newsfilter
← Interviews
Baron Capital59 min

The Next Wave of AI Investing: Why Active Management Matters

Michael Lippert, Ashim Mehra, Matt Camuso, Mike Lippert

Webinar, Conference Presentation
  • Market Context & Paradigm Shift

    • Generative AI adoption occurred in two years, a pace significantly faster than the internet (decade) or mobile (five years).
    • Mike Lippert characterizes AI as the fastest-moving inflection point in human history since the internet or industrial revolution.
    • The market recently experienced a reversal where "AI losers" (software, IT services, real estate) began reverting after 2.5 years of underperformance against "AI winners" (semiconductors, equipment).
    • Active management is deemed critical to separate "signal from noise" amidst rapid product announcements, deal speculation, and ex-post narratives.
  • Investment Strategy: "Act Two" of AI

    • Act One focused on the AI build-out (infrastructure, chips, hardware).
    • Act Two focuses on the application layer, inference, and "agentic AI" where AI agents perform tasks autonomously.
    • The Barron Technology ETF (BCTK) portfolio as of late September is split approximately 56% into infrastructure/compute and 42% into application sciences.
    • The firm avoids betting on a single "winner" model, instead positioning for a multi-model world where specific use cases drive value.
    • Portfolio construction emphasizes diversification across nine distinct layers, with no single layer exceeding 21% of the portfolio.
    • Position sizing is determined by conviction, valuation, and future growth potential rather than index weightings; the largest position is 7.5% and the top 10 holdings represent 55% of the fund.
  • Competitive Advantage Framework

    • Capital Intensity: Unlike prior tech cycles, AI is capital-intensive; firms with access to capital, power generation, and data centers possess a structural moat.
    • Data Moats: Companies with proprietary or specialized data (e.g., cybersecurity, observability) can train differentiated models, creating barriers to entry.
    • Scalability & Talent: Competitive advantage now relies on the ability to hire specialized talent and move at a "startup pace" despite large scale.
    • Revenue Models: Successful companies align pricing with customer outcomes rather than just token counts, as efficiency varies by task.
    • Re-underwriting: The firm constantly retests investment theses weekly or monthly, willing to sell if the competitive advantage erodes or the CEO leaves.
  • Sector-Specific Outlooks & Trends

    • Power & Infrastructure: Power generation is identified as the potential "weakest link" in the AI value chain in the US, driving interest in energy providers and "powered shell" data centers.
    • Memory & Storage: Demand is expected to surge due to the rise of agentic inference, which requires extensive data retention.
    • Software Sector: Contrary to passive "AI loser" labels, the firm adds software names at the application layer (e.g., Axon, Hinge Health) that generate unique data or offer specialized utility.
    • Agentic AI: The inflection point is shifting from "chatbots" to "agents" that execute work, starting in software development and expanding to broader knowledge and physical work.
    • Private Market Intelligence: The team actively engages with private companies (including a high-conviction unnamed entity) to identify disruptions and future public opportunities.
    • The "Harness" Layer: Investment interest exists in companies that route queries to optimal models (open, frontier, or proprietary) to minimize cost and maximize accuracy per task.
  • Risk Management & Liquidity

    • BCTK maintains liquidity support of ~$475 million based on the implied liquidity of underlying holdings, despite an average daily trading volume of ~$1.5 million.
    • The strategy includes significant exposure to mid- and small-cap stocks (25% of the portfolio) to capture growth before it reaches large-cap indexes.
    • The firm avoids concentration risks inherent in passive indices (e.g., Nasdaq 100), which reconstitute annually and overweight past winners.
    • Cash flow analysis focuses on Return on Invested Capital (ROIC) rather than just CapEx spending; high spending is acceptable only if it generates superior returns.
  • Q&A Highlights & Forward-Looking Statements

    • Monetization: The firm believes AI value will eventually diffuse across the economy, moving beyond infrastructure suppliers to specific industry applications.
    • Valuation Cycles: While current AI infrastructure growth is hyper-fast, earnings will eventually normalize into a mature, cash-flow-generating phase similar to Apple's current model.
    • Model Strategy: The base case assumes a "harness" or "router" ecosystem where companies optimize costs by selecting the best model for specific tasks rather than relying on a single monolithic model.
    • Selling Discipline: Stocks are sold not only for valuation targets but also when the core investment thesis breaks (e.g., loss of key leadership, erosion of competitive advantage).
    • Long-Term Horizon: The team targets 5-to-15-year investments, having held positions in names like NVIDIA since before the ChatGPT moment (2018).