The market has stopped buying into vague promises and started rewarding companies for the actual "mileage" of their neural networks. A study led by Yale economist Aleh Tsyvinski confirms that the intensity of AI token consumption now correlates directly with stock prices. The performance gap between aggressive AI adopters and conservative laggards has reached an impressive 0.64% per week. On an annualized basis, this translates into a chasm that is impossible for investors to ignore.
This phenomenon, dubbed the "AI Premium," is no mere speculation. It is based on an analysis of a massive dataset comprising 380 trillion tokens processed through the OpenRouter platform. The researchers, including Nicola Borri from Luiss University and analysts from the University of Rochester, examined data extending through early 2026 (incorporating projected usage models), covering millions of users and over 400 models—from GPT and Claude to DeepSeek.
Key Research Findings
The AI effect is no longer a private party for Big Tech. Significant market cap growth is now being recorded in the consumer sector, retail, and capital-intensive manufacturing. Investors are no longer waiting for a "revolution"; they are evaluating specific "exposure"—the depth to which algorithms have integrated into a company's core business processes. The premium for AI implementation has become a stable market factor, comparable in significance to traditional metrics like liquidity and risk profiles.
We are witnessing a shift from evaluating marketing slogans to auditing actual compute power. The main question remains: when will investors stop buying expectations and demand hard currency in quarterly reports?
In our view, current market valuations represent a massive down payment. Capitalization is growing not because of immediate profits, but based on expectations of future productivity gains. For now, the market is willing to pay for tokens, but the line of credit for "expected efficiency" has its limits.