As powerful AI models become cheap and universally accessible, having one is no longer a defensible moat. The real, lasting advantage for a business now comes from assets that a better model can't easily replace: proprietary customer data, deeply integrated user workflows that are difficult to replicate, and long-term client relationships.
Contrary to the belief that cheaper technology levels the playing field, it often benefits established companies most. These incumbents can leverage the now-cheaper tool to enhance their existing products and reach more customers. New startups built solely on the cheap tool often fail because they lack a unique business foundation and are easily copied.
Decreasing the cost per AI query (inference) paradoxically drives up total expenditure. As AI becomes more efficient and accessible, its usage skyrockets for new applications, increasing aggregate spending on infrastructure like chips and data centers. This mirrors the Jevons paradox, where more efficient steam engines increased, rather than decreased, Britain's coal consumption.