names is battery life, which Google and others still need to overcome. He frames it as the plausible mainstream human-computer interface to watch rather than something he is acting on today.
Sergey Brin Co-founder, Google 25:28
Google runs TPUs, still buys Nvidia
Asked whether the dependency on Nvidia will eventually be abstracted away by transpilers and alternative silicon, Brin says Gemini mostly runs on Google's own TPUs, while Google also supports Nvidia, is one of the big purchasers of Nvidia chips, and offers them to Google Cloud customers alongside TPUs. He adds that at this stage the stack is for better or worse not that abstract: given the sheer amount of computation these models require, the exact chip, how the memory works and how communication works are pretty big factors, and AI itself is not yet good enough to reason through that for you. That leaves Google with a vertically integrated silicon position for its own frontier models plus continued heavy Nvidia purchasing, and no near-term commoditization of the hardware layer.
Up Next

This All-In Podcast video, published May 20, 2025, features Sergey Brin discussing AI-SECTOR, ROBO, Humanoid robots, Smart glasses, GOOG, NVDA. 5 trade ideas extracted by AI with direction and confidence scoring.

Speakers: Sergey Brin  · Tickers: AI-SECTOR, ROBO, Humanoid robots, Smart glasses, GOOG, NVDA