He Quantified 200 Years of Disruption | Kai Wu on Separating Software Survivors from Value Traps

Episode Summary

Executive Summary: The episode argues that software stocks’ selloff may be creating opportunity, but only for businesses with real moats beyond code and meaningful AI adoption. Using patent-based disruption mapping and intangible-value factors, Kai shows that traditional value investing fails in exposed industries, while a framework that incorporates brand, IP, human capital, and network effects better separates winners from value traps during technological disruption.

Main Topics: Software Valuations and the AI Selloff (Priority: 5/5): The conversation opens with the unusual reality that software stocks now trade at a discount to the market, reflecting fear that AI will disrupt software economics. What Makes a Value Trap (Priority: 5/5): Kai defines value traps as companies that look cheap because their earnings are headed toward zero due to disruption, using Blockbuster, Borders, RadioShack, and McClatchy as examples. Measuring Disruption Systematically (Priority: 5/5): He explains a two-step NLP and patents-based framework to identify disruptive technologies and measure industry exposure in real time. Why Traditional Value Broke Down (Priority: 5/5): Traditional value works in insulated industries but fails in exposed ones, especially after 2010, because disruption hurts cheap companies in vulnerable sectors. Intangible Moats and Complementary Assets (Priority: 5/5): Drawing on David Teece, Kai argues that the firms that win disruption often possess complementary assets such as brand, distribution, human capital, and network effects. AI Adoption Plus Moats as the Winning Lens (Priority: 4/5): For software, the best outcomes come from firms that both adopt AI aggressively and have strong intangible moats; this combination helps distinguish survivors from losers. High Dispersion Creates Opportunity (Priority: 4/5): Disruption-era software stocks show extreme return dispersion, which increases the payoff to stock picking if an investor has a real edge.

Key Arguments: Software is trading at historically low relative valuations, even at a discount to the market, but cheapness alone is not enough to justify buying. Value traps occur when prices fall faster than fundamentals because markets anticipate disruption before reported earnings and sales fully collapse. A patents-and-NLP framework can identify disruptive waves and quantify which industries are exposed, allowing a more systematic analysis than anecdotes. Traditional value factors still work in insulated industries, but they fail in exposed sectors because disruption overwhelms any cheapness signal. The deterioration of value investing since around 2010 is explained largely by the rising share of the market exposed to technological disruption. Companies survive disruption not just by adopting the new technology, but by leveraging complementary assets such as brand, distribution, human capital, and network effects. Intangible-value factors capture these moats better than book-value/earnings-based value metrics and therefore work more consistently across time and sectors. Software companies are unusually aggressive AI adopters, which suggests they recognize the threat and may be able to use AI to improve margins if they survive. High dispersion in disrupted sectors means better opportunity for skilled stock pickers because winners and losers separate more sharply than in normal markets.

Data Points: Software forward P/E relative to S&P 500: 10% discount - Current valuation basis for software stocks after the selloff, described as unprecedented in the sample. Historical average software valuation premium: 32% premium - Average forward P/E premium of software versus the S&P 500 over roughly 20 years. Market-cap share exposed to innovation: 40% to mid-70s (~75%) - US market cap exposed to technological innovation increased materially over the past 20 years. Names basis exposure share: 72% vs. 78% - Exposure measured by number of stocks rather than market cap, showing similar upward trend. Traditional value spread in exposed vs insulated industries: -7 percentage points per year - Baseline spread showing value underperformance in exposed sectors relative to insulated sectors. Intangible value factor return in agreed-cheap names: 4.2% annualized - Full-sample performance for stocks that were cheap on both traditional and intangible value metrics. Intangible value factor return in agreed-expensive names: -5.1% annualized - Full-sample performance for stocks that were expensive on both metrics. Return for expensive-on-traditional but cheap-on-intangible names: 2.8% annualized - Examples like Amazon and Apple performed well despite appearing expensive on traditional value. Return for cheap-on-traditional but expensive-on-intangible names: -1.6% annualized - Value-trap bucket containing examples like Macy's. Value factor performance in exposed vs insulated sectors: Worked in insulated; failed in exposed - Core empirical result explaining the post-2010 deterioration of traditional value. Software stocks down in the setup chart: 30% or more over one year - Subset of software names used to study AI-disruption fear and valuation dispersion. Disruption-scare stock upside frequency: 10% double over next year - Historically, beaten-down exposed stocks had a much fatter right tail than the overall market. Disruption-scare stock downside frequency: 60% lose more than half - Historically, these stocks also had a much fatter left tail than the overall market.

Pivotal Quotes: "The code is not the moat." — Kai: Central takeaway for software investors: technology alone is not enough; complementary assets matter. "Value investing is not dead, it's maybe just being disrupted." — Kai: Explains why traditional value signals fail when applied broadly to industries facing technological change. "The biggest cost center for these companies, right? It is the production of code." — Kai: Why AI could potentially improve margins for software survivors by lowering software labor intensity.

Implications: Investors should stop treating low valuation as a buy signal in disrupted sectors and instead assess AI adoption plus intangible moats. In software, winners will likely be firms that combine real defensibility with effective AI use; losers may look cheap for a long time before collapsing.