Thinking Fast & Slow by Daniel Kahneman | Psychological Biases Explained Simply (TIP793)

The software sector is experiencing a "bloodbath," with major names like Adobe and Salesforce down significantly. Institutional flows have rotated heavily out of software and into AI hardware/momentum trades. Investors are exhibiting "availability bias" by chasing the hot AI hardware narrative and dumping software due to fear. This creates a dislocation where profitable, entrenched software companies are trading at distressed valuations simply because they are currently "unloved." Contrarian opportunity to acquire high-quality software businesses while the market is distracted by the AI hardware momentum trade. The "AI disruption" thesis for broad horizontal software (like Adobe/Salesforce) might be more valid than for vertical software, leading to genuine value destruction.

Institutional capital is flowing aggressively out of software and into AI hardware/semiconductor names, driven by momentum. The speaker warns against "appealing fictions"—stories that investors desperately want to be true to justify high valuations. He cites Buffett’s warning about the 1999 tech bubble, implying the current frenzy in AI hardware might be decoupling from base rates and probability. While not an explicit short, the commentary suggests extreme caution regarding the "hot" sectors (AI Hardware) where valuations are driven by stories rather than probabilities. The AI hardware boom could continue longer than rational analysis suggests (irrational exuberance).

Constellation Software shares on the TSX are down over 50% from their May 2025 highs. The drop is driven by founder Mark Leonard stepping down and fears that AI will disrupt the software industry. The market is reacting emotionally (System 1 thinking) to the narrative of AI disruption and the leadership change. However, the new CEO, Mark Miller, is a developer-turned-investor with a 30-year track record within the company. The business model relies on "mission-critical" VMS with high switching costs, which are unlikely to be easily displaced by AI. In fact, AI may lower development costs for incumbents. The stock is now trading at a "high teens multiple" of 2026 earnings, which is viewed as compelling for a compounder of this quality. The thesis relies on the durability of the VMS moat despite the AI narrative. Mark Miller fails to replicate Leonard's capital allocation success; AI disruption in VMS proves to be structural rather than just a narrative fear; management has not yet stepped in to buy shares aggressively despite the price drop.