According to CoinMeta and a report by OneMillion.AI, three economists conducted a backtest using real AI usage data in conjunction with the US stock market trends. The study covered the period from January 2024 to April 2026 and utilized approximately 380 trillion token units. The researchers compiled a “AI Usage Index” by combining weekly token usage, consumption amounts, and growth rates of active users, to observe the stock price responses of each US stock over the past 13 weeks. The results indicated that stocks that tended to outperform the market when AI usage accelerated and underperformed when it slowed down were categorized as the high-exposure group. The backtest revealed an average difference of 0.641 percentage points per week between the two groups; even after controlling for factors such as company size, valuation, profitability, and stock price momentum, there was still a difference of about 0.56 percentage points. With a sample period of only 28 months and the data covering only a small portion of global AI usage, and given that users were predominantly developers, the long-term effectiveness of this method remains to be verified.