Earlier this year, Sam Altman, OpenAI’s CEO, envisioned a future where intelligence serves as a utility, akin to electricity, and individuals purchase it on a meter. This prediction has garnered considerable attention, potentially transforming AI tokens from niche topics to widespread consumer phenomena.
Understanding AI Tokens
AI usage is measured in tokens, representing text and data chunks read and generated by AI models. The higher the workload of a model, the more tokens it processes. While flat-rate subscriptions for average consumers persist, businesses face costs based on token consumption. Companies, especially in the tech sector, have seen their AI expenditures rise with increased usage, necessitating strategies like ‘tokenminimizing’—a method coined by a writer at The Information—employed by firms like Uber and Amazon.
AI Tokens and Economic Impact
Tokens not only measure AI usage, but their digital trail aids economists and researchers in analyzing the financial impact of AI. Nicola Borri, Aleh Tsyvinski, and Yukun Liu use data from 380 trillion AI tokens to explore how AI development influences stock markets. Their inquiry focuses on aligning stock prices with overall AI consumption changes.
“The stock market reflects investor expectations, providing insights into companies positively or negatively impacted by AI.”
Though stock prices provide a glimpse into expected economic shifts, investors’ predictions may not always be accurate, as evidenced by historical financial bubbles. Despite uncertainties, Borri, Tsyvinski, and Liu’s work shows how AI growth has influenced Wall Street’s perceptions. Companies perceived as AI beneficiaries exhibit a pattern termed ‘AI premium’ with increased stock returns, extending beyond tech to diverse industries impacted by AI.
New Data Source for Analysis
The study leverages OpenRouter, a platform centralizing hundreds of AI models, offering developers the ability to manage AI costs and analyze significant token usage data. Their analysis covers January 2024 to April 2026, incorporating data representing 2 percent of global AI usage. The economists combine weekly changes in tokens, spending, and active users into a metric dubbed ‘AI Factor,’ measuring overall AI consumption.
Intriguingly, the ‘AI premium’ appears strongest in the U.S. and Europe, with frontier AI models driving stock sensitivity. Industries ranging from airlines and utilities to financial institutions are deemed AI beneficiaries. Conversely, emerging markets show weaker AI premiums.
Cautionary Notes and Market Implications
AppLovin, Carvana, and other S&P 500 companies top lists of AI premium stocks, while Moderna and Estée Lauder are seen as potential AI losers. Despite its insightful perspective, this working paper, not yet peer-reviewed, presents data from sophisticated AI users via OpenRouter, not reflecting average consumer behavior. Moreover, market efficiency questions remain, cautioning against indiscriminate investment based on identified AI winners and losers.
The paper’s significant contribution may lie in its innovative approach to measuring AI’s market spread. While not financial advice, it highlights potential pathways for economists to explore AI’s economic integration. As we navigate this era of AI measurement-maxxing, enhancing data collection remains crucial in addressing broader economic questions surrounding AI.