a16z Podcast: "Is there an AI Bubble?” with Gavin Baker and David George

Date: October 30, 2025
Host: Andreessen Horowitz (A)
Guests: Gavin Baker, Atreides Management (B); David George, a16z (A)

Overview

This episode takes a deep, data-driven look at the state of AI investment and adoption. Gavin Baker and David George tackle the perennial question: Are we in an AI bubble? Drawing lessons from past tech cycles (notably the 2000 telecom/internet crash), they explore today’s infrastructure build-out, competition among AI model and hardware providers, business model evolution, and changing market dynamics for everything from SaaS to semiconductors and robotics.

Key Discussion Points & Insights

1. Are We in an AI Bubble? (00:00–05:42)

• Comparing Tech Cycles:

  • Gavin Baker, having lived through the 2000 bubble, draws a crucial distinction:

    “At the peak, 97% of the fiber that had been laid was dark. Contrast that with today, there are no dark GPUs.” — Gavin Baker, 00:15

  • In 2000, vast infrastructure was unused (dark fiber); today, all GPU capacity is being strained.
  • Valuations differ significantly; Cisco in 2000 peaked at ~150–180x earnings; Nvidia is closer to 40x now.

• Infrastructure & ROI:

  • Trillions of dollars are being funneled into data centers—more than the interstate highway system (when adjusted for inflation).
  • Current ROI on major AI infrastructure remains strong.

    “Since they ramped up capex, [AI infrastructure companies] have seen, call it, a 10 point increase in their ROIC.” — Gavin Baker, 04:29

• Conclusion:

  • Both agree: No bubble, evidence supports real value and demand.

2. Adoption: From Internet Era to AI Era (05:42–07:22)

• Barriers Are Lower:

  • Unlike the internet, which required building demand and supply sides, AI tools can reach billions instantly through APIs and cloud.

    “All you have to do is light them up via API or turn on your website… you can get to instant distribution. A billion people right away.” — David George, 05:50

• Deep Pockets Backing AI:

  • The companies betting big (Google, Meta, Microsoft, etc.) generate hundreds of billions in free cash and hold enormous cash reserves, giving “an $800 billion buffer.”

• Competitiveness & ‘Race Mentality’:

  • Quote attributed to Larry Page:

    “I’m happy to go bankrupt rather than lose this race.” — recounted by Gavin Baker, 07:05

3. Round-Tripping, Competitive Dynamics, and Nvidia’s Role (07:22–10:07)

• Round-tripping Concerns:

  • Small-scale “round tripping” of funds (e.g., Nvidia investing in partners who in turn buy Nvidia chips) is happening, but is rational and not destabilizing.

• Competition:

  • Nvidia’s main competitor is Google (with TPUs/DeepMind), not other chipmakers.

    “Nvidia’s biggest competitor… is Google, because Google owns the TPU chip… you could argue they are the leading AI company today.” — Gavin Baker, 08:08

• Strategic Moves:

  • Nvidia’s actions are considered “100% rational.”

    “Jensen’s one of the two best CEOs, along with Elon, I have ever known. And I think he's playing a strong hand really well.” — Gavin Baker, 10:02

4. Model Companies, Infrastructure, and Market Structure (10:07–12:59)

• Early Stage of the Cycle:

  • Drawing internet analogies (e.g. Netscape to ChatGPT), the panel notes how early we are; humility in forecasting is necessary.

    “If ChatGPT is to AI as Netscape Navigator was to the Internet… it’s just very early.” — Gavin Baker, 10:29

• Infrastructure’s Staying Power:

  • Owning “raw ingredients of data, capital and distribution,” today’s Tech Titans (MAG 7) have strong positioning but face existential risks if they fail to execute.
  • On Google’s reaction to ChatGPT:

    “ChatGPT was Pearl Harbor for Google and we’re going to see how they responded and they're slowly starting to respond.” — Gavin Baker, 12:25

• Gross Margins in AI:

  • AI companies, particularly model providers, will have structurally lower gross margins than SaaS/software did at their peak.

    “It’s going to be a long time before we see… a frontier lab with gross margins anywhere near SaaS or Internet era margins.” — Gavin Baker, 13:41

5. SaaS and Business Model Shifts in the Age of AI (13:58–18:42)

• Fear of Margin Compression:

  • Application SaaS is evolving. Success in AI will necessarily mean lower gross margins, which shouldn’t be feared but seen as a signal of product use.

    “Don’t be scared… look at declining gross margins kind of as a mark of success rather than a badge of shame.” — Gavin Baker, 16:20

  • Historical reference: Amazon’s margin journey; Microsoft’s successful transition to cloud.

• Investor Perspective:

  • Investors can accept lower margins, if supported by higher adoption and revenue.

    “If you’re an application SaaS company… don’t be scared and look at declining gross margins… as a mark of success rather than... shame…” — Gavin Baker, 16:20

6. Consumer AI, Search, and Interface Futures (18:42–21:45)

• AI as Portal:

  • The consumption model is shifting from directing users to third-party sites (as with Google Search), to outcome-based agent experiences.

