Loading...
Three economists used approximately 38 trillion token data on OpenRouter to construct an AI usage index and conduct backtesting on US stocks. The study covers the period from January 2024 to April 2026, and the results show that the strategy of buying high exposure groups and shorting low exposure groups generates an average weekly yield difference of 0.641 percentage points. After controlling for relevant factors, the difference is still about 0.56 percentage points. This method is limited by the sample period and data coverage, and its long-term effectiveness needs to be verified.