Not Once in 18,900 Tries | Michael Mauboussin on What History Says About AI Growth

Key Concepts

• Base Rates (Outside View): A forecasting framework that evaluates the probability of an outcome by looking at historical data from a relevant reference class rather than building projections from the bottom up.
• Intangible Assets: Non-physical assets (e.g., software, R&D, brand, training) that are typically expensed on the income statement rather than capitalized on the balance sheet.
• The Four S’s of Intangibles: Scalability, Sunkenness, Spillovers, and Synergies.
• Reference Class Forecasting: A methodology for predicting project outcomes based on the historical performance of similar large-scale projects.
• Economic Profit: Calculated as (Return on Invested Capital - Cost of Capital) × Invested Capital.
• Operational Effectiveness: The ability to perform similar activities better than rivals, which empirical evidence suggests is a significant driver of competitive advantage.

1. Assessing the AI Boom via Base Rates

Michael Mauboussin discusses the plausibility of OpenAI’s revenue projections (growing from $3.7B in 2024 to $185B+ by 2029).

• The Data: Using the Compustat database (1950–present), Mauboussin analyzed 18,900 firm-years for companies with $2B–$5B in revenue. He found that a 108%–118% compound annual growth rate (CAGR) over five years is a nine-and-a-half standard deviation event—essentially unprecedented in history.
• The Argument: While not physically impossible, such growth is not a "base case" scenario. Investors should treat these projections as outliers rather than expected outcomes.
• Diffusion Models: OpenAI’s rapid user adoption (e.g., ChatGPT reaching 100 million users in two months) is a positive indicator, but the "three legs of the stool"—great product, great talent, and massive capital requirements—remain difficult to execute simultaneously.

2. Large-Scale Projects and Capital Intensity

Drawing on Bent Flyvbjerg’s research on 16,000 projects, Mauboussin highlights the risks of the current AI infrastructure buildout:

• The Statistics: Less than 9% of large projects are completed on time and on budget; only 0.5% deliver what was promised on time and on budget.
• The AI Context: AI data centers face significant hurdles, including permitting, energy, and cooling. Reports suggest 25%–50% of data center projects face delays, reinforcing the "base rate" expectation that these projects will likely exceed cost and time estimates.

3. Intangible Assets and Market Dynamics

Mauboussin explains why intangible-intensive businesses exhibit "fatter tails"—more extreme successes and more frequent bankruptcies.

• The Four S’s:
  • Scalability: High upfront costs but low marginal costs for distribution.
  • Sunkenness: Intangible investments (e.g., failed software) have little recovery value compared to physical machinery.
  • Spillovers: Best practices (like the 3-point shot in the NBA or smartphone features) disseminate rapidly, making it hard to maintain a long-term lead.
  • Synergies: Innovation is the recombination of digital building blocks, allowing for faster growth.
• The "Secret Sauce": Large companies (Magnificent 7) are growing faster than historical norms because they invest heavily in proprietary software that creates both scale and differentiation, which does not easily diffuse to competitors.

4. Value Capture and Competitive Strategy

• Value Distribution: Using the Brandenburger and Stuart framework (Willingness to Pay vs. Price vs. Cost), Mauboussin argues that much of the value created by AI will likely accrue to the consumer due to competition.
• Preemptive Deterrence: Companies are engaging in "peacocking"—making massive, sometimes aspirational, capital commitments to scare off competitors. However, this is a high-stakes game where firms must commit capital before the economics of the business are fully understood.
• Operational Effectiveness: Even if AI becomes a commodity, companies that implement it with superior operational effectiveness will outperform.

5. Accounting Quirks and Investor Discipline

Mauboussin critiques the "asset-light" narrative:

• The Misconception: Because accounting rules expense R&D and marketing (intangibles) rather than capitalizing them, companies appear "asset-light" when they are actually investing heavily.
• The Solution: Investors should break down SG&A into maintenance (sustaining current revenue) and discretionary investment (future growth).
• Value Investing: Adjusting book value to include capitalized intangible investments can improve the "value factor" signal, helping investors identify truly cheap companies that traditional price-to-book metrics might misclassify.

Synthesis and Conclusion

The core takeaway for investors is to focus on the Return on Invested Capital (ROIC) and the amount of capital being invested, regardless of whether that investment appears on the balance sheet or the income statement. Mauboussin emphasizes that while AI is a transformative technology, the historical base rates for growth and project execution suggest that investors should be cautious of "unprecedented" projections. Success in the AI era will likely be determined by a company's ability to manage capital intensity, maintain talent, and execute with superior operational effectiveness.