• Distribution Remains King:

  • Companies with the largest user bases (Google, Meta) hold strong advantages. Newcomers with novel AI browsers face an uphill battle unless they differentiate — Google’s caution is strategic.

• The Feedback Flywheel:

  • RL (Reinforcement Learning) from large user bases amplifies product improvement—a dynamic similar to classic consumer internet flywheel effects.

7. Chips, Hardware, and the Semiconductor Battle (22:47–25:41)

• Nvidia vs. Google TPU:

  • The AI hardware landscape is defined by the contest between Nvidia and Google’s TPU.

    “Nvidia is no longer just a semiconductor company… now arguably a data center level company with the level of architecting they’re doing.” — Gavin Baker, 23:17

  • Broadcom and AMD are attempting to offer alternatives, but “most of those ASICs are going to fail,” predicts Baker.

• Market Structure:

  • If Google sells TPUs externally, it may shake up the landscape. Amazon’s silicon team is talented (Trainium 3 is expected to be materially improved).

8. Business Models & Sources of Disruption (25:41–29:28)

• Outcome-based Payment:

  • Tasks with clear metrics (e.g. customer support) shift pricing to resolution/outcome.

    “Humans were fundamentally paid based on outcomes… and a lot of AI will be augmenting humans, but probably also replacing some humans.” — Gavin Baker, 27:01

• Affiliate/Marketplace Pressures:

  • Much-discussed inefficiencies (like Google’s relentless focus on search advertising over outcomes/marketplace) will be narrowed by AI.

• Work and Society:

  • Citing Elon Musk, the vision is for work to become optional as AI augments or replaces more human labor.

9. Robotics: Tesla vs. Chinese Entrants (29:28–30:29)

• Humanoids Are Coming:

  • Robotics is described as “very real,” with competition boiling down to “Tesla versus the Chinese.”

    “It’s going to be Tesla versus the Chinese… in the same way it’s Tesla versus the Chinese in cars.” — Gavin Baker, 29:36

• Learning from Humans/YouTube:

  • Humanoid form factors are seen as inevitable, because they can learn from observing humans, and demonstration is easier.

    “It's easier for a human being to put on a suit and show the robot how to do it." — Gavin Baker, 29:54

Notable Quotes & Moments

| Timestamp | Quote | Attribution | |-----------|-------|-------------| | 00:15 | “At the peak, 97% of the fiber that had been laid was dark. Contrast that with today, there are no dark GPUs.” | Gavin Baker | | 04:29 | "Since they ramped up capex, [AI infrastructure companies] have seen, call it, a 10 point increase in their ROIC." | Gavin Baker | | 07:05 | “Larry Page apparently internally said ‘I’m happy to go bankrupt rather than lose this race.’" | Gavin Baker (on Google) | | 10:02 | "Jensen’s one of the two best CEOs, along with Elon, I have ever known." | Gavin Baker | | 10:29 | “If ChatGPT is to AI as Netscape Navigator was to the Internet… it’s just very early.” | Gavin Baker | | 12:25 | "ChatGPT was Pearl Harbor for Google." | Gavin Baker | | 13:41 | "It’s going to be a long time before we see... a frontier lab with gross margins anywhere near SaaS or Internet era margins." | Gavin Baker | | 16:20 | "Don’t be scared… look at declining gross margins... as a mark of success rather than... shame..." | Gavin Baker | | 27:01 | "Humans were fundamentally paid based on outcomes… a lot of AI will be augmenting humans, but probably also replacing some humans." | Gavin Baker | | 29:36 | "It’s going to be Tesla versus the Chinese… in the same way it’s Tesla versus the Chinese in cars.” | Gavin Baker | | 29:54 | "It's easier for a human being to put on a suit and show the robot how to do it." | Gavin Baker |

Timestamps of Major Sections

• 00:00–05:42: AI Bubble? Comparisons, data center build-out, ROI debate
• 05:42–07:22: How AI adoption differs from the internet, financial strength of Big Tech
• 07:22–10:07: Nvidia, round-tripping, and competition
• 10:07–12:59: Model companies, market structure, margins
• 13:58–18:42: SaaS, margin debate, business model shifts
• 18:42–21:45: Consumer AI, portals, distribution power
• 22:47–25:41: Semiconductors: Nvidia vs. Google, ASICS, Broadcom/AMD
• 25:41–29:28: Business models, outcome-based pricing, societal impact
• 29:28–30:29: Robotics, Tesla vs. China, humanoid inevitability

Tone

The tone is candid, data-rich, and filled with analogies to past tech eras. Baker’s style mixes humility (“humility is an important virtue for an investor”) with clear, forceful opinions, and David George pushes for practical implications for founders, investors, and incumbents.

Conclusion

Gavin Baker and David George urge investors and operators alike to avoid the recency and fear-informed mistakes of previous bubbles. AI’s infrastructure is being used, and Big Tech’s financial footing is strong. While business models are under pressure — especially for SaaS and consumer applications — now is the time for clear-sighted strategy, not panic. The next frontiers, from chips to robots, will be defined by execution, talent, and the ability to adapt to genuinely lower-margin but higher-scale businesses.