When Agents Replace Labor | $11 Billion Tech Manager on What Investors Miss About AI

Key Concepts

• Inevitabilities: Long-term, high-probability outcomes in technology that persist despite market noise and short-term volatility.
• Agentic AI: Software systems designed to act autonomously, performing tasks (e.g., customer service, coding) that were previously human-dependent.
• Compute Scarcity: The concept that high-performance computing (GPUs, HBM) is becoming a constrained resource, similar to oil, rather than a commodity with a declining cost curve.
• S-Curve Adoption: A framework for identifying where a technology sits in its lifecycle (emerging, inflecting, compounding, or mature).
• Economic Bottlenecks: Critical components or services in a value chain that are essential for a new technology to function, often representing the best investment opportunities.
• Token Economics: The shift in software pricing from per-user licensing to outcome-based or token-usage-based models.

---

1. Investment Philosophy: Identifying Inevitabilities

Tony, a portfolio manager at T. Rowe Price, emphasizes that successful long-term investing requires separating market "noise" from "signal." He defines an "inevitability" as a trend that is highly likely to occur over a 3–10 year horizon.

• Core Argument: Investors should focus on the "driver of the driver." For example, in the semiconductor space, the inevitability was the need for more and better compute power as Moore’s Law slowed down.
• Current Inevitability: The transition from software designed for human input to software designed for autonomous agents. This shift is expected to uncap economic productivity by removing labor as the primary constraint on output.

2. The AI Buildout: A "Space Race" with Multiple Moons

Tony rejects the idea that the AI market is a zero-sum game. He describes the current buildout as a "space race with multiple moons," where various tech giants (Google, Meta, Tesla, OpenAI) are pursuing different applications of intelligence.

• Market Size: The Total Addressable Market (TAM) for AI is likely 10x larger than initial estimates because it creates new productivity enhancements rather than just replacing existing workflows.
• Capex Justification: Unlike the 2000s fiber-optic bubble—where utilization was low and costs were declining—current AI infrastructure (GPUs, HBM) is at maximum capacity, and the cost of production is increasing (an "inflated cost curve"). This suggests the current spending is grounded in real, immediate demand.

3. Portfolio Construction Framework

Tony utilizes a "sleeve-based" approach to balance risk and reward:

• Compounders (60%): Established companies that outperform the benchmark through the cycle.
• Emerging Tech (20–30%): High-growth, early-stage companies that provide "idiosyncratic alpha" (e.g., Lumentum Holdings in optical networking).
• Value/Inflection Sleeve: Mature companies undergoing a business model shift or experiencing a specific catalyst.

4. The Future of Work and Economic Impact

Tony maintains a centrist view on the economic impact of AI:

• Productivity vs. Job Loss: While some roles will be displaced, technology is becoming easier to use (e.g., natural language as the new programming language). This lowers barriers to entry for innovation.
• Human Element: Creativity and judgment remain the most critical human skills. AI acts as an "elevator," allowing individuals to perform more sophisticated work rather than replacing them entirely.
• Quote: "English is the new programming language."

5. Sell Discipline and Risk Management

Tony’s sell discipline is driven by valuation and market feedback:

• The "Beat and No-Go" Signal: If a company reports strong earnings (beating expectations) but the stock price fails to rise, it serves as a signal that the market may be fully valued or that there is hidden risk.
• Humility: He emphasizes the importance of trimming positions to avoid "brutal" drawdowns, noting that it is better to recycle capital into new opportunities than to hold onto a winner that has reached a valuation extreme.

6. Synthesis and Conclusion

The main takeaway is that we are currently transitioning from the investment phase of AI to the inferencing and deployment phase. While cyclicality and macro shocks are inevitable, the fundamental shift toward agentic and physical AI is a decade-long evolution. Investors should focus on companies that control the "economic bottlenecks" of this new infrastructure—such as memory (HBM) and optical networking—while remaining disciplined about valuation and the "rate of change" in market sentiment.