This week, the idea that engineers could receive half their salary in AI tokens—computational units for tools like Claude, ChatGPT, and Gemini—has sparked debate. CEO Jensen Huang of Nvidia floated this at GTC, suggesting it as a recruiting tool.

The concept isn’t new; Tomasz Tunguz from Theory Ventures has been discussing token compensation since February. Using data from Levels.fyi, he proposed adding $100,000 in tokens to a top-quartile engineer's salary, bringing the total to $475,000.

The practical consequence is clear: agentic AI has taken off, with tools running continuously and consuming vast amounts of compute. Engineers are competing on internal leaderboards tracking token consumption, and generous budgets have become standard perks.

However, engineers should proceed with caution. More tokens mean more power but also higher expectations. If a company funds an engineer’s equivalent of a second engineer's worth of compute, the pressure to produce at twice the rate could be intense.

The real issue, according to financial services CFO Jamaal Glenn, is that this can lead to a shift in how companies view headcount costs. As token spend per employee approaches or exceeds salary, questions about human relevance arise. Tokens might seem like a perk but could actually make it easier for companies to freeze cash compensation while pointing to increased compute as an investment.