newsfilter
← Interviews
All-In Podcast29 min

Bill Maris: How Google Could Crush AI Competitors, Why Small Funds Win, and AI's Atari Stage

Bill Maris, Chamath, Jason, Friedberg, Thomas, Sachs, Alex Komoroske

Fireside Chat

Bill Mares and Section 32 Overview

  • Bill Mares, founding CEO of Google Ventures (GV), is returning to the investment world with a new fund, Section 32.
  • Section 32 has raised $150 million as its initial fund size, a departure from Mares's previous experience managing multi-billion dollar funds.
  • Mares states a core investment mandate: "We're going to invest for a financial return. Any other metric is impossible to measure and therefore won't succeed."
  • Mares previously founded Google's data center business, incubated Waymo, Google X, and Calico, and served as Google's VP of Special Projects.
  • Mares describes the current technological landscape as a century-long shift where the next 100 years will see change "orders of magnitude" larger than the previous century due to AI.

Four Key Lessons on Investing

  • Lesson 1 (Seeing the Future): Mares compares his 1997 experience of spotting a server in a Wall Street closet to glimpsing the future of the internet, leading him to quit his job to build a data center business.
  • Lesson 1 (Entrepreneurial Risk): Mares illustrates the necessity of "insanity" in entrepreneurship by recounting tarring a leaking apartment roof during a thunderstorm in Vermont to protect servers, choosing to risk personal injury over data loss.
  • Lesson 2 (Predicting Adoption): Mares uses photos of inaugurations (1989 vs. 2005) to show how quickly technology adoption (cameras, live-streaming) becomes ubiquitous, emphasizing that successful entrepreneurs see "secret" futures others ignore.
  • Lesson 3 (Computer Science): Mares argues one should never "bet against computer science," citing his use of AI/machine learning (when internally labeled "machine learning" to avoid "AI" stigma) to design Google Ventures' portfolio construction.
  • Lesson 3 (GV Performance): Mares reports that Google Ventures (2009–2018) achieved estimated returns of 4.1x, performing within the top quartile of VC returns despite early internal skepticism.
  • Lesson 4 (Fund Size vs. Performance): Mares presents data indicating small funds outperform large funds, asserting "small funds outperform large funds" is a mathematical reality.
  • Lesson 4 (Performance Metrics): Data cited shows funds smaller than $750 million averaged 4.76x DPI returns, while funds larger than $1 billion averaged 2.42x.
  • Lesson 4 (Portfolio Math): Mares calculates that to achieve a 3x return, a $500 million fund needs $15 billion in exits, whereas a $7 billion fund requires $210 billion in exits, a figure that exceeds most years' total venture-backed IPO/M&A value.
  • Section 32 Track Record: Section 32 has deployed six funds averaging $400 million each, all performing in the top decile, with early investments in CrowdStrike, Cohere, and Coinbase.

Market Dynamics and Future Trends

  • AI and Google's Leverage: Mares predicts Google will eventually cut token costs by 80%, using its capital as a weapon to undercut competitors like OpenAI and Anthropic, potentially threatening their business models.
  • Valuation Risks: Mares warns that companies staying private longer to avoid public scrutiny are forcing overpriced assets onto 401(k) holders and retail investors once they eventually go public, creating "bag holders."
  • Bimodal Venture Returns: Despite some fund's "excessive" returns, Mares notes venture capital remains bimodal, with 75% of funds losing money and a small handful of winners generating massive returns.
  • AI Investment Strategy: Mares likens current AI to the Atari era, projecting a shift to the "PlayStation 10" stage within five years; he focuses on infrastructure (controllers, physics engines, GPUs) rather than large model training.
  • Life Sciences Shift: Mares notes a flight of capital to India and China for biotech due to U.S. regulatory hurdles, NIH/CDC funding cuts, and an "anti-science vibe," while U.S. investors focus on computational biology.
  • Deep Tech Viability: Mares suggests deep tech is becoming more tractable for investment as AI enablement and physics engines accelerate timelines for long-cycle capital-intensive projects.

Panel Discussion on Fund Structures

  • Andreessen Horowitz Comparison: Panelist Sachs questions if Section 32's model contradicts the need to scale to firms like Andreessen Horowitz (a16z), citing the "late-stage" preference of larger funds.
  • Incentive Misalignment: Sachs argues that large funds (e.g., $5 billion) can return 1.01x and still raise the next fund, whereas a small fund (e.g., $500 million) must return 3x to survive, yet the larger fund manager earns more absolute carry on the lower multiple.
  • Founder Incentives: Sachs notes founders often accept inflated valuations from massive funds (e.g., $250M for 1% at a $4B valuation) rather than strategic advice from smaller funds, distorting market prices.
  • Market Correction: Sachs predicts the current preference for late-stage investing and giant capital bases is unsustainable and that the "pendulum will swing back" toward earlier-stage, smaller funds.
  • Check Size Determination: Sachs emphasizes that fund size dictates strategy, noting that a $500 million fund can diversify across 20–25 names, whereas larger funds face pressure to write $50M+ checks that push valuations out of proportion.