newsfilter.io
Interview, Fireside Chat

David Cahn: Why Servers, Steel and Power Are the Pillars Powering the Future of AI | E1186

20VC with Harry StebbingsDavid Cahn, Harry Stebbings
Watch on YouTube →

AI Capital Expenditure and Strategic Sentiment

  • Consensus on Overspending: Major tech leaders (Mark Zuckerberg, Sundar Pichai) acknowledge aggressive overbuilding of AI infrastructure is a strategic necessity rather than a calculated, risk-free investment; they recognize the high risk if AGI does not materialize.
  • Oligopoly Dynamics: The "prisoner's dilemma" drives the CapEx race, as Microsoft, Google, and Amazon (representing $7 trillion in combined market cap) must spend aggressively to defend their dominance and prevent competitors from eroding their cloud monopolies.
  • Startups vs. Incumbents:
    • Pro-Startups: High CapEx by incumbents creates an oversupply of compute, lowering costs for startups (consumers) and increasing their gross margins.
    • Anti-Startups: Massive capital requirements erect significant barriers to entry, potentially cementing the power of the existing oligopoly and preventing new entrants.
  • Investment Horizon: Venture capitalists and executives are investing based on a long-term view (living "many years in the future") where AI fundamentally transforms society, accepting that immediate ROI is unlikely.

The Industrialization of AI: "Servers, Steel, and Power"

  • Physical Asset Scarcity: AI is becoming an industrial revolution where the data center is the primary asset; no frontier model will ever be trained on the same data center twice due to rapid hardware obsolescence.
  • Scaling Laws: As models grow, scaling laws become the dominant factor, requiring data center architecture to evolve from 100,000 GPU clusters to potential 300,000 GPU clusters.
  • Three Critical Resources:
    • Servers: Continued innovation and competition in chips (NVIDIA, AMD, Broadcom) driven by the need for better price-to-performance ratios (e.g., B100 vs. H100).
    • Steel/Construction: Massive demand for real estate developers (Cyrus One, QTS) and general contractors (DPR) to build new facilities.
    • Power: The current energy grid cannot support AI demand; capital will flow to new generation (solar, batteries) driven by AI economics rather than pure regulation.
  • Vertical Integration: Successful AI requires deep coupling between model teams and data center teams; companies like Meta (Zuck) and Elon Musk are vertically integrating, whereas separate entities like OpenAI and Microsoft face integration challenges.
  • Financial Engineering: Shift toward off-balance sheet financing via long-term leases (20 years) to mask the immediate cash impact of CapEx, effectively turning these commitments into debt instruments backed by the credit of tech giants.

Founder Assessment and Venture Capital Strategy

  • The "Four-Dimension" Founder Framework: To build a $10B+ company, founders need a combination of:
    • Science applied to Technology (Engineering).
    • Science applied to Human (Self-mastery/Hardcore mindset).
    • Intuition applied to Technology (Product vision, e.g., Ivan Zhao, Brian Chesky).
    • Intuition applied to Human (Leadership).
  • Customer Validation vs. Churn: Lessons from past investments (Marketo, UiPath, Snowflake) show that customers often claim they will churn or build solutions in-house but continue paying for expensive, complex products; investors should prioritize revenue retention over verbal feedback.
  • Sequoia Culture:
    • "Slugging": The only definition of success is generating billion-dollar gains; the firm operates on high constraints (1-2 deals per year) to force conviction.
    • Rebuilding: The firm "breaks down and rebuilds" investors to adhere to a high bar for excellence, where every investment must be a potential "YouTube" or "Instagram."
  • Sourcing Philosophy: David Khan identifies as an "outbound hunter," believing founders do not want to meet VCs; he utilizes creative tactics (e.g., Cameo videos, daily Loom messages) to earn the right to speak with founders.

Market Dynamics and Specific Insights

  • China Competitiveness: While China is often cited as two years behind the US, Khan warns against underestimating their capabilities and believes the US must assume strong competitor performance to maintain its lead.
  • Open vs. Closed Source: Khan favors a mixed ecosystem; he does not believe AGI is imminent enough to warrant total restriction, viewing the coexistence of open (Llama) and closed models as beneficial for the world.
  • Compute as Currency: "Compute" is a euphemism for physical assets; the "currency" is the ability to build and maintain the physical infrastructure, which faces significant construction and energy constraints.
  • Chip Wars: Betting against Jensen Huang (NVIDIA) is considered impossible; however, new entrants and competitors (AMD, Broadcom) will increase, driven by the massive gross margins in the sector.
  • First Investment Success: Starburst Data was Khan's first major "win," secured through diligent customer calls on Databricks and a personal relationship built over 12 months.
  • Personal Philosophy: Khan cites being a twin as the most influential factor in his life, fostering a sense of non-conformity and hyper-competitiveness that drives his work ethic (waking at 5 AM, biking to work).