A new metric called "tokenmaxxing" is dividing Silicon Valley, as software developers race to spend computing tokens on AI tools to climb internal company leaderboards. The practice sees engineers competing for titles like "Token Legend" based on their consumption of tokens, the unit of computing that determines how AI work is priced.
The trend, reported by The Information, is particularly visible at companies like Meta, where an employee-made "Claudeonomics" dashboard tracks usage. While some executives champion it as a sign of embracing new technology, critics argue it incentivises performative and inefficient use of expensive AI resources within corporations.
The Token as a New Currency
To understand tokenmaxxing, one must first understand the token. Large language models break words into numerical inputs, with each token representing roughly three-quarters of a word. AI models charge based on the number of tokens processed, making token spend a direct measure of AI usage and cost.
Tokenmaxxing, therefore, is the drive to maximise this spend. Companies including Meta and OpenAI have implemented token leaderboards, The New York Times has previously reported. The fintech company Ramp cited Gartner data on X, calling rising AI expenditure a "$1 trillion blind spot," noting that monthly business spending on AI has quadrupled in the past year.
A Culture of Competition and Critique
For some in tech, high token spending is a badge of honour. Y Combinator CEO Garry Tan wrote on X: "We've been tokenmaxxing longer than most people," quoting a post that criticised companies being "stingy" with tokens. An xAI employee lamented that the industry turns every good idea "into theatre," while another user posted: "i personally spend thousands of dollars on tokens every week... feels insane but i can't stop tokenmaxxing."
However, significant criticism has emerged. Linear COO Cristina Cordova compared ranking engineers by token spend to ranking a marketing team by budget burn, stating: "Don't mistake a high burn rate for a high success rate."
Khosla Ventures partner Jon Chu labelled it an "absolutely stupid policy," claiming on X that "plenty of my Meta friends told me folks have been building bots that just run in a loop burning tokens as fast as they can due to this policy."
Nuanced Views and Executive Endorsement
Other voices offered more measured analysis. Cursor employee Edwin Wee Arbus called token spend a "useful, fast proxy, but slightly flawed," comparing it to Body Mass Index (BMI) for health. Persona software engineer Arush Shankar noted, "Token spend is always an output not an input... It's a signal but not THE signal."
Nvidia CEO Jensen Huang has stressed the importance of engineers using tokens, stating he would be "deeply alarmed" if a $500,000 engineer did not consume at least $250,000 worth. In contrast, "The Pragmatic Engineer" newsletter author Gergely Orosz called the practice wasteful, writing: "Devs game everything and anything seen as a target for more bonus or promos. This was no different."
The Broader Implications
The debate highlights a fundamental shift in how productivity is measured in the AI era. BEP Research founder Ben Pouladian suggested on X that "compute [is] the bottleneck for innovation," adding that "in the AI era, every employee becomes a compute consumer."
As companies grapple with soaring AI costs, the tokenmaxxing phenomenon raises critical questions about whether tracking resource consumption drives genuine innovation or merely encourages a new form of corporate gaming, with significant financial implications for the tech industry.