AI "tokens" — the small chunks of text and data that AI models read and generate — are emerging as a standard unit for measuring artificial intelligence usage across the economy, much as the kilowatt-hour became the benchmark for electricity consumption. This matters because businesses increasingly pay for AI based on token consumption rather than flat fees, and economists are now using aggregated token data to work out which companies stand to gain or lose from the AI boom, offering a new lens on how AI is reshaping markets well beyond the tech sector.

Economists Nicola Borri, Aleh Tsyvinski and Yukun Liu analysed 380 trillion AI tokens generated between January 2024 and April 2026, finding that firms perceived as major AI beneficiaries earned notably higher stock returns, an "AI premium" of around 0.64 percentage points a week, strongest in US and European markets but weaker in China. The effect reaches well beyond tech giants, touching airlines, utilities, manufacturers, retailers and even waste-management firms; predicted winners include AppLovin, Carvana, Lumentum, Expand Energy and Baker Hughes, while Moderna, Estée Lauder, ON Semiconductor, Skyworks Solutions and Aptiv were flagged as likely laggards. Separately, companies such as Uber and Amazon have reportedly moved from unrestrained "tokenmaxxing" to cost-conscious "tokenminimizing" after AI expenses climbed sharply. The researchers caution that their data comes from OpenRouter, representing only around 2% of global AI usage, has not been peer-reviewed, and reflects the behaviour of sophisticated users rather than the average consumer, so the findings are indicative rather than definitive.