For Korea's so-called 386 generation, "tokens" meant the coin-shaped bus fare pieces dropped into fare boxes. But in the artificial intelligence era, the word carries an entirely different meaning. Today, a token is the smallest unit through which large language models (LLMs) like ChatGPT process text — and a form of computing resource that consumes massive semiconductor processing power and electricity.
The compensation systems symbolizing Silicon Valley's wealth have evolved from generous base salaries to stock options promising jackpot dreams, and then to restricted stock units (RSUs) offering guaranteed rewards. Recently, however, an entirely new form of compensation has emerged at the heart of global tech innovation: AI tokens.
"Fall Behind If You Don't Use AI Tokens" — Jensen Huang's New Talent Theory
The person who ignited this unusual debate is Jensen Huang, CEO of Nvidia. He raised a provocative idea at the company's recent annual developer conference, GTC.
"If an engineer earning $500,000 (approximately 700 million won) is not consuming at least $250,000 worth of AI tokens, that is concerning," Huang said. "We should also consider providing engineers with tokens worth about half their base salary on top of their pay."
His reason for emphasizing computing resources rather than cash or stock is clear. The nature of work has shifted from "writing code directly" to "directing AI." As illustrated by the recently spotlighted open-source AI assistant OpenClaw, AI now spawns its own sub-agents to write code and fix bugs even while users sleep.
The top talent companies now seek is no longer the developer who types fastest. It is the "conductor" who manages numerous AI agents and leverages millions of tokens per day to produce overwhelming results. According to The New York Times, engineers at Meta and OpenAI compete on internal leaderboards over who consumes the most tokens. Token usage has become the badge that proves skill and productivity.
Tokens Worth Hundreds of Millions — Reward or Pressure?
Behind Huang's enticing proposal, however, lies capitalism's cold calculus. An engineer at Ericsson, the Swedish telecom equipment maker, said, "The cost of the AI I use — Claude — could exceed my salary."
Why would a company willingly hand out AI resources more expensive than a human's salary? The answer is simple. If a developer earning 300 million won receives 300 million won worth of AI tokens, the company will demand not the output of two people, but 20 — or even 200.
It amounts to an unspoken ultimatum: "We'll buy you the best equipment, so go build a skyscraper by yourself." There are rosy forecasts — like OpenAI CEO Sam Altman's vision of "Universal Basic Compute" — suggesting computing power will become the currency of the future. But from management's perspective, the question has shifted from "How much do we pay a person?" to "How much computing can we attach to reduce headcount?"
A Consumable Resource, Not Cash or Equity — Is the Labor Market Shifting?
The most fundamental warning concerns the quality of compensation. Silicon Valley venture capitalist Jamal Glenn sharply criticized AI token packaging as "a gimmick that inflates the appearance of compensation without cash or equity."
We receive stock or stock options to share in the fruits of a company's growth. Over time, asset values rise and serve as proof of our market worth when changing jobs. AI tokens, however, are gone once consumed. Whether you use one million or 10 million tokens a month, the amount deposited in your bank account is zero.
Companies gain a perfect justification to freeze salary increases by declaring, "We are investing hundreds of millions of won worth of computing resources in each employee." Meanwhile, a chilling question will surface on the CFO's spreadsheet: "If the AI token cost per employee has surpassed the employee's salary, wouldn't it be better to cut staff and buy more tokens with the savings?"
Howard Marks, chairman of Oaktree Capital, has noted that AI will replace a substantial portion of knowledge work. Goldman Sachs has also warned that 25% of U.S. labor hours could be automated.
Ultimately, this "token salary" debate that began in Silicon Valley suggests the history of human labor has crossed a river of no return.
In the not-so-distant future, what lands on your salary negotiation table may not be cash or stock but access rights to top-tier AI servers. An era is approaching in which an individual's market value will be determined by how effectively they can control and wield AI computing resources.
Will we become masters who command AI, or will we be reduced to "biological components" temporarily hired to keep massive AI systems running? At this great inflection point in the technology paradigm, it is time for serious reflection.
A bonus of "300 million won" — but not in cash? The truth behind the "token bonus" that has turned Silicon Valley upside down.
