OpenAI CEO Sam Altman has stated a vision of artificial intelligence functioning as a utility that businesses and consumers purchase through metered usage. This concept has prompted considerable interest across the AI industry, with tokens potentially emerging as the primary measurement unit for AI consumption, analogous to kilowatt-hours in the electricity sector.
Tokens represent the fundamental units of processing in AI systems, encompassing small chunks of text and data that models process when responding to prompts or completing tasks. While consumer-facing AI services typically operate on subscription models, many businesses face token-based pricing structures tied to their actual usage. As corporate AI adoption has accelerated across multiple sectors, expenses associated with token consumption have risen substantially. Following an initial period of unconstrained AI experimentation within technology companies, organizations like Uber and Amazon have implemented measures to control spending and optimize their token usage patterns.
A recent working paper by economists Nicola Borri, Aleh Tsyvinski, and Yukun Liu examines 380 trillion tokens processed between January 2024 and April 2026 to assess how AI consumption affects financial markets. The researchers analyzed data from OpenRouter, a platform aggregating multiple AI models, and identified which stock prices correlate with fluctuations in overall AI usage. They discovered an “AI Premium” whereby companies perceived as major AI beneficiaries generate higher returns than those viewed as potential losers, with a differential of approximately 0.64 percentage points weekly.
The analysis reveals that the AI Premium extends beyond technology companies, with financial markets appearing to anticipate AI benefits across diverse industries including airlines, utilities, manufacturers, retailers, banking, and waste management. The effect proves strongest among United States and European equities and correlates particularly with usage of advanced frontier AI models. The researchers identified AppLovin, Carvana, and Lumentum as leading beneficiaries, while Moderna, Estée Lauder, and ON Semiconductor ranked among anticipated losers.
Significant limitations qualify these findings. The research remains a non-peer-reviewed working paper, and OpenRouter data represents only sophisticated heavy users optimizing costs rather than typical consumers with standard subscriptions. While the study suggests token usage could provide economists unprecedented real-time data on AI adoption, matching actual stock price movements to AI benefits remains speculative, and current valuations may already incorporate market expectations about AI’s economic impact.
Article Attribution | Read More at Article Source
Article summary produced by Claude AI