Wanan Yossingkum/Getty Images
Earlier this year, OpenAI CEO Sam Altman said: “We see a future where intelligence is a utility, like electricity or water, and people buy it from us through a meter.”
Many other AI companies seem to be betting on a similar future. If this vision actually becomes a reality, “AI tokens” could break out of the world of nerdy tech and business talk and become a much more familiar part of our lives. The number of AI tokens used by businesses and consumers could even become one of the defining metrics of a new industrial era – the AI equivalent of the kilowatt-hour for electricity: the standard way we measure and pay for AI usage.
Think of tokens as the counter that runs in the background every time you use AI. Every time an AI model reads your prompt, writes an answer, or completes a task, that work is measured in tokens. These are tiny blocks of text and other data that are read and generated by AI models. In general, the more work a model does, the more tokens it typically processes.
While AI companies still tend to offer flat-rate subscriptions to average consumers, they are increasingly charging companies and developers based on the number of tokens they use. At the same time, companies, especially in the technology sector, are increasingly using AI in their work. As their AI usage has increased significantly, many companies have realized how expensive token-based pricing can become.
After a period in which tech workers dabbled in a form of AI that everyone could use for their own benefit—what some called “tokenmaxxing”—companies like Uber and Amazon have limited their use of AI and reduced their skyrocketing token bills (as one astute writer noted). The information recently referred to as “token minimizing”).
But tokens aren’t just how AI companies measure usage and charge many of their business customers. They also leave a kind of digital paper trail that a growing number of economists and other researchers are using to track AI deployment and study its economic impact.
In a new working paper, Nicola Borri, Aleh Tsyvinski and Yukun Liu do essentially exactly that. Using data from 380 trillion Using AI tokens, these economists seek to understand how the increasing use of AI is changing financial markets. They ask a simple question: As overall AI consumption changes over time, which companies tend to see their stock prices rise – and which tend to fall?
To be clear, the economists don’t track how many AI tokens individual companies use – which, let’s face it, would be a lot cooler. Instead, they look at the rising tide of overall AI usage and ask which stocks are rising and falling with it.
Part of their analysis is classical financial theory. The idea is that stock prices tend to reflect investors’ future expectations, and so can provide insight into the companies that Wall Street believes will be most positively or negatively affected by the rise of AI.
Of course, investors can be very wrong. Hello, the long history of financial bubbles. However, if you want to know what the markets expect AI to do for the economy, stock prices are an interesting place to look.
Not surprisingly, economists are finding that as AI use has increased, financial markets have treated some companies very differently than others. The companies considered the biggest AI beneficiaries enjoyed higher stock returns – a pattern the researchers call the “AI premium.” What’s more interesting is that they find that it’s not just tech stocks that seem to deserve this premium. Their findings suggest that investors expect AI to benefit a broad range of companies and industries across the economy.
“The story of AI is no longer just a Silicon Valley story,” says Tsyvinski about the results of her work. “Financial markets are already seeing an impact on Main Street.”
The most exciting thing about this study is probably less its results than that it offers a glimpse into future research opportunities. Instead of relying on surveys, earnings calls, or company announcements, economists may be able to study the spread of AI by following the digital paper trail created by AI tokens. AI tokens could become a powerful new data source, allowing researchers to track AI usage in near real-time and study its economic impact with a precision not possible with previous technological revolutions.
Table of Contents
An exciting way to measure the AI economy
Borri, Tsyvinski and Liu’s analysis relies on a relatively new data source: OpenRouter.
OpenRouter is a kind of one-stop shop for AI models. Instead of signing separate contracts with OpenAI, Anthropic, Google and dozens of other AI companies, developers can access hundreds of models through a single interface. As companies and workers become more aware of their rising AI token bills, they are turning to platforms like OpenRouter to compare prices, upgrade to cheaper models for less complex tasks, and manage their token spending. In doing so, OpenRouter has collected extensive data on the use of AI tokens (and this data is anonymized to protect the identity of users).
The economists analyzed the use of 380 trillion AI tokens between January 2024 and April 2026. This corresponds to around 2 percent of monthly global AI usage. They then combine weekly growth in tokens, spending, and active users into a broad measure of AI consumption, which they call the “AI factor.” Next, they estimate which companies’ stock returns change the most when this factor changes.
The economists find that companies whose stock prices were most sensitive to an increase in overall AI consumption subsequently achieved significantly higher returns. The companies that Wall Street appears to view as the biggest beneficiaries of AI outperformed those deemed the least likely beneficiaries by about 0.64 percentage points per week – the “AI premium” described in the paper.
Of course, the markets aren’t always right. These AI-related stock moves could reflect the wisdom of the crowds — or simply a herd of enthusiastic investors hoping to profit from a technological revolution that ultimately disappoints. In other words, take these findings with a grain of salt.
Perhaps the most interesting of their findings is that the AI premium can be found far beyond the world of technology. Markets seem to believe that companies in industries ranging from airlines and cruise ships to utilities, industrial manufacturers, retailers, banks and even waste management companies could all benefit as AI transforms the economy. They also find that this “AI premium” is strongest for companies in the US and Europe and is significantly weaker in China and other emerging markets. Perhaps relatedly, they note that stocks are most sensitive to increased use of the most technologically advanced “frontier” AI models.
In their paper, the economists even provide a list of S&P 500 companies with high AI premiums. The top 5 are: AppLovin, Carvana, Lumentum, Expand Energy and Baker Hughes.
They also identify companies that Wall Street appears to view as the biggest AI losers. The final five are: Moderna, Estée Lauder Companies, ON Semiconductor, Skyworks Solutions and Aptiv.
Their analysis has a number of caveats. First, it is a working paper that still needs to be peer-reviewed. Second, OpenRouter’s token data likely provides a distorted picture of AI usage. People using OpenRouter are likely demanding, heavy AI users trying to save money by switching between competing models to get the most bang for their buck. Your data does not represent the type of average consumer who may have and continue to use a ChatGPT, Claude or Gemini monthly subscription.
Perhaps most importantly, even if researchers correctly identified the companies that investors expect will benefit from AI, that doesn’t mean those expectations will turn out to be correct. And if markets are reasonably efficient, much of that optimism could already be reflected in today’s stock prices. Therefore, buying the stocks that you identify as potential AI-related winners and selling the ones that could be AI-related losers is by no means a good financial strategy. This is not financial advice!
Still, the paper’s greatest contribution may not be what it says about today’s stock market. Perhaps it points to a new way for economists to measure the spread of AI in the economy. We welcome this era of AI measurement maxxing. There are still many big economic questions about AI, and better data is rarely a bad place to start.
https://www.npr.org/sections/planet-money/2026/07/28/g-s1-135475/ai-tokens-could-become-the-kilowatt-hour-of-the-ai-age
