AI tokens — the small units of text and other data that AI models read and generate — could become a standard tool for measuring the spread of artificial intelligence through the economy, according to a new working paper from economists Nicola Borri, Aleh Tsyvinski, and Yukun Liu.

In a study that analyzed 380 trillion AI tokens from OpenRouter, a platform that gives developers access to hundreds of AI models through a single interface, the researchers combined weekly growth in tokens, spending, and active users into a broad measure they call the “AI Factor.” They then estimated which S&P 500 companies’ stock returns move most strongly with changes in that factor.

The economists found that companies whose stock prices were most sensitive to increases in overall AI consumption subsequently earned significantly higher returns. Firms that Wall Street appeared to view as the biggest AI beneficiaries outperformed those seen as least likely to benefit by about 0.64 percentage points per week — a pattern the paper describes as an “AI premium.”

Tsyvinski says that the findings suggest investors expect AI to benefit a wide range of companies across the economy. “The story of AI is no longer just a Silicon Valley story,” Tsyvinski says. “Financial markets already see Main Street being impacted.”

The top five S&P 500 companies identified in the paper as having high AI premiums are AppLovin, Carvana, Lumentum, Expand Energy, and Baker Hughes. The bottom five — companies Wall Street appears to view as AI’s biggest losers — are Moderna, Estée Lauder Companies, ON Semiconductor, Skyworks Solutions, and Aptiv.

The economists found that the AI premium is strongest for companies in the United States and Europe and much weaker in China and other emerging markets. They also found that stocks are most sensitive to increased use of “frontier” AI models — the most technologically advanced systems.

The analysis relies on data from OpenRouter, which has amassed rich information on AI token usage as businesses and workers have turned to the platform to compare prices and manage their token spending. The researchers said OpenRouter’s data represents approximately 2 percent of monthly global AI usage and is anonymized to protect user identities.

OpenAI CEO Sam Altman declared earlier this year that the company sees AI becoming a metered utility. “We see a future where intelligence is a utility, like electricity or water, and people buy it from us on a meter,” Altman said.

AI companies increasingly charge businesses and developers based on the number of tokens they use, according to the report. Companies such as Uber and Amazon have responded to soaring token bills by placing guardrails around AI use.

The economists cautioned that their findings come with several caveats. The paper has not yet been peer reviewed. OpenRouter’s token data likely provides a skewed picture of AI usage, the researchers said, because its users are sophisticated, heavy AI users shopping between models to get the best value rather than average consumers with monthly subscriptions to services like ChatGPT, Claude, or Gemini.

Even if the researchers have correctly identified the companies investors expect to benefit from AI, the paper noted, those expectations may not prove correct, and much of that optimism may already be reflected in current stock prices.

The broader significance of the research may lie less in its stock market findings than in the data source it uses. The economists said AI tokens could allow researchers to track AI usage in near real time and study its economic effects with precision not possible during past technological revolutions, offering a new way to measure the spread of AI through the economy.