Five Cases in Creating and Destroying Value

Case studies are most useful when they recover the decisions and uncertainties behind an outcome. In AI the outcome often arrives faster than the evidence: a model can define a category within weeks and be overtaken within a year, and a company can be valued at a hundred times its sales before anyone has seen a full year of its gross margin. The task is to separate what decision-makers could have known from what happened later.

The five cases below were chosen to test all four factors of the intelligence equation, to balance value created against value captured or destroyed, and to span the stack from memory chips to consumer voice. SK Hynix shows a commodity supplier that found the one input the AI boom could not route around. Palantir shows capability that survived real work in the least forgiving customers there are, and a valuation that ran ahead of it. ElevenLabs shows a specialist that turned misuse and rights from liabilities into parts of its product. Stability AI shows an open model that created enormous value for others and almost none for itself. Zillow Offers shows what happens when a model good enough to inform consumers is trusted to commit a balance sheet. Figures come from the research companion, which records their sources and confidence. Interpretations are analytical judgments, not claims that one decision caused an outcome.