Original title: AI Premium
Nicola Borri, Yukun Liu, Aleh Tsyvinski · June 2026
Investors want to know how real-world artificial intelligence use affects the economy. To find out, researchers tracked actual AI activity instead of just market hype. They analyzed user activity, spending, and data flow across hundreds of models. Then, they compared these trends directly to stock market movements to measure how tied different companies are to AI growth. They found that companies whose stock prices move strongly with actual AI adoption earn higher subsequent investment returns. This extra return is biggest for companies connected to serious, advanced use, like long prompts, paid accounts, and closed models, rather than casual use. This payoff also spreads into consumer and capital-heavy businesses, but it does not appear in emerging markets. Finally, jobs focused on communication and interacting with people show more positive market exposure to AI, while analytical and scientific jobs show negative exposure. The team also found early signs of an agentic economy.
You are a farmer deciding which crops to plant. You watch which crops have sold well, but the market keeps shifting depending on which crop you choose, so last year's prices tell you nothing about this year's. In this paper, the crop is a company's stock, and the shifting market is AI usage. The researchers found that companies whose stocks rise and fall with AI usage earn higher returns — but only when the AI use is serious, paid, and intensive, not casual.
The AI Premium puts a number on AI's economic impact. It shows which companies the market thinks will benefit — and which workers are seen as exposed. Interactive jobs are priced as more AI-exposed, while analytical and operations jobs are not.
The abstract doesn't say whether the premium survives trading costs, how long it persists, or whether AI exposure causes higher returns rather than reflecting risk. It also doesn't say how representative OpenRouter's licensed data is, beyond covering about 2% of global monthly AI token consumption.
Ask anything about this paper. Answers come from its abstract, and say so when the abstract doesn't cover your question.
AI Premium — Nicola Borri, Yukun Liu, Aleh Tsyvinski
Abstract licence: CC0
Explanation generated by Synthesis from the abstract — it is not the original paper. This is a preprint, so it hasn't necessarily been peer reviewed.