AI Tokens Emerge as the New Kilowatt-Hour for Measuring AI Use

OpenAI CEO Sam Altman envisions a future where "intelligence is a utility, like electricity or water, and people buy it from us on a meter." That future may be closer than it appears—and AI tokens are positioned to become its fundamental unit of measurement.

Tokens represent the tiny chunks of text and data that AI models process when reading prompts, generating responses, or completing tasks. While consumer AI services typically charge flat monthly fees, businesses and developers increasingly pay based on token consumption. As companies integrate AI more deeply into their operations, many have discovered these token-based bills can balloon quickly.

The shift has sparked what some call "tokenmaxxing"—a period of unrestricted AI experimentation—followed by "tokenminimizing" as companies like Uber and Amazon impose guardrails to control costs. But tokens serve a purpose beyond billing: they create a digital trail that researchers can use to measure AI's economic footprint with unprecedented precision.

Tracking AI's Market Impact Through Token Data

A new working paper by economists Nicola Borri, Aleh Tsyvinski, and Yukun Liu demonstrates this potential. Using data from 380 trillion AI tokens processed through OpenRouter—a platform that lets developers access hundreds of AI models through one interface—the researchers examined which companies' stock prices move in sync with overall AI consumption growth.

Their analysis, covering January 2024 through April 2026, found that stocks most sensitive to AI usage increases earned significantly higher returns—about 0.64 percentage points per week more than companies viewed as less likely to benefit from AI. The researchers call this the "AI premium."

More surprisingly, this premium extends well beyond Silicon Valley. Financial markets appear to expect AI benefits across airlines, cruise lines, utilities, manufacturers, retailers, banks, and even waste management companies. "The story of AI is no longer just a Silicon Valley story," Tsyvinski noted. "Financial markets already see Main Street being impacted."

The researchers identified AppLovin, Carvana, Lumentum, Expand Energy, and Baker Hughes as the S&P 500 companies with the highest AI premiums. At the opposite end, Moderna, Estée Lauder Companies, ON Semiconductor, Skyworks Solutions, and Aptiv appeared as stocks investors view as potential AI losers.

Why it matters

If tokens become the standard metric for AI consumption—analogous to kilowatt-hours for electricity—they could provide economists and business leaders with real-time visibility into AI adoption patterns across industries. This represents a fundamental improvement over past technological revolutions, where researchers relied on surveys and delayed company disclosures to understand economic impact. The ability to track AI usage through token data could enable more precise policy decisions and strategic planning as the technology reshapes the economy.

Important Caveats

The research comes with significant limitations. The paper hasn't been peer-reviewed, and OpenRouter's user base likely skews toward sophisticated, cost-conscious AI power users rather than typical consumers. More fundamentally, investor expectations captured in stock prices can be spectacularly wrong—the history of financial bubbles offers ample evidence. Even if markets correctly identify AI winners, much of that optimism may already be priced in.

Still, the paper's most valuable contribution may be methodological rather than predictive. By demonstrating how token data can track AI's economic spread, it opens new research possibilities for understanding this technological shift as it unfolds.

These details were first reported by NPR's Planet Money.