AI Usage Index Backtest Yields 0.64% Weekly Spread Between High and Low Exposure Groups

Backtest of AI usage-linked stock selection shows 0.64% weekly spread between high and low exposure groups, remaining significant after controlling for size and valuation factors.

Woofun AI data shows that three economists conducted a backtest linking actual AI usage to U.S. stock market performance from January 2024 to April 2026, utilizing approximately 380 trillion tokens on OpenRouter. The researchers created an "AI Usage Index" based on weekly token usage, expenditure, and active user growth, then categorized stocks into high-exposure and low-exposure groups based on their price reactions over a 13-week period.

The strategy involved going long on the high-exposure group and shorting the low-exposure group, resulting in an average weekly spread of 0.641 percentage points. After controlling for company size, valuation, profitability, and momentum, the spread remained at approximately 0.56 percentage points.

However, the sample covers only 28 months, and OpenRouter represents a small, developer-centric fraction of global AI usage, leaving long-term effectiveness unvalidated.