by Editorial Team, AdvisorAnalyst
The instinct is understandable. Software stocks have been obliterated. Names like GoDaddy, Adobe, Workday, Atlassian, and Salesforce are trading at forward P/E ratios below 13.5 — roughly half the market multiple. For value-oriented advisors trained to buy quality on the cheap, the setup looks almost textbook.
Kai Wu thinks that instinct could be dangerous.
In his May 2026 paper AI Disruption: Moats and Value Traps1, Sparkline Capital's Founder and CIO delivers a methodical, data-dense argument that what looks like a blanket bargain is actually a dispersion-rich stockpicking environment — one where the wrong framework doesn't just underperform, it destroys capital.
The Ghost of Blockbuster
Wu's opening salvo is historical. He invokes Blockbuster, Borders, RadioShack, and McClatchy — firms that appeared statistically cheap as disruption approached, only to collapse into bankruptcy. The trap wasn't ignorance. It was a measurement problem.
"Falling prices relative to lagged fundamentals created the illusion of cheapness — luring value investors onto sinking ships."
This is the value trap mechanism in its purest form: price collapses fast, revenue erodes slowly, and the ratio looks attractive precisely when the underlying business is terminal. Wu argues the same dynamic is live in software today, with AI playing the role that streaming played for Blockbuster.
Value Investing Is Short Disruption
The paper's most striking empirical contribution is its dissection of the value factor itself. Using a disruption classification system built on patent clustering and machine learning — applied to a century of data — Wu separates industries into "exposed" and "insulated" segments and runs the standard value factor through each.
The result is unambiguous. In insulated sectors, traditional value has worked consistently for thirty years. In exposed sectors, it has been actively destructive, particularly since 2010. The conclusion Wu draws is blunt: "Value may not be dead — but it is being disrupted."
The failure is structural. Traditional metrics like P/E, P/B, and P/FCF systematically miss the assets that actually determine survival in technological disruption. They sold Amazon. They bought Borders. The factor's long and short legs both contributed to losses — a rare and damning finding.
The Intangibles Framework
Wu's proposed solution draws on David Teece's 1986 paper Profiting from Technological Innovation, which observed that innovating firms often fail to capture the returns from their own inventions. The real scarce assets, Teece argued, are the complementary capabilities surrounding the core technology — distribution, brand, human capital, regulatory credibility, network effects.
Applied to AI, this reframes the investment question entirely. The issue isn't whether a software firm can be replaced by cheaper code. It's whether that firm possesses the complementary moat that makes replacement economically and practically unattractive for its customers.
Wu's intangible value framework operationalizes this across four pillars: intellectual property, brand equity, human capital, and network effects. Unlike traditional value, intangible value has produced consistent excess returns in both exposed and insulated industries since 1995 — and critically, it has avoided the post-2010 drawdown that crippled conventional factor strategies.
"Unlike traditional value, intangible value also proved an effective way to pick stocks in exposed industries."
The Software Selloff Through This Lens
Applied to the current moment, the framework produces a nuanced verdict. Among software stocks down 30% or more over the past year, Wu finds the average intangible value score is positive — genuine opportunity exists. But he also finds "an abnormally long left tail of value traps — stocks, such as Duolingo, which appear unattractive even after huge drawdowns. Many of these stocks look cheap on traditional multiples but are screened out on intangible value."
This is the key distinction for advisors. The sector is not uniformly cheap. It is disperse. The stocks worth owning share two characteristics: strong complementary moats — trusted brand, embedded workflows, switching costs, trained professional ecosystems — and active AI adoption. Wu identifies Salesforce as an example of a name scoring well on both dimensions.
Critically, Wu argues that for enterprise software incumbents, the moat was never primarily the code. "Even before AI coding, customers had the option to build software in-house or switch to cheaper startups with more modern infrastructure — yet most did not. Rather than reinvent the wheel, they chose to outsource to trusted partners."
Disruption's Rising Tide
Wu widens the lens. AI, he notes, is a general-purpose technology. His data shows that over 72% of U.S. companies by count — representing 78% of market capitalization — now face disruption. The traditional value investor's refuge in insulated sectors is rapidly disappearing. Even U.S. small-cap stocks show 68% of market cap exposed.
"Soon, value investors will have nowhere to hide from disruption's all-seeing gaze."
Dispersion Is the Opportunity
Perhaps the paper's most actionable insight is its treatment of what Wu calls "Disruption Scare Stocks" — technologically exposed names down 30% or more. Historically, these stocks produce median forward returns in line with the broader market. They are not systematically doomed. But their return distribution has dramatically fatter tails: 10% double over the next year, while 16% are cut in half. Interquartile range runs 1.8 times the broader market.
This is the setup. "Disruption reshuffles the competitive landscape, producing a wider spread between winners and losers. Higher dispersion can be a boon for investors, as it enhances opportunities for stockpicking alpha."
Wu's backtests show the intangible value factor produced significantly higher absolute returns within Disruption Scare Stocks precisely because of this amplification effect — similar Sharpe ratios, far larger magnitudes.
Key Takeaways for Advisors
Traditional value screens are unreliable in disruption. P/E and P/B ratios systematically misprice companies facing technological change. Using them as a primary signal in software today invites the Blockbuster outcome.
The moat question matters more than the multiple. Brand equity, switching costs, embedded workflows, and trained professional ecosystems are the durable assets. Code is increasingly commoditized; complementary assets are not.
Dispersion is the edge. The software selloff is not a sector call — it is a stockpicking environment. The same dynamics that are destroying some names are creating durable entry points in others.
AI adoption is not optional. Firms actively integrating AI into their platforms are not just defending their moats — they may be expanding them, with meaningful long-term margin implications from reduced software development costs.
The disruption tide is rising everywhere. Advisors should assume most of their equity exposure now sits in "exposed" territory. Frameworks built for an insulated world will increasingly fail.
Wu's conclusion is measured but direct. The software selloff is real, the risks are real, and some of these names are genuinely cheap. But separating the survivors from the value traps requires moving beyond the metrics that built conventional value investing — and toward a definition of cheapness that accounts for the intangible assets disruption actually rewards.
Footnote:
1 Wu, Kai. "AI Disruption: Moats and Value Traps." Sparkline Capital, May 2026
