380 Trillion AI Tokens Show Markets Rewarding AI-Ready Firms

Financial markets are delivering measurably higher returns to companies positioned to benefit from artificial intelligence adoption, according to new research analyzing one of the largest datasets of real-world AI consumption ever studied.

Yale economist Aleh Tsyvinski and colleagues examined 380 trillion AI tokens—the fundamental units AI models process—from OpenRouter, a platform routing requests to more than 400 AI models including GPT, Claude, and Deepseek. The dataset, representing 2% of global monthly AI usage from January 2024 through April 2026, came from millions of anonymous users.

The research team found that companies with the highest AI exposure earned approximately 0.64% higher weekly stock returns compared to those with the lowest exposure—a financial advantage they term the "AI Premium."

Why it matters

This research moves beyond surveys and estimates to quantify AI's economic impact using actual consumption data at unprecedented scale. The findings suggest investors are already pricing in productivity gains from AI across multiple sectors, not just technology companies. The workforce implications are particularly striking: markets expect AI to create opportunity in communication-oriented roles while displacing routine analytical work, potentially reshaping career paths in science and healthcare.

Beyond the tech sector

The AI Premium extends well beyond Silicon Valley. Retail, consumer durables, and capital-intensive industries like manufacturing all show positive market responses to rising AI consumption. This breadth indicates investor confidence that AI-driven productivity gains will materialize across the economy.

The premium concentrates in the United States, Europe, and other developed markets closely connected to frontier AI development and infrastructure. China and emerging markets show less pronounced effects.

Sophisticated usage drives value

Not all AI consumption carries equal weight in market valuations. Investors place higher value on intensive use of proprietary, advanced models by experienced, paying customers using sophisticated prompts. Casual experimentation with free or open-source models matters less.

"Despite the widespread and fast adoption of AI tools by everyday users, the AI premium is mostly determined by the exposure to the frontier AI consumption by sophisticated and professional users," said Nicola Borri, associate professor of finance at Luiss University and study co-author.

Workforce implications

Combining stock market data with government labor statistics and AI consumption patterns, the researchers identified which occupations face the greatest disruption. Jobs involving non-routine tasks—persuasion, teaching, communication, and interpersonal interaction—show more positive exposure to AI consumption than routine analytical work.

Health care positions and roles requiring scientific analysis appear more vulnerable. "Occupations in science are among the most negatively affected, which may seem surprising," Tsyvinski noted, distinguishing between routine lab work and cutting-edge research at major universities.

The rise of agentic AI

The dataset captured a significant shift toward agentic AI—systems that autonomously complete tasks rather than simply answering questions. In 2024, agentic AI represented a small fraction of consumption. By 2026, it accounted for more than half of all AI tokens processed. Early evidence suggests exposure to this "agentic economy" is beginning to influence company valuations.

"Previous studies on the economic effects of AI have mostly relied on surveys and estimates, but our work is based on unprecedented amounts of real-world data, providing a much more granular view of how AI consumption is affecting the economy," said Tsyvinski, the Arthur M. Okun Professor of Economics at Yale.

The findings were first reported by Yale News. The study appears as a National Bureau of Economic Research working paper, co-authored by Borri and Yukun Liu, associate professor of finance at the University of Rochester.