What are AI Tokens, the incentive Nvidia CEO Jensen Huang said he will offer engineers worth half their salary
Nvidia CEO Jensen Huang has a new idea for luring top engineering talent—and it has nothing to do with ping pong tables or free lunches. At the company’s GTC conference in San Jose, Huang floated giving engineers an annual token budget worth roughly half their base salary, on top of their regular pay.“They’re going to make a few hundred thousand dollars a year in their base pay. I’m going to give them probably half of that on top of it as tokens so that they could be amplified 10X. Of course we would,” Huang told the packed SAP Center crowd.
Tokens are quickly becoming the new currency of AI work
So what exactly is a token? Think of it as the unit of measurement for AI computing—the way kilometres measure distance or kilowatts measure electricity. When you type a prompt into ChatGPT or Claude, the model breaks it down into small chunks called tokens. The word “unbelievable,” for instance, might be split into “un,” “believe,” and “able.” Generating 750 words of text takes roughly 1,000 tokens. More complex tasks—writing code, running AI agents for days—churn through far more.
AI companies charge based on token consumption. OpenAI, for example, prices its most advanced model at $15 per million tokens. When an engineer uses a tool like Claude Code, every line of generated code draws from their organisation’s token quota. These costs add up fast—one Vercel engineer recently racked up a $10,000 bill deploying AI agents to build a new service in a single day.
Engineers are already asking ‘How many tokens come with the job?’
Huang’s pitch isn’t coming out of nowhere. Thibault Sottiaux, engineering lead at OpenAI’s Codex, noted in February that job candidates were increasingly asking about compute access during interviews—some OpenAI employees are burning through over 10 billion tokens a week. Silicon Valley is already treating token budgets as a fourth pillar of compensation, alongside salary, bonus, and equity.“It is now one of the recruiting tools in Silicon Valley: ‘How many tokens comes along with my job?'” Huang said, adding the answer is obvious—engineers with more tokens are simply more productive.
Token costs are rising as AI use grows across companies
Nvidia, naturally, has skin in this game—its chips are what produce those tokens at scale. But the broader trend is real. Companies like Zapier and Kumo AI, per a WSJ report, are already tracking token use per employee, hunting for wasteful patterns or, on the flip side, star performers squeezing outsized output from every prompt.Token budgets as compensation may sound novel today. Within a few years, leaving them off a job offer could be the thing that sounds strange.
