The emergence of reasoning models like o1 marks a shift away from the zero-marginal-cost economics that defined the internet era. Since the 2010s, hyperscalers have relied on Aggregation Theory, where scaling products to billions of users incurred minimal incremental expense. Today, the high cost of compute is making technology capital-intensive again.
As Doug O’Laughlin observes, the primary limit to AI improvement is now economic rather than technical. This transition forces a reckoning for business models built on internet-scale distribution. Read the full story at Stratechery (subscription).
Source · Stratechery
