Squawk Pod: Davos 2026 – Google DeepMind CEO Demis Hassabis

Date: January 24, 2026
Host: Andrew Ross Sorkin (CNBC)
Guest: Demis Hassabis, CEO of Google DeepMind

Episode Overview

This special Squawk Pod episode, reporting from the World Economic Forum in Davos, Switzerland, features an in-depth conversation between CNBC’s Andrew Ross Sorkin and Demis Hassabis, CEO of Google DeepMind. The discussion explores the rapid evolution and future of artificial intelligence, challenges and opportunities in AI innovation, competition among top labs, industry "frothiness," and the broader economic and societal implications of transformative AI technologies like Gemini.

Key Discussion Points and Insights

The AI Race and Google's Journey

• (03:24) Sorkin frames the ongoing rivalry between DeepMind, OpenAI, and Anthropic, asking what catalyzed DeepMind’s leap in 2026.
• Hassabis: Google and DeepMind’s “long journey” was about harnessing years of research, TPUs, and infrastructure; the recent breakthrough came from efficiently organizing these assets across the company.
  • Quote (03:45): “We always had all the ingredients… [This year] we’ve kind of organized that all in a really efficient, efficient way.”

Productization and Gemini’s Momentum

• (04:14–05:08) The focus shifted from research to speedy deployment across Google’s ecosystem — Gemini 3 now powers products like Search and will increasingly feature in Gmail and beyond.
• On Apple partnership: Hassabis calls Apple's adoption of Gemini a "massive vote of confidence" after a rigorous evaluation, underscoring Gemini’s advancement over competitors.
  • Quote (05:08): “It’s a massive sort of vote of confidence in the quality of our model… Gemini came top.”

AGI (Artificial General Intelligence): Hype vs. Reality

• (05:58–07:10) Sorkin probes whether simply scaling compute will achieve AGI, or if paradigm-shifting breakthroughs are still required.
• Hassabis’ “in-between view”: While scaling current models still yields “amazing gains,” major hurdles—true creativity, continual learning, long-term reasoning—remain. The recipe for AGI may require two to three big new breakthroughs or just more scale, and DeepMind is investing in both paths.
  • Quote (06:24): “It’s an empirical question… we’re still getting lots of amazing gains out pushing the existing paradigms.”

Differentiation and Convergence of Models

• (07:10–07:44) Sorkin questions whether leading models are growing too alike.
• Hassabis: There’s more differentiation at the frontier, with each model excelling in different domains—for example, Gemini’s strength in multimodal understanding and image generation.

The AI “Bubble” Debate

• (07:44–08:18) Hassabis distinguishes between froth in startup investments and value in real product adoption. He sees parts of the sector as overheated (e.g., vast sums for seed-stage startups with no tech) but affirms there are meaningful advances in practical applications.
  • Quote (07:50): “If you look at things like the new hot startups… that seems a little bit frothy to me and perhaps unsustainable.”

Funding, Scale, and Industry Sustainability

• (08:18–09:22) With massive capital requirements, Sorkin wonders if independent labs can keep up. Hassabis feels secure with Google’s balance sheet and product ecosystem, highlighting natural fits for AI (Search, Email, Chrome) and confidence in their competitive position regardless of market cycles.

Compute Limits and “Deep Seek” Hype

• (09:22–10:30) Sorkin asks if a game-changing innovation might let us leapfrog current compute needs.
• Hassabis: While emerging efficiencies may arrive, big compute will remain necessary for training, serving, and exploration. Notably, the Deep Seek moment was “overblown” as it quietly depended on existing Western models for training.

Chip Lifecycles and Google’s “Full Stack” Advantage

• (10:30–11:28) Sorkin discusses hardware depreciation—are today’s data centers “replaced” too quickly?
• Hassabis: Google’s advantage is its integration of custom TPUs, cloud, and research. Old chips can be repurposed for lighter tasks, so investments retain value beyond cutting-edge AI training.

Competitive Landscape: OpenAI, Anthropic, Meta, and Others

• (11:28–12:46) Sorkin queries the threat from Meta, Elon Musk, and China (“xi”).
• Hassabis: Acknowledges all as "extraordinary companies led by ambitious leaders"—the field remains highly competitive, with outcomes still uncertain.

Future Market Structure: “Winners,” Moats, and Personalization

• (12:46–13:51)
  • Hassabis predicts multiple (2–4) major AI players may co-exist, especially in enterprise, but the bar for catching up rises every year.
  • Sorkin asks about business moats, particularly personalization data as a sticky edge.
  • Quote (13:13) Hassabis: “The key is going to be the quality… and the capabilities.”

Impact on Jobs, Skills for the Next Generation, and Societal Shifts

• (13:51–14:58)
  • No major disruptions in employment yet, but entry-level roles may start to shift.
  • Hassabis’ advice to young people: Become “unbelievably proficient with the new tools” and leapfrog traditional ladders via AI-native skills.
  • Quote (14:40): “Get unbelievably proficient with the new tools, immerse yourself in it, become native with it, and then leapfrog… the incumbent people on that [career ladder].”

Notable Quotes & Memorable Moments

• On Gemini leapfrogging competition (05:08):
  “It’s a massive sort of vote of confidence in the quality of our model. So we’re very pleased with that partnership [with Apple].”

• On AGI progress and breakthroughs (06:24):
  “It’s an empirical question… there’s some missing capabilities still. They don’t do continual learning, they don’t have true creativity yet, they don’t do long-term planning and reasoning.”

• On industry “bubble” warning signs (07:50):
  “If you look at things like the new hot startups that raising billions of dollars in a seed round with, you know, no product or technology yet, that seems a little bit frothy to me and perhaps unsustainable.”

• On the unrelenting pace of competition (12:46):
  “I work 100 hour weeks. I’ve been doing that for the last three years and I see no end to that. And I think all the leading labs are doing that… it’s progressively a harder, harder problem [to catch up].”

• Advice for the next generation (14:40):
  “Get unbelievably proficient with the new tools, immerse yourself in it, become native with it, and then leapfrog… whatever professional ladder you’re trying to get onto.”

Segment Timestamps

• Introduction and Context: 00:49–02:27
• DeepMind’s AI Journey: 03:24–04:14
• Gemini in Google Products: 04:14–05:08
• Apple Partnership and Model Maturity: 05:02–05:58
• AGI Roadmap & Model Limitations: 05:58–07:10
• Model Differentiation: 07:10–07:44
• AI Bubble Debate: 07:44–08:18
• Sustainability and Capital Needs: 08:18–09:22
• Compute, Chips, and Deep Seek: 09:22–10:30
• Full-stack and Repurposing Chips: 10:30–11:28
• Competitive Field (Meta, Elon Musk, China): 11:28–12:46
• AI’s Future Market Structure: 12:46–13:51
• Jobs, Youth Advice, and Disruption: 13:51–14:58

Conclusion

This episode offers an insider’s look into the high-stakes world of AI, where foundational research is rapidly meeting commercial deployment, and competition grows fiercer by the month. Demis Hassabis remains pragmatic about both the promise and the perils—technical, financial, and societal—of the generative AI race. The conversation ends with a clear call to prepare for a future where adaptability and AI proficiency are keys to opportunity.