Nvidia CEO Jensen Huang proposes AI tokens as engineering pay

Earlier this week at Nvidia’s annual GTC event, CEO Jensen Huang floated the notion that engineers should receive roughly half their base salary in AI tokens. According to Huang, top engineers might burn through $250,000 a year in AI compute. By this weekend, The New York Times reported on a trend called “tokenmaxxing”—where engineers compete on internal leaderboards that track token consumption. The publication reported that engineers at companies including Meta and OpenAI are participating in these competitions. While a person writing an essay might use 10,000 tokens in an afternoon, an engineer running a swarm of agents can consume millions of tokens in a single day.

This shift follows a broader discussion initiated in mid-February by Tomasz Tunguz, a venture capitalist at Theory Ventures. Tunguz reported that tech startups in the Bay Area were already adding inference costs—the costs associated with running AI models—as a fourth component to engineering compensation. Using data from the compensation tracking site Levels.fyi, Tunguz analyzed a top-quartile software engineer salary of $375,000. Adding $100,000 in token compensation brings the fully loaded compensation to $475,000, meaning compute represents roughly one-fifth of the total package. This conversation has accelerated since the late January release of OpenClaw, an open-source AI assistant designed to run continuously—churning through tasks and spawning sub-agents. OpenClaw uses agentic AI, which refers to AI systems that take sequences of actions autonomously over time.

The trend is visible beyond Silicon Valley. In Stockholm, an engineer at Ericsson reportedly spends more on Claude than his salary, with his employer covering the cost. However, tokens may become the fourth pillar of engineering compensation, a shift that presents risks for tech workers. While a large token budget can boost short-term productivity, it lacks the compounding value of cash or equity. On the East Coast, Jamaal Glenn points out that token budgets do not vest or appreciate. If companies successfully normalize tokens as a form of pay, they could keep base salaries flat while pointing to compute allowances as compensation, effectively shifting financial risk to employees.

Why it matters

The normalization of AI tokens as a component of engineering compensation could fundamentally alter the financial logic of headcount and job security. Companies may use compute allowances to keep cash compensation flat, shifting the balance of tech compensation from compounding assets to temporary operational utility.