Silicon Valley is flirting with a new form of pay: AI tokens—budgets for compute that let engineers run agents and crank out code faster. Nvidia CEO Jensen Huang fanned the flames by suggesting top engineers could warrant token allotments equal to roughly half their base salaries, reframing compute as a recruitment tool and performance lever. Proponents argue access to more tokens boosts productivity; skeptics note the strings attached, including higher output expectations and the risk that rising compute spend shifts management thinking from adding headcount to automating tasks. VCs and finance chiefs caution that token budgets don’t vest or appreciate, making them an easy way to pad offers without raising cash or equity. As agentic AI proliferates and token consumption soars, the economics of engineering compensation—and who, or what, does the work—are rapidly being renegotiated.
Related articles:
NVIDIA GTC keynote: Why AI compute budgets are exploding
Amazon Bedrock pricing: the economics of generative AI in production
