A new working paper from National Bureau of Economic Research (NBER) finds that investors are demanding a hefty premium to hold shares of companies most exposed to the AI transition thereby showing evidence that markets view artificial intelligence as a source of disruptive risk to be compensated for, and not a guaranteed windfall.
According to the study titled “AI Premium”, investors are demanding a significant premium to hold shares of companies most exposed to AI.
The researchers Aleh Tsyvinski from Yale University, Nicola Borri from Luiss University and Yukun Liu from University of Rochester, analysed 380 trillion AI tokens generated across more than 400 large language models using licensed data from OpenRouter, a platform that routes user requests to models from providers including OpenAI, Anthropic, Google and Meta.
The study, circulated by the National Bureau of Economic Research (NBER) was published in July 2026 and is yet to be peer-reviewed. Using the dataset by OpenRouter covering January 2024-April 2026, represents roughly 2 percent of global monthly AI token compensation. The researchers built a weekly index of worldwide AI usage to measure how strongly stocks of AI companies moved with shocks to that index.
According to the study, companies whose returns swung most sharply with AI consumption earned 64.1 basis points per week more than the least sensitive firms but the researchers caution against reading this as proof that AI exposure makes businesses stronger.
Instead, they attribute the gap to a “transition-risk” channel: new technologies reallocate value across firms, sectors and workers in unpredictable ways, and investors require higher expected returns as compensation for holding the companies most tied to that upheaval.
“Our result that the AI premium is positive and quantitatively sizable is most consistent with the argument that when this reallocation is systematic, investors require compensation for holding the firms most exposed to it,” read an excerpt form the report.
Tellingly, the most AI-sensitive firms in the S&P 500 are not chipmakers or software giants but names like AppLovin, Carvana and Expedia while several semiconductor firms, including AMD and Qualcomm, sit among the most negatively exposed.
The researchers created a new measure called the AI Factor to track growth in global AI usage. They then estimated each company’s AI Beta, which measures how closely its stock price moved with changes in AI activity and finally they examined the AI Premium, whether investors rewarded companies that were more exposed to AI with higher stock returns.
The study disclosed that firms with higher AI Betas earned higher subsequent returns than firms with lower AI Betas. A value-weighted portfolio of high-exposure firms outperformed a comparable portfolio of low exposure firms by 64.1 basis points per week.
“We also show that the AI premium is larger in regions where AI development and adoption is more advanced—about 18 basis points per week across developed markets, such as Europe, but absent in emerging markets, including China,” the researchers claimed in the study.
The study also examined how AI could affect different jobs across the US economy. It found that occupations requiring communication, persuasion, teaching and the installation and repair of new systems were more closely linked to AI. In contrast, jobs that depend mainly on analytical, scientific or operations control skills showed lower AI exposure. The researchers noted that occupations with stronger communication and interaction skills had higher market implied AI exposure.
“We build a market-implied mapping of AI exposure onto the tasks and skills across the US economy. We show that the nonroutine, interactive work—communication, persuasion, instruction, and the installation and repair of new systems—loads more on the AI factor, and analytical, scientific, and operations-control skills load less,” the researchers mentioned.
Also Read: AI Creating Two-Track Labor Market, AI ‘Professionals’ Get 42% Higher Wage
