Nathaniel Whittemore reviews an Exponential View report that aggregates sources on AI company revenues and aims to avoid double counting spending across the supply chain. He presents the report’s estimates and growth comparisons as its analysis, not independently audited figures.
The episode discusses capital expenditure on computing infrastructureAI compute infrastructure is the hardware, facilities, networks, power, cooling, storage, and software used to train and run AI models., the relationship between revenue and depreciation, and the report’s argument that some older hardware continues to generate value. It notes that revenue covering current expense does not settle the question of recovering cumulative investment.
Whittemore also covers falling token prices alongside growing usageAI inference economics examines the revenue, cost, utilization, pricing, and margins involved in serving trained models., shifts in value between infrastructure and applications, and reported differences in company revenue growth. Earlier headlines include proposed agent regulationGovernance sets responsibilities, rules and oversight for decisions about how AI systems are used., model access, GPU pricing and memory costs; proposals and reported negotiations remain attributed.
Watch on YouTube
