AI Disruptees (May 2026)
AI Summary
📖 Deep Analysis
Theme and Background
This chapter explores the potential threat posed by artificial intelligence (AI) technology to the "moats" of information and data analytics companies. Market concerns are widespread that AI will disrupt business models built on proprietary data barriers, but the report argues these fears may be overstated. Historical data shows that investors have performed poorly in predicting industry winners and losers. For example, at the peak of the TMT bubble in 2000, investment bank Robertson Stephens ran a full-page ad in The Wall Street Journal declaring "Thanks, Old Economy. We'll take it from here," only to see the market turn immediately afterward. The "everything bubble" of 2020-21 similarly confirmed the failure of market pricing.
Core Arguments
1. AI has a limited impact on information companies with proprietary data moats and may instead enhance the value of data (as AI needs high-quality data to optimize models).
2. The market's panic over "AI losers" is excessive, creating opportunities for long-term investors to buy high-quality assets at a discount.
3. The risk of AI replacing online classified platforms is overestimated; consumer behavioral inertia, proprietary data, and embedded workflow tools constitute real barriers.
4. Seeking mispriced targets among so-called "losers" is more likely to generate excess returns than chasing AI winners.
Key Arguments and Data
- Deficiencies in market forecasting ability:
An Economist study examined the performance of 80 industries entering bear markets (relative decline >20% over three months) between 2005 and 2026. Share prices continued to fall only 50% of the time, making predictions no better than a coin toss.
- Resilience of proprietary data companies:
- Experian's credit data is delivered through machine-to-machine digital channels, making it less vulnerable to changes in per-seat pricing models.
- RELX's Risk division integrates diverse data (insurance risk, identity theft, cybersecurity), and AI tools enhance the value of data insights.
- RELX's Legal division (13% of profit) is more exposed to large language model (LLM) disruption because case law is publicly available, but AI startups still rely on RELX's LexisNexis and Thomson WestLaw as "systems of record."
- Latest results for the Legal division: sales growth of 9%, profit growth of 12%.
- Legal AI companies like Harvey and Anthropic's plugins are competitors, but LexisNexis's addressable market is $25 billion (legal tech products and services), while the core legal research market is only $5 billion.
- Competitive landscape for online classified platforms:
- Only 0.5% of Rightmove's searches come from LLM sources (3.5 years after ChatGPT's launch).
- UK app downloads and traffic have recently stabilized.
- Auto Trader's share price has fallen 45% over the past 12 months, but the report notes that used cars suffer from information asymmetry (Akerlof's "lemons market" theory). LLMs' probabilistic, unstructured processing struggles to replace Auto Trader's structured data system (vehicle models, configurations, technical specifications, etc.).
Companies/Assets Involved
| Company/Asset | Role | Key Data | Bullish/Bearish |
|---|---|---|---|
| RELX | Holding (Marathon portfolio) | Risk division integrates diverse data; Legal division accounts for 13% of profit, sales +9%, profit +12%. | Bullish: AI enhances data value, Legal disruption risk limited (still needs its data). |
| Experian | Holding (Marathon portfolio) | Credit data delivered machine-to-machine; Decision Analytics segment accounts for 21% of revenue. | Bullish: Data format and delivery model less impacted by AI. |
| Rightmove | Analyzed (not held) | LLM-sourced searches <0.5%; App downloads and traffic recently stable. | Defensive: Consumer behavioral inertia, proprietary data. |
| Scout24 | Analyzed (not held) | German real estate portal; seller data valuable to brokers. | Defensive: Similar to Rightmove. |
| Auto Trader | Holding (Marathon portfolio) | Share price -45% (12 months); Company embedded in dealer workflows (valuation, inventory, CRM). | Bullish: Information asymmetry and structured data advantage undervalued. |
| SAP | Not held | Index heavyweight; software threatened by AI. | Avoid. |
| Thomson WestLaw | Competitor | RELX's main competitor in the Legal segment. | Neutral. |
Investment Implications
1. Increase exposure to mispriced "AI losers": The report explicitly states that Marathon has recently added to positions in the information companies and online classified platforms mentioned above, as current valuations imply excessively pessimistic expectations.
2. Avoid the pure AI-winner bubble: While global equity markets are being driven to new highs by a handful of AI-themed stocks, the report argues that winners are already fully priced, offering poor risk-reward.
3. Focus on structural data barriers: Machine-to-machine delivery, structured databases, and embedded workflows are key characteristics that resist LLM replacement; companies relying on purely public data (e.g., legal case law) face higher risk.
4. Be patient for catalysts: The report criticizes strategies of "waiting until uncertainty is resolved before buying," arguing that markets price in catalysts quickly, making it more effective to buy at a discount and hold for the long term.