The most important indicator: "Token count" in the AI era = "User count" in the Internet era
Multiple top investment institutions view "Token count" as one of the core growth indicators in the AI era, similar to "user count" in the internet era. Coatue Management has incorporated AI Token growth into its investment decisions, considering it more important than tariffs. Token growth has become an important battleground in the AI competition, with Google's monthly inference volume rising to 480 trillion Tokens, a 50-fold increase within a year, achieving a scale advantage of 6 times over Microsoft
In the competition to assess the commercialization of AI, a key metric is being pushed to the forefront: Token. This metric is akin to "user numbers" in the internet era, becoming the core measure of the actual adoption scale and growth potential of AI companies.
On June 19, during a speech at the Economic Club of New York, Philippe Laffont, founder of Coatue Management, which manages $54 billion in assets, referred to AI Token as the "atomic unit of AI" and viewed the growth of AI Tokens as a more significant market influence than the U.S. deficit and tariffs, with "Token > Tariffs" becoming a major investment logic.
Tokens are the smallest units of text processed by AI. AI foundational models like OpenAI's GPT-4.1 process information by breaking down text into Tokens, with each Token typically containing only a few letters rather than complete words. AI model manufacturers usually charge developers based on the number of Tokens.
The growth of Tokens has also become an important battleground in the AI competition among tech giants, who are competing to disclose Token data. A previous article from Wall Street Journal cited a Barclays research report showing that Google's monthly inference volume has soared to 480 trillion Tokens, a 50-fold increase within a year.
Google has demonstrated an overwhelming advantage in AI inference traffic. In the first quarter of 2025, the total number of AI inference Tokens processed by Google reached approximately 634 trillion (634T), while Microsoft's processing volume during the same period was about 100 trillion (100T). The growth momentum is even more astonishing: as of April 2025, Google's monthly inference volume has surged to 480 trillion, a 50-fold increase from 9.7 trillion a year ago. This means that Google's current AI inference scale is about six times that of the combined Azure and ChatGPT from Microsoft.
The importance of the Token metric is fully reflected in corporate earnings reports. According to AlphaSense AI research, the number of corporate earnings call and presentation mentions of AI Tokens has more than doubled compared to the same period last year.
A Major Logic Behind Investment Bets on Token Growth
Brad Gerstner, CEO of Altimeter Capital, a supporter of OpenAI, viewed Google's "parabolic" Token growth as a bullish signal for AI adoption during a CNBC program.
Coatue Management has applied the Token growth theory to its significant investment decisions. The company informed attendees at the annual East Meets West conference in Montecito that Token growth is leading to a "new inflection point in computing demand," benefiting the stock prices of companies like Broadcom and TSMC.
AI cloud service provider Together AI has processed 24 trillion Tokens per year this month, a sixfold increase compared to the third quarter of last year. Coatue has invested in this company while also betting on chip giant NVIDIA, cloud computing newcomer CoreWeave, and OpenAI The Information reported that despite the limitations of Token analysis—huge differences in Token prices across different models and a significant decline in overall Token prices over the past year—AI investors seem satisfied as long as consumption and revenue continue to rise. As NVIDIA CEO Jensen Huang stated, Tokens are becoming a new commodity driving the prosperity of AI, as important as electricity.
Google's "Token Empire": The Cost Truth Behind Scale Advantages
Wall Street Insight cites Barclays analysis showing that Google exhibits overwhelming advantages in AI inference traffic. Data indicates that by April 2025, Google's monthly inference volume will have surged to 480 trillion Tokens, a 50-fold increase from 9.7 trillion Tokens a year ago.
Even more surprising is the cost control capability. According to Barclays' estimates, based on the Gemini 2.5 rate, the inference cost for Q1 2025 is about $750 million, accounting for only 1% of search revenue. Even if Token prices continue to rise at a rate of 4 times, the cost proportion will still be far below the 18% of core infrastructure costs for search.
The key catalyst is not paid subscriptions, but free scenarios like AI Overviews in Search. The user base of Google Search is about 5-6 times that of ChatGPT, and the growth rate of free AI Tokens (50 times) far exceeds the growth rate of paid large model revenue (3-4 times).
