AI Tokens as a Measure of the New Industrial Age

Sam Altman of OpenAI has proposed a future where intelligence is purchased much like electricity or water. Many AI companies are embracing this vision, potentially making ‘AI tokens’ a regular part of our daily interactions. These tokens could become a crucial measure of AI usage, similar to kilowatt-hours for electricity. Every time an AI processes a prompt or completes a task, it involves tokens, which are small data pieces that represent the model’s workload. More tasks mean more tokens processed.

Currently, AI companies mostly offer flat-rate subscriptions for regular consumers but are increasingly charging businesses based on token usage. As businesses, especially in tech, ramp up AI use, they encounter the high costs of token-based pricing. Companies like Uber and Amazon are now monitoring AI token usage closely to manage expenses.

Beyond usage measurement, AI tokens provide a digital paper trail. Economists and researchers are leveraging this data to study AI’s economic impact. A new study by Nicola Borri, Aleh Tsyvinski, and Yukun Liu analyzes data from roughly 380 trillion AI tokens to explore how AI growth affects financial markets. They examine which companies’ stock prices rise or fall as AI consumption changes over time.

The study suggests that stock prices tend to reflect investor expectations about the future. Although not always accurate—like past financial bubbles—these prices offer a glimpse into market predictions regarding AI’s impact. These economists found a link between AI token usage and stock returns, particularly for companies perceived as AI beneficiaries. Interestingly, the findings indicate that the AI advantage is not restricted to tech firms. Investors foresee a diverse range of industries benefiting from AI advancements.

This study is notable not only for its findings but also for its method. Instead of relying on traditional sources like surveys or earnings calls, researchers can now use AI tokens to track usage and economic effects with unprecedented precision.

One significant source of token data is OpenRouter, a platform allowing access to many AI models via a single interface. This platform facilitates businesses and developers to compare prices and manage costs effectively. OpenRouter’s anonymized data shows trends in AI token usage, covering about 2 percent of global AI usage from January 2024 to April 2026. Growth in tokens, spending, and active users forms an ‘AI Factor’ used to understand stock return sensitivities.

The study reveals that companies with stock prices sensitive to AI token consumption earn higher returns. This ‘AI Premium’ is applied not just to tech stocks but also across various industries, demonstrating a broad economic impact expectation. However, results should be approached cautiously, as financial markets aren’t always correct.

Some specific findings suggest that companies like AppLovin and Carvana benefit the most from AI, while others such as Moderna and Estée Lauder are less favored. However, these insights come with caveats. The working paper isn’t peer-reviewed yet, and OpenRouter’s data might not represent average consumers.

Though the analysis suggests a new way to gauge AI’s economic spread, it isn’t meant as financial guidance. It does offer a promising approach for economists to quantify AI’s influence, paving the way for data-driven insights into AI’s growing role in the economy